Part 1 — What is Squid Hunt Progressive

What this book covers

This book is a strategy manual for Squid Hunt Progressive — one of four variants in the Squid family of NLHE-with-tokens poker games. Hunt Progressive is the variant trained on the QuintAce model under the codename SquidType::DOUBLE. This primer establishes the ruleset; the rest of the book is what the trained model has learned about playing it.

Where Squid Hunt Progressive sits in the Squid family

Four variants share the Squid backbone: every main pot winner collects a "squid" token, and at game end, players without squids pay a chip penalty. The variants differ in how a squid count translates into chip payout:

In the canonical Squid-family pedagogical order — Stand-up → Hunt Regular → Hunt Progressive → Blood Battle — this book is the third step. Hunt Progressive is where the squid game stops being simple. The first two variants give you a single threshold (have a squid, or don't) or a flat slope (every squid is +1). Hunt Progressive introduces discrete jackpots: cross s = 3, every pot suddenly pays 2× weight; cross s = 5, every pot pays 4×. The strategy reorganizes around chasing those crossings.

For a deeper comparison of all four shapes, see the Squid Family Variants Primer.

The Squid Hunt Progressive ruleset

Hunt Progressive is 6-max NLHE with two rules layered on.

Rule 1 — Squids accumulate. Every time you win a main pot, you collect a squid (a win token). There's no per-player cap. A player who wins five pots holds five squids.

Rule 2 — Squids translate to chips through a tiered multiplier curve at game end. The weight function is:

weight(s) = s              if s < 3        (Tier 1, 1× multiplier)
weight(s) = 2 × s          if 3 ≤ s < 5    (Tier 2, 2× multiplier)
weight(s) = 4 × s          if s ≥ 5        (Tier 3, 4× multiplier — capped)

Two discrete cliffs: at s = 3 the multiplier doubles, at s = 5 it doubles again. Past s = 5 there are no more cliffs — every additional squid is worth a flat +4 weight units.

The 4 → 5 jump is +12 weight units — the largest single jump in any Squid variant, including Blood Battle's +9 peak. This is what makes Hunt Progressive distinct: the strategy doesn't follow a smooth curve; it organizes around two sharp threshold crossings.

The intuition this curve invites is "race to the cliffs — the next squid is dramatically more valuable when it crosses you into a new tier, and barely matters between tiers." That intuition is approximately right, with one important nuance: at s = 4 (one pot from the +12 cliff) the model plays most aggressively in the entire game; at s = 3 (just past the +4 cliff) it plays wider but more passive — limp-heavy participation, saving aggression for the next cliff. The chapters in this book unpack why.

What's the same as Stand-up Game

The basic frame carries over. You're playing 6-max NLHE with 100bb stacks, standard blinds, all the usual positions. Squids are awarded only to the main pot winner — split pots and side pots award no squid. The val parameter scales how much each squid is worth in chips, trained on five settings: 1, 2, 3, 5, and 10 bb. The reward formula is still chip-EV plus the change in your forward-looking squid value, zero-sum at settlement.

If you've read Stand-up Book 2, the rule frame is familiar. The strategy is not.

What's different from Stand-up — three rule changes

1. No per-player cap. Stand-up caps you at 1 squid. Hunt Progressive lets squids stack. A player who wins five pots has five squids and a 4× multiplier on every additional pot.
2. More squids in the pool. Total squids handed out per game is N + 4 — that's 10 in 6-max (Stand-up has N − 1 = 5). The pool is bigger to allow accumulation.
3. A tiered multiplier on the squid count. Stand-up's weight is binary (0 or 1). Hunt Progressive's weight has two discrete jumps at s = 3 and s = 5.

How games end

The game ends at the FIRST of two conditions:

• Z = 1 — only one player has zero squids (Z = 1 trigger, same as Stand-up and Blood Battle)
• t = T — all N + 4 = 10 squids have been distributed (then any players still at 0 pay the loser penalty)

The Z = 1 trigger means you'll always be playing in a state where at least two players still have zero squids (Z ≥ 2). The strategic question is how many — and where in the gradient from "everyone still desperate" (Z = 6, fresh state) to "only two left" (Z = 2, the cliff before the game ends) the table currently sits.

The trained model reflects this: it has only learned to play states with Z ≥ 2. It was never trained on Z = 1 (the terminal state) or Z = 0 (which never arises because the game ends before it can). When this book talks about strategy at "the start of the game" or "near the end," it means the gradient from Z = 6 down to Z = 2.

What's at stake — the payout math

At settlement, each holder receives val × weight(s) × Z (Z = the number of zero-squid players at termination). Each loser pays val × sum_of_holder_weights.

Worked example, val = 2 (reproduces AceSense product spec §III.1 verbatim). 4 players. Terminal state: A has 3 squids (Tier 2), B has 5 squids (Tier 3), C and D have 0 squids each.

• Z = 2 (two losers: C and D)
• weight(A = 3) = 2 × 3 = 6 (Tier 2)
• weight(B = 5) = 4 × 5 = 20 (Tier 3)
• Sum of holder weights: 6 + 20 = 26
• A receives: 2 × 6 × 2 = 24 chips
• B receives: 2 × 20 × 2 = 80 chips
• Each loser pays: 2 × 26 = 52 chips
• Sums to zero: 24 + 80 − 52 − 52 = 0 ✓

Compare to Blood Battle at val = 2 in the same setup. Blood Battle weight: A = min(5, 3) × 3 = 9, B = min(5, 5) × 5 = 25. Loser pays 2 × (9 + 25) = 68. Hunt Progressive's loser pays 52 chips — about 24% less than Blood Battle's 68. The cliff structure produces lower total weight at this state because Hunt Progressive multiplies by tier (2× and 4×), while Blood Battle multiplies by min(5, s) (which equals 5 for any s ≥ 5).

Where Hunt Progressive's stakes get bigger than Blood Battle's is when one player rides the 4× tier far. Same val = 2, terminal state with one player at 8 squids (other terms equal): Hunt Progressive weight = 4 × 8 = 32. Blood Battle weight = 5 × 8 = 40 — actually still bigger. The shapes are close enough at high s that magnitude isn't the differentiator. The strategy differentiator is the cliff structure, not the absolute payout.

Two strategic axes

Strategy in Hunt Progressive depends on two axes — both well-tested across the in-distribution Z = 2 to Z = 6 range.

Axis 1 — The co-desperate count (Z). How many players still have zero squids? Z = 6 means everyone (fresh state at the start of a game). Z = 2 means only two — the game is one squid from ending. As Z decreases, the implicit ante per pot grows — fewer desperates left to share the eventual penalty, and the game is closer to ending. Strategy shifts gradually: hero plays wider, opponents adjust. Continuous, not stepped — the model treats the gradient smoothly across Z = 2 through Z = 6.

Axis 2 — Hero's own tier — three weight-function regimes, with within-tier nuance. The weight function has three tiers (1× / 2× / 4×). Each tier has internal structure that Part 6 unpacks; the high-level picture:

• 1× tier (s = 0, 1, 2) — pre-first-cliff. Hero is at the same per-pot weight as Stand-up Game. Three sub-states with very different incentives:
• s = 0 (desperate): two motives stack — removing lone-loser risk (insurance) AND starting the race toward the +4 cliff at s = 3 (head start). Maximum participation, even at terrible chip-EV.
• s = 1 (dead zone): post-head-start, but no immediate cliff in sight. The lowest-incentive state in the game — hero coasts. Don't expect aggression here the way you would from s = 0 or s = 2.
• s = 2 (cliff-1-imminent): one pot from the +4 conversion. Activity rises again.
• 2× tier (s = 3, 4) — between cliffs. Hero just crossed cliff 1 and now plays at 2× per-pot weight.
• s = 3 (post-cliff-1): limp-heavy participation, coasting as ammunition for cliff 2.
• s = 4 (cliff-2-imminent): peak aggression in the entire game. One pot from the +12 conversion at s = 5 — raise-heavy, bigger sizings.
• 4× tier (s ≥ 5) — past-peak. Hero captured both cliffs. Each additional squid is worth a flat +4 — a moderate fixed bonus per future pot won. Most selective; close to chip-EV with a small accumulation bonus.

The cliff signature: between cliffs, hero coasts; right before a cliff, hero pushes.

The two axes interact: a pre-cliff (s = 2) hero plays differently when six players are still desperate (fresh state) versus when only two are (one squid from game-end). The chapters of this book walk through the interaction across spots — preflop opens, BB defense, flop c-bet, etc.

Terminology for Hunt Progressive

Language to avoid:

• "Hero-has" without a count — ambiguous in Hunt Progressive. Prefer the tier name or s = N.
• Treating Hunt Progressive as "Stand-up but bigger numbers" — the cliff structure produces qualitatively different strategic incentives. The 1× tier is like Stand-up; the 2× and 4× tiers are different games entirely.
• Treating Hunt Progressive as Blood Battle. The shapes are close at low s but the cliff structure means s = 2, s = 4, and s = 5 behave categorically differently from Blood Battle's smooth ramp at the same states.

Two known implementation gaps

The AceSense product spec defines two extras for Hunt Progressive that the trained model does not know about. Both are more impactful in Hunt Progressive than in Blood Battle because they interact with the cliff thresholds.

• Super Squid — a randomly selected pot during the game gives a 2× squid (counts as 2 instead of 1). In Hunt Progressive this can jump a player from s = 2 (Tier 1, weight = 2) to s = 4 (Tier 2, weight = 8) in a single hand — a much larger discrete jump than the same Super Squid would produce in Blood Battle. Declared in code, not implemented in training.
• First-hand double squid — the winner of the first hand of every game gets 2 squids immediately (split pots void). In Hunt Progressive this means a first-hand winner starts 2 steps from the s = 3 cliff instead of 3 steps — materially changing early-hand aggression incentives. Declared, not implemented.

Both gaps are tracked as open ENG items requiring retraining. Until closed, this book's findings are scoped to: Super Squid off, no first-hand bonus, default tier thresholds (3, 5).

What this primer does NOT cover

• Strategic findings — those are the rest of the book. This primer just establishes the rules and the tier framework.
• Squid Hunt Regular — sibling Squid-family variant with linear weight = s. Not yet trained on the model. Engineering ticket pending. Will get its own book once trained.
• Blood Battle — the smooth-quadratic Squid variant (weight = min(5, s) × s). Trained, separate scope, separate book (Book 7).
• Configurable tier thresholds — the AceSense product spec allows custom thresholds per game; the CUDA implementation hardcodes (3, 5). Research is on the default config only.
• Super Squid + first-hand double squid — not implemented in training (see above).
• Z = 0 or Z = 1 strategic states — out of scope. Z = 1 is the game's terminal state; Z = 0 never arises during play. The training sampler enforces Z ≥ 2 by construction.

Where the rules come from

The rules and multiplier table trace to the engineering team's ground-truth reference at engineering-department/gameplay-ai/projects/llm-verifier-game-expansion/squid-double/GAME-RULES.md. The folder name and the code identifier SquidType::DOUBLE use the codebase enum name; the canonical product name we use throughout this book is Squid Hunt Progressive. Rules cross-checked against the AceSense product spec §III ("Double Mode" / 翻倍模式) — perfect rule match, including the worked example reproduced above. Core mechanics match; two product features (Super Squid, first-hand double squid) are spec'd in product but not yet trained.

The strategic findings in the rest of this book are the model's learned response to these rules, queried only at in-distribution states (Z ≥ 2) per meta_seed_sampler.h:654-684.

Draft · Squid Hunt Progressive Part 1 primer · 2026-05-01 (Phase 1 — pre-pull-campaign)

Part 2 — The Implicit Ante and the Future-Loser Pool

This is the foundation chapter. Every chapter that follows cites this one. If you internalize it, the rest of the book teaches you how to apply it.

The thesis: Hunt Progressive's strategy follows from one unified mental model. Every pot pays an implicit ante — a quantity bigger than the chips in the pot, smaller than the entire game's prize pool, and ignored entirely if you're carrying NLHE intuition over. Once you can put a number on it, you can read every situation in the book.

§2.1 — Uri's swing formula: putting a number on the implicit ante

Pick a concrete scenario. You're in 6-max Hunt Progressive, val=3 (each squid weight unit is worth 3 BB at settlement), 100bb stacks. You win a pot. What's that pot actually worth?

The chips in the pot: whatever was in there. That's the visible part — same as NLHE. But you also picked up a squid, which converts to chips at game-end. How much, in BB?

Uri's swing formula tells you. In a state where N players are still desperate (zero squids, eligible to be the lone-loser):

swing per pot = 6 × val / (N − 1)

The "swing" is what changes between winning this pot and not winning it — what you give up if someone else takes the pot instead of you. (Not "how much the squid is worth in expectation right now" — that's a different number that doesn't drive the decision.)

Plug in the table state. Fresh game, all 6 desperate (N=6), val=3:

swing = 6 × 3 / 5 = 3.6 BB

Every pot at the start of a Hunt Progressive game has 3.6 BB of squid-equity riding on top of the chips you can see. You play a 4-bet pot for an extra 3.6 BB beyond the chips on the table. That's the implicit ante.

Now move to mid-game. Two desperates left (N=2), val=3:

swing = 6 × 3 / 1 = 18 BB

Same val, same chip stacks, same hand. The implicit ante just went 5× larger. The pot is much more valuable to win, and more valuable to deny.

This is where Hunt Progressive parts ways from NLHE. Pot odds, ICM-style adjustments, even the standard "fold equity vs hand equity" math all assume the chips are the prize. In Squid, the chips are part of the prize. The implicit ante is the rest of it.

But the formula is only one read. The implicit ante is the size of what's at stake; it doesn't tell you who pays it or what makes the size grow. §2.2 unpacks the second read — the squid economy has two sides, and reading both is the dominant strategic skill.

Cite: uri-squid-invisible-ante §1.

§2.2 — Where the squid value comes from

A trap to avoid: thinking the squid you just won was paid by your opponent.

The chips in the pot moved from one player's stack to another — that part is zero-sum, same as cash. But the squid-equity you just gained didn't come from your opponent on this hand. It came from a settlement event at game-end with two sides.

The settlement formula is literal: at game-end, each loser pays val × sum_of_holder_weights, and each holder receives val × weight(s) × Z. Two sides, two reads:

1. Who pays — the future-loser pool. The set of players still at 0 squids when the game terminates. The pool starts at 6 players, shrinks as desperates capture squids, and the game ends when only one player is left at zero. Whoever's most likely to be that player is funding the squid prize for everyone else.
2. How much they pay — the holder amplification. The loser's bill scales with the SUM of holder weights. When a holder crosses a cliff (s=2 → s=3 = +4 weight; s=4 → s=5 = +12 weight), the loser owes more, even though the pool itself didn't shrink. Cliff crossings are amplification events, not pool events — but they show up in the same line of the settlement formula.

Both reads matter. The pool side tells you who funds the prize. The amplification side tells you how big the prize gets. Hero's strategic skill is reading both halves.

Three consequences worth foregrounding:

• The opponent who pays the chips for this pot is rarely the player who pays the squid. They might be a co-holder afterward; a different bystander becomes the future-loser instead.
• A holder's cliff crossing is a transfer event between holders too — the crossing holder's share of the loser payment grows; other holders' shares shrink in relative terms (though their absolute payouts also grow because the loser's bill grew).
• Reading the future-loser pool plus reading where holder amplification is forming is the dominant strategic skill. Part 6 §6.3 walks through this for opener-side amplification (the marquee deny/commit/fold pattern); §6.5b extends it to off-pot amplification (any cliff-imminent player at the table reshapes hero's defense).

One refinement that the data forces. Hero responds to amplification potential — cliffs in front, not cliffs already crossed. When a holder is cliff-2-imminent (s=4), hero treats the upcoming +12 amplification as the dominant signal and shifts toward call-heavy ("be in the pot when it pays"). When the same holder has already crossed into the 4× tier (s=5+), the amplification is in the bag and no further cliff is coming — hero shifts back toward chip-pressure mode. Phase 8 confirms this directly (Slice 18d, in §6.5b): same Z, same opener, but a holder past s=5 produces opposite play vs holders one cliff away. Amplification potential, not current amplification, drives the strategic adjustment.

Cite: uri-squid-invisible-ante §6.

§2.3 — The "less zero-sum" corollary

This is where the math has practical consequences.

In cash NLHE, every BB you win comes from a BB someone else loses on this hand. Per-hand zero-sum, no exceptions. So when you bluff, you're trying to make a calling station fold — because that's the only way a bluff has value in a per-hand zero-sum world.

In Hunt Progressive, the chip transfer is still per-hand zero-sum (the pot conserves chips). But the squid-equity transfer is not. The squid you collect comes from a settlement event with two sides separate from the chips in this pot — the loser pool funds it, holder amplification sizes it. Your bluff can have positive squid-equity even if your opponent never folds, because the squid is paid by someone else (and the size of that payment scales with how many cliffs the holders cross between now and game-end).

This is what Uri means by "less zero-sum than NLHE": per-hand chip transfers are zero-sum, but per-hand squid-equity transfers are not.

The practical consequence is the result that confuses NLHE players the most: in a Hunt Progressive multiway pot, multiple players can have profitable contradictory plays simultaneously. A desperate player can profitably defend (they want any squid). A cliff-imminent player can profitably call wide (their +4 or +12 cliff conversion is huge — and amplifies what every loser pays). A safe mid-tier player can profitably bluff (their bluff equity comes from the settlement event, not from the caller's stack). All three of those plays could be the right answer at the same table at the same moment — because each player's squid-equity is sourced from the same settlement event but capturing different parts of it.

The cash reflex — "deny equity from the calling station" — is partially wrong here. You don't need the caller to fold for your bluff to make money. You need the future-loser pool + amplification structure to favor you more after this pot than before. Sometimes those two conditions overlap. Sometimes they don't.

Cite: uri-squid-invisible-ante §6.

§2.4 — Hunt Progressive extension: the swing scales with hero's tier

Uri's formula was written for Stand-up Game, where every pot win adds +1 weight unit. In Hunt Progressive, that's only true at Tier 1 (s < 3). Past the cliffs, the per-pot weight changes.

Read this carefully. At Tier 1 the implicit ante matches Uri's Stand-up formula exactly — Hunt Progressive at s < 3 plays like Stand-up plus a forward jackpot. The cliffs are when the formula spikes: at s = 4 with two desperates left and val=3, winning this pot is worth 216 BB of squid-equity, because winning it crosses you into the 4× tier and re-mints all your accumulated squids at the new multiplier.

That's why the data shows hero=s=4 as the most aggressive state in the entire game (Part 3, Part 4). The implicit ante there is a mortgage payment, not a tip.

Two things this table tells you:

1. The cliffs aren't just "more weight." They're discrete jumps in the implicit ante. Hero's per-pot stake jumps 12× at cliff 2, 4× at cliff 1.
2. The implicit ante grows with hero's accumulated stash. Past cliff 2 you're at +4 weight per pot — four times Stand-up's baseline — for as many pots as the game has left.

This is the core asymmetry between Hunt Progressive and Blood Battle. Blood Battle's smooth quadratic ramp produces a continuously growing per-pot stake. Hunt Progressive's tier structure produces step changes — abrupt reorganizations of the implicit ante at s = 3 and s = 5. The strategy reorganizes in steps too.

§2.5 — What this mental model predicts

If the mental model is right, three things should fall out of the data.

Prediction 1 — Hero's defense varies with opponent's tier. Why hero defends differently against a cliff-1-imminent opener than against a cliff-2-imminent one isn't about the chips in this pot. It's about how much the loser payment will inflate if the opener crosses their cliff (the holder-amplification side from §2.2). Part 6 walks through this — at hero=BB s=3, BB raises 80% vs cliff-1-imminent CO (small amplification, deny it cheaply) and calls 84% vs cliff-2-imminent CO (huge amplification — +12 weight per remaining loser — too valuable to miss; just be in the pot when it pays). The math comes out of the table above.

Prediction 2 — The cliffs are decision-reshape moments, not gradual. The implicit ante jumps discretely at s = 3 and s = 5, so the strategy should jump too. Part 3 confirms — hero's raise rate at val=3 climbs gradually through s=0 to s=4, then peaks at the cliff-2-imminent state, then drops post-cliff.

Prediction 3 — Within-tier states have different incentives even at the same multiplier. At s = 0 hero is in Tier 1 and has a +1 marginal squid coming up. At s = 1 hero is in the same tier, same marginal weight — but plays much more passively in the data, because the next cliff is now two pots away rather than directly in front. The mental model gives you the multiplier; the position relative to the next cliff gives you the urgency. Part 4 unpacks the seven states.

The whole book is the trained model's answer to "what does this mental model imply at every spot?" The framework predicts; the data confirms (or surprises us, in which case we treat it as an editor's question).

§2.6 — What the mental model does NOT predict

The framework gives you a number for the implicit ante and a frame for reading both halves of the squid economy (pool + amplification). It doesn't directly predict everything.

It doesn't tell you why s = 1 is the most passive state in the game. The marginal weight at s = 1 is +1 — same as s = 0 and s = 2. The mental model says all three should play similarly. The data says s = 0 plays max-participation (insurance + head start), s = 2 plays cliff-1-imminent aggression, and s = 1 sits in the middle as a low-incentive trough. Part 4 gives the explanation: it's about position relative to the next cliff, which the formula doesn't capture directly.

It doesn't tell you why hero=s=3 is the peak responsiveness state for the deny/commit/fold pattern. The Goldilocks finding (Part 6 §6.3) emerged from the data. The mental model is consistent with it after the fact — at s=3, hero's own forward stake is moderate, letting CO's state do the steering — but the framework alone wouldn't have predicted s=3 sharply over s=2 or s=4.

It doesn't tell you about val-conditioning. At val=10 the Goldilocks pattern moves from hero=s=3 to hero=s=1. The mental model is about the direction of strategic response; the trained model adds val-specific shape that the formula alone doesn't capture. The val-stability audit walks through which findings survive across vals and which are val=3-specific.

The right way to use this chapter: read the mental model as the framing, then let the data refine your expectations. When the data matches the model, you've earned a heuristic. When the data surprises the model, you've found something the framework doesn't yet capture — and we flag it as an editor's question.

§2.7 — How to read the data in this book

Each chapter section opens with a one-line setup: what's held, what varies. For example, §6.3's setup says "hero=BB at 3 squids throughout this section; Z=2 fixed; opener's squid count varies." Read that setup line first. Anything not mentioned is at the book's default: hero state varies, table held at Z=3 fixed, val=3, 100bb stacks. Anything else (varying Z, val, stacks, opp tier) gets called out explicitly.

The reason for this discipline: when comparing two solver outputs, multiple things can vary at once. The setup line tells you what's been held constant so the comparison is apples-to-apples.

The full methodology framework — the formal naming for each "what varies, what's held" paradigm, the cleanliness audit of every pull, the val-stability audit — lives in the methodology appendix at the end of the book for readers who want the rigor. Most readers don't need it. The setup line is enough.

Reader takeaway from Part 2. Every pot in Hunt Progressive carries an implicit ante given by 6 × val / (N − 1) times hero's per-pot marginal weight. The squid value comes from a settlement event with two sides: the future-loser pool funds it (who pays) and holder amplification sizes it (how much each loser pays). Reading both halves is the dominant strategic skill — and the data shows hero responds to amplification potential (cliffs in front), not current amplification (cliffs already crossed). That makes Hunt Progressive less zero-sum than NLHE in a per-hand sense. Every chapter from here is the trained model's answer to "what does this framing imply at this specific spot?"

Draft · Squid Hunt Progressive Part 2 · Pass 1 · 2026-05-02 · v0.1.0

Part 3 — Three Lines to Cross: The Safety Line and Two Cliffs

Part 2 told you the implicit ante per pot scales with hero's tier. This chapter shows what the data says about the moments when that scaling jumps. There are three of them in Hunt Progressive — and they do not look alike.

§3.1 — Three transitions, three different mechanisms

The strategy organizes around three crossings:

The safety line is qualitatively different from the cliffs. Crossing s = 0 → s = 1 doesn't change your weight multiplier (still 1×) — it changes your status (from "candidate lone-loser" to "no longer at risk"). The cliffs are quantitative — they change how every existing squid converts to chips.

Three crossings, three different mechanisms, three different shapes in the data.

The canonical sequence — hero state varies, table held at Z=3, val=3, CO opens (Pull 13 — paradigm-clean H-Z):

Read top to bottom. Three things jump at you: the dip at s = 1, the climb to a peak at s = 4, the drop after s = 5. Three transitions, three mechanisms. Walk through them in order.

§3.2 — The safety line (s = 0 → s = 1) — status, not multiplier

You're at zero squids. You are the only player who can become the lone-loser if the game ends right now. Capturing your first squid removes that risk entirely. You become "safe" — no longer in the future-loser pool — and you've also taken your first step toward the +4 cliff at s = 3.

Two motives stack at s = 0: insurance against being the lone-loser, plus the head start to cliff 1. The mental model from Part 2 says hero should play max-participation here, and the data agrees — VPIP 92.6%, limp 88.9%, almost any pot is worth contesting. ⚠️ Val-conditioning: the limp-pure shape (raise% under 1%) is val=3-specific. At val=1 and val=10, hero=s=0 raises 37–44% (Phase 5 Slice 16). The max-participation finding holds across vals; the limp-pure character is the val=3 form.

Crossing the safety line removes one of the two motives — the insurance is paid. Only the head start motive remains, and the next cliff is now two pots away. So at s = 1 hero plays a different game: VPIP drops to 79.3%, raise rate collapses to 0.1%, sizing stays at 7.1 bb (when hero raises at all). This is the dead zone — the post-safety-line plateau, the lowest-incentive state in Hunt Progressive. The mental model alone wouldn't have predicted it; Part 4 unpacks why the data lands here.

The safety line's signature in the data is the VPIP drop at s = 1, not a raise spike. Hero participates less because half the reason to play just got banked.

§3.3 — Cliff 1 (s = 2 → s = 3) — the +4 conversion

s = 2 is one pot from the +4 conversion to the 2× tier. In Part 2's framework: at val=3 with two desperates left, the implicit ante for crossing this cliff is 12 BB of squid-equity per pot (6 × val / (N − 1) × 4 — the formula multiplied by the cliff conversion). That's a meaningful chunk on top of the chips, but it's not catastrophic if hero misses this pot.

So the cliff-1-imminent shape is wider, not sharper. VPIP rises to 93.8% (back up from the dead zone), but raise rate stays low at 6.9%. Limp 86.9%, sizing 7.2 bb. Hero plays max-participation again — now with the cliff motive replacing the insurance motive — but the cliff is small enough that cheap wide participation captures it better than aggression. Take any pot, don't commit chips.

After the cliff (s = 3, post-cliff-1), hero's per-pot weight is now 2× — the implicit ante doubled. The shape changes: VPIP 96.1% (slightly higher), limp drops sharply (87% → 70%), raise climbs (7% → 26%). Hero starts converting wide passive participation into aggressive participation. The "coasting" framing applies — you're stockpiling for cliff 2 — but the per-pot weight is high enough that you start playing some pots harder.

§3.4 — Cliff 2 (s = 4 → s = 5) — the +12 conversion

s = 4 is where Hunt Progressive shows its teeth.

You're one pot from the +12 conversion to the 4× tier. Plug into Part 2's formula at val=3 with two desperates left: the implicit ante for crossing this cliff is 72 BB of squid-equity per pot — three times the implicit ante for cliff 1, six times the Stand-up baseline. This is the largest discrete strategic event in any Squid variant. Bigger than Blood Battle's smooth-quadratic peak. Bigger than anything in Stand-up or Squid Hunt Regular.

The data reflects this. At s = 4, Z=3, val=3:

• Raise 48.5% — the highest raise rate in the entire seven-state sequence.
• VPIP 95.6% — still essentially every pot, but now a near-coin-flip between limp and raise.
• Sizing 5.7 bb — smaller than the wide-limp tiers above (where hero raises rarely with size 7+ bb), because the goal is to get to a flop with leverage, not to fold opponents out.

The pattern flips from cliff 1's "wide and cheap" to cliff 2's "wide and aggressive." Hero's per-pot stake is now so big — 72 BB at val=3 N=2, and growing as Z drops further — that hero can't afford to play passively. You raise to put leverage on the pot, then bet through it on the flop.

✅ Val-stable in direction. Phase 5 Slice 16 confirms the s = 4 raise peak at val=1 (47.3%), val=3 (48.5%), and val=10 (65.8%). Magnitude grows with val; direction is robust. The +12 cliff matters at every val we tested — it's the structural feature, not a val=3 artifact.

After the cliff (s = 5, past-peak): raise drops to 37.6%, sizing drops to 4.6 bb, VPIP drops to 86.2%. The implicit ante per pot is still high (4× weight every pot), but no more cliffs to chase. Hero's posture downshifts from "race to the next cliff" to "accumulate at 4× weight." Different gear.

§3.5 — Post-transition plateaus

Each transition has a distinctive after state. Three plateaus:

s = 1 (post-safety, dead zone). Covered above and unpacked in Part 4. The strange one — strategy DROPS into a passive trough rather than continuing forward.

s = 3 (post-cliff-1, coasting toward cliff 2). VPIP very high (96%), limp drops (70% from 87%), raise rises (26% from 7%). Hero is now in 2× tier weight every pot. The "coasting" framing is mostly accurate — wide cheap participation continues — but with 2× weight on the line, even passive pots accumulate faster.

s = 5+ (post-cliff-2, grinding). Each additional squid is +4 weight, flat. No more cliffs. The strategy converges toward chip-EV with a moderate accumulation bonus — and the data shows raise rate climbing slowly back up: 37.6% at s=5, 49.2% at s=6, 56.9% at s=7. Why the climb? At 4× weight the implicit ante per pot is high enough that hero's selection narrows toward stronger hands and bigger plays, even though no cliff is imminent.

The three plateaus look different because the reason hero is at each state is different. Post-safety hero coasts because the next motive is far away. Post-cliff-1 hero accumulates at higher per-pot weight, balancing wide participation with rising aggression. Post-cliff-2 hero grinds chip-EV at maximum weight, narrower but still aggressive.

§3.6 — Why the three transitions are asymmetric

Read the marginal-value column from the table at the top: +1, +4, +12. The proportional jumps are 1×, 4×, 12× over the Stand-up baseline.

This asymmetry is the structural fact about Hunt Progressive. The three transitions are not equally weighted. Cliff 2 dominates. The peak aggression in the entire game is at cliff-2-imminent (s = 4), not at cliff-1-imminent (s = 2) — because the cliff math says it should be.

Compare this to Blood Battle (Book 7), the smooth-quadratic Squid variant. Blood Battle's marginal weight at the analogous transition (s = 4 → s = 5) is +9, not +12. The biggest jump in Blood Battle is smaller than the biggest jump in Hunt Progressive — and Blood Battle's ramp is gradual, so no single state plays as sharply as HP's s = 4. (See OQ-7 in Part 6 §6.3 — Hunt Progressive's three-zone defense shape comes from this same +12 discontinuity, which Blood Battle lacks.)

Compare to Stand-up. Stand-up has only the safety line; no cliffs. Strategy organizes around being desperate or being safe — a binary. Hunt Progressive turns that binary into a graduated structure with two more reorganization points.

Three transitions, three magnitudes. The book's central insight — and what the rest of the chapters apply at every spot — is reading WHICH transition is in front of you and WHO it's in front of.

§3.7 — How transitions reshape the squid economy

Part 2 framed squid value as a settlement event with two sides: the future-loser pool (who pays) and the holder amplification (how much each loser pays). Each transition acts on a different side.

The safety line is a pool event. When you go from s = 0 to s = 1, your candidacy to be the lone-loser ends. The pool shrinks by one. The remaining desperates' candidacy grows — there are fewer of them to share the eventual loss. (Uri's swing formula encodes this: as N decreases, the per-pot swing grows because each remaining desperate is more likely to be the future-loser.) The amplification side is unchanged — your weight on the holder side just moved from 0 to 1.

The cliffs are amplification events. Crossing cliff 1 or cliff 2 doesn't change the pool — the same set of players are still desperates after the cliff as before. What changes is the holder side of the settlement. After CO crosses cliff 2, CO's weight goes from 2 × 4 = 8 to 4 × 5 = 20, and the sum_of_holder_weights jumps by +12. Every future loser owes +12×val more chips at game-end. This is a transfer event between the loser pool and the crossing holder — and it shows up in the literal val × sum_of_holder_weights line of the settlement formula.

So the three transitions are not the same kind of event:

This is why §6.3's three zones look the way they do. Hero's defense vs cliff-1-imminent CO is "deny the small amplification — it's cheap to fight." Hero's defense vs cliff-2-imminent CO is "commit, because the amplification event is worth being in the pot for." The differential isn't about CO's hand; it's about how big an amplification event the holder side is one pot away from.

A refinement the data forces (Phase 8 §6.5b): hero responds to amplification potential, not current amplification. A holder past s=5 has already amplified — no further cliff is coming for them. Hero's "commit" reflex weakens once the amplification is in the bag. This is why the safety line + two cliffs are decision-reshape moments specifically: each one represents a future shift in the squid economy that hero can position around.

The remaining chapters apply this spot by spot.

Reader takeaway from Part 3. Three transitions: safety line (a pool event), cliff 1 (a +4 amplification event), cliff 2 (a +12 amplification event). Asymmetric in magnitude, asymmetric in mechanism. The peak aggression in Hunt Progressive lives at cliff-2-imminent (hero=s=4) because that's the largest single amplification event in any Squid variant — every future loser's bill jumps by +12 weight when hero crosses. The +12 cliff is what makes Hunt Progressive Hunt Progressive rather than Stand-up or Blood Battle.

Draft · Squid Hunt Progressive Part 3 · Pass 1 · 2026-05-02 · v0.1.0

Part 4 — The Seven States — Hero's Squid Count as a Strategic Variable

Part 3 walked through the transitions — the moments of change. This chapter walks through the states — what it's like to be at each squid count, between transitions. Every hero squid count is a different game. The decision chapters (Parts 5, 6, 7) reference this vocabulary; the cheatsheet at §4.5 is the one-page reference you'll keep coming back to.

The seven states are the hero counts that exist in-distribution: s = 0, 1, 2, 3, 4, 5, 6+ (treated as one cohort past cliff 2). The shape across them isn't a smooth curve — it's a sequence of qualitatively different strategic postures, with the transitions from Part 3 separating them.

§4.1 — Three weight-function tiers, with internal structure

The weight function gives you three tiers (1× at s < 3, 2× at 3 ≤ s < 5, 4× at s ≥ 5). That's the math foundation. But each tier has internal structure the math doesn't capture — within Tier 1, the s = 0 state behaves nothing like the s = 1 state, even though both have the same per-pot weight.

Two pressures act on every hero state:

• Weight pressure. How much is this pot worth in squid-equity right now? Determined by hero's tier (1×, 2×, or 4×). Higher weight pressure = more reason to play any given pot harder.
• Time pressure. How soon does the next cliff arrive? Determined by hero's distance to s = 3 or s = 5. Cliff-imminent states feel urgent; post-cliff states feel relaxed.

The seven sub-states are the cross-product of these two pressures plus the safety-line status (desperate or safe). That's why the within-tier shape exists — same multiplier, different distance to the next reorganization, different posture.

§4.2 — Within Tier 1: three sub-states

Tier 1 is the 1× weight tier — same per-pot weight as Stand-up Game. But Hunt Progressive's Tier 1 contains three meaningfully different states.

s = 0 — desperate / safety-line-imminent.

You are the only player who can become the lone-loser if the game ends now. Two motives stack: insurance (capture any squid to exit the future-loser pool) and head start (start the race toward the +4 cliff at s = 3 from the lowest possible position).

Posture: maximum participation. At Z=3, val=3, CO opens, hero's VPIP is 92.6% — almost any pot is worth contesting. Limp 88.9%, raise 3.7%, sizing 7.2 bb when raising. ⚠️ Val-conditioning: the limp-pure shape (raise% near 0) is val=3-specific. At val=1 and val=10, hero=s=0 raises 37–44% (Phase 5 Slice 16). The max-participation finding is robust across vals; the limp character is the val=3 form.

Why limp-pure at val=3? At val=3 specifically, the implicit ante per pot peaks relative to chip-EV — cheap participation captures squid value without committing chips against an opponent who can outdraw. At extreme vals (val=1 where chip-EV dominates, val=10 where the cliff math dominates) hero shifts toward more aggressive opens.

s = 1 — the dead zone.

You crossed the safety line. The lone-loser threat is paid; the cliff at s = 3 is two pots away. The marginal value of the next squid is just +1, and there's no immediate reorganization to chase. This is the lowest-incentive state in Hunt Progressive — and the data shows it sharply.

At Z=3, val=3, CO opens: VPIP drops to 79.3%, limp 79.2%, raise 0.1%, sizing 7.1 bb (when hero raises at all). Compare to s = 0 above (VPIP 92.6%) and s = 2 below (VPIP 93.8%) — the s = 1 dip is real, not noise.

The dead zone is HP-M9 — an emergent finding, not derivable from the multiplier curve alone. The mental model from Part 2 says all three Tier 1 states should play similarly (same +1 marginal weight). The data says they don't, and the explanation is position relative to the next cliff: s = 0 has insurance + head start motives; s = 2 has cliff-1-imminent motive; s = 1 has neither, just the slow climb toward cliff 1. Hero coasts.

Practical consequence: don't expect aggression from a s = 1 hero the way you would from a s = 0 or s = 2 hero. The dead zone is the variant's quietest state. ✅ Val-stable in direction — Slice 16 confirms s = 1 is the lowest-aggression state at val=1 (raise 17.7%) and val=10 (17.5%) too; the trough is robust.

s = 2 — cliff-1-imminent.

One pot from the +4 conversion to the 2× tier. The cliff motive is now in front of hero — but the cliff itself is small. Hero's posture: wider, not sharper.

At Z=3, val=3: VPIP 93.8% (back up from the dead zone), limp 86.9%, raise 6.9%, sizing 7.2 bb. Wide cheap participation captures the cliff better than aggression here, because the +4 conversion (12 BB squid-equity per pot at val=3 N=2) doesn't yet justify chip commitment against opponents who can outdraw post-flop.

§4.3 — Within Tier 2: two sub-states

Tier 2 is the 2× weight tier. Hero just crossed cliff 1 and now plays at double per-pot weight.

s = 3 — post-cliff-1, coasting.

You're past the first cliff. Each new squid is worth +2 weight (twice Stand-up's baseline), and the next cliff (the +12 at s = 5) is two pots away. Posture: wider AND somewhat more aggressive, balancing accumulation with leverage on the bigger pots.

At Z=3, val=3: VPIP 96.1% (the highest VPIP in the seven-state sequence), limp drops sharply (87% → 70%), raise climbs to 26.2%, sizing 6.9 bb. Hero is converting wide passive participation into aggressive participation. Limp-heavy is still the dominant mode, but now you're raising about a quarter of the time you enter a pot.

The coasting framing applies: you're stockpiling for cliff 2, but at 2× weight every pot you enter is materially more valuable than it was at s = 2. Don't fold marginal stuff just because the next cliff is two pots out — accumulate.

s = 4 — cliff-2-imminent.

The peak state. One pot from the +12 conversion to the 4× tier — the largest discrete strategic event in any Squid variant.

At Z=3, val=3: VPIP 95.6%, limp drops further (47.1%), raise 48.5% — the highest raise rate in the seven-state sequence. Sizing 5.7 bb (smaller than the wide-limp tiers above, because the goal is to get to a flop with leverage, not to fold opponents out).

The shape flips from cliff 1's "wide and cheap" to cliff 2's "wide and aggressive." Hero's per-pot stake is so big — 72 BB of squid-equity at val=3 N=2 — that hero can't afford to play passively. Raise to put leverage on the pot, then bet through it on the flop.

✅ Val-stable in direction. Slice 16 confirms the s = 4 raise peak at val=1 (47.3%), val=3 (48.5%), and val=10 (65.8%). The cliff-2-imminent peak is the most robust finding in the chapter; magnitude grows with val.

The contrast between s = 2 and s = 4 is the cleanest illustration of the +4 vs +12 asymmetry. Both are cliff-imminent states. Both have similar VPIP. But s = 2 limps 87% / raises 7%, and s = 4 limps 47% / raises 48%. Same posture toward the cliff (be in pots), opposite postures within those pots (cheap vs leveraged) — because the cliff math is different by a factor of three.

§4.4 — Tier 3 (s ≥ 5): past-peak grinding

You captured both cliffs. Each additional squid is +4 weight, flat. No more cliffs.

The strategy converges toward chip-EV with a moderate accumulation bonus. The data shows raise rate climbing slowly with s: 37.6% at s=5, 49.2% at s=6, 56.9% at s=7. VPIP narrows somewhat (86–90% range) as hero gets more selective. Sizing stabilizes around 4.5 bb.

Why the slow raise climb past s = 5? At 4× weight, the implicit ante per pot is high enough (about 16 BB at val=3 with two desperates) that hero's marginal hands have a bigger reason to play hard than they did at 2× weight — even though no cliff is in front. Each accumulated squid is worth +4 weight at the current multiplier; not as decisive as a cliff conversion, but not negligible.

The grinding framing is right. You're past the major reorganizations. Strategy is "accumulate at high weight, narrow up, keep the foot on the gas in the pots you do enter."

§4.5 — The seven-state cheatsheet

One reference card. Everything else in this chapter compresses into this:

(Numbers from Pull 13 — H-Z at Z=3 val=3 CO opens. Z=3 chosen as the "mid-game default"; magnitudes shift at other Z values. The shape of the sequence is robust.)

§4.6 — How the seven states map to the squid economy

Part 2 framed squid value as a settlement event with two sides: the future-loser pool (who pays) and holder amplification (how much each loser pays). Hero's state determines hero's relationship to BOTH sides.

Pool side — distance from being the future-loser. s = 0 hero IS in the pool. s = 1 hero is just out (once safe, always safe — you can't lose your squid in Hunt Progressive). s ≥ 2 hero is irrevocably out of the pool, with progressively more weight relative to it.

Amplification side — hero's share of the eventual loser payment. Each squid you hold is a draw on the future-loser's stack. The cliffs upgrade hero's draw size dramatically — cross cliff 2 and hero's share of the loser payment goes from 2 × s weight to 4 × s weight. Hero's tier is a direct read on hero's share of the eventual squid prize. Hero's amplification potential — distance to the next cliff hero will cross — modulates how aggressively hero plays for forward squids.

So the seven states modulate hero's position on BOTH sides:

This is the framing the decision chapters use. Reading hero's state means reading hero's relationship to both sides of the settlement event, not just hero's chip stack.

§4.7 — The "down-to-two" analog — Hunt Progressive shape

In Stand-up Game (Book 2), the down-to-two state — when only two players are still desperate (Z = 2) — is the highest-leverage state in that variant. Both desperates have full implicit ante on the line; one of them is going to lose; everyone else is watching.

Hunt Progressive has an analog, but it's shaped by tier as well as by Z. Framework attribution: Nick Petrangelo (idea-source).

The Hunt Progressive down-to-two state combines low Z (high implicit ante per pot) with the tier composition on the holder side. Phase 8 Slice 18d tested two specific configurations at hero=BB s=3, CO=0, Z=2:

Both at the same Z, same opener, same hero state. Hero plays opposite ways depending on the holder structure — and the direction is what the two-sides framing predicts:

• Spread holders one cliff away each: multiple amplification events still in front. Hero defers (call-heavy) to be in pots when those cliffs cross.
• One holder past peak (already amplified): no further amplification coming on that side; hero shifts to chip-pressure mode (raise-heavy). The amplification is already in the bag.

This is the empirical version of the "amplification potential, not current amplification" finding from Part 2 §2.2. The data forces it: down-to-two doesn't have a single shape; it has at least two, and the discriminator is whether the holders have amplification potential remaining or not.

The Stand-up down-to-two reflex carries over partially: tighten up if you're at risk, push if you're not. Hunt Progressive layers a second consideration on top: are there cliff-imminent players at the table who'd amplify the loser payment if they cross? If yes, defer (be in the pot). If no — even at low Z — push for chips.

⚠️ Coverage caveat: Slice 18d tested 2 configurations. Other holder structures (e.g., two cliff-imminent holders, mixed past-peak + cliff-imminent) are open work. The qualitative direction is data-backed; the full surface across holder configurations is not.

Reader takeaway from Part 4. Seven hero states, three pressures (weight, time, status), one cheatsheet. Each state is a different game with different motives, posture, and numbers. The decision chapters (Parts 5, 6, 7) cite these states by name — when you see "cliff-2-imminent hero" in §6.4, you should already know that means s=4 and have the posture in mind. ⚠️ Val-conditioning callout: the magnitudes in §4.5 are at val=3 (the canonical training value). Direction is val-stable across canonical vals (1, 3, 5); shape sharpens at val=3. See val-stability-audit.md for per-state verdicts.

Draft · Squid Hunt Progressive Part 4 · Pass 1 · 2026-05-02 · v0.1.0

Part 5 — Preflop Opens

How Hunt Progressive's structure shows up when hero is the opener, not the defender. Three things to learn here: how much wider hero opens vs the cash baseline (§5.1), the val=3 limp-pure shape (§5.2), and how the seven-state structure replicates across positions (§5.3). The rest is val and Z gradients (§5.4–§5.5) and a heuristic check (§5.6).

§5.1 — The widening — Hunt Progressive vs three baselines

The single biggest fact about preflop opens in Hunt Progressive: hero plays much more than in NLHE Cash. At fresh state (Z=6, all desperate), val=3:

The widening from cash to Hunt Progressive is +42 to +56 percentage points. UTG goes from playing 19% of hands in cash to 63% in HP. CO goes from 29% to 87%. BTN approaches 100%.

Hunt Progressive's VPIP shapes match Blood Battle's closely (both are accumulation modes). What separates HP from BB is the raise-vs-limp mix, not the VPIP. Stand-up sits in between cash and the accumulation modes; it's the bridge variant.

The structural reason traces back to Part 2: every pot pays an implicit ante. At fresh state val=3 with N=6 desperates, that ante is 3.6 BB per pot — enough that hands which fold for chip-EV in cash become profitable to enter for the squid-equity. Cite: uri-squid-invisible-ante §3a (the limp-return finding extends here from Stand-up).

§5.2 — The val=3 limp-pure peak

Look at the raise% column for Hunt Progressive vs Blood Battle at fresh state val=3:

Hunt Progressive at val=3 fresh state collapses to near-pure-limp at UTG, MP, and CO — raise rates under 1%. Blood Battle still raises 3–10% at the same positions. This is a Hunt Progressive distinguishing finding: at val=3 the limp-pure peak is sharper than in any other Squid variant tested.

Why val=3 specifically? At val=3, the implicit ante per pot peaks relative to chip-EV. The cliff structure adds a forward jackpot premium that makes wide cheap participation maximally profitable — every limp is a step toward the cliffs at s=3 and s=5. Raising commits chips against opponents who can outdraw post-flop; limping captures the squid value at lower chip cost.

⚠️ Val-conditioning: this is a val=3-specific shape. At val=1 hero opens raise-heavy (Slice 16: UTG 30%+ raise, CO 38%+ raise — chip-EV dominates). At val=10 hero also opens raise-heavy (UTG 19%+, CO 44%+ — cliff math dominates). The limp-pure regime exists in a narrow val band centered on val=3.

The val ladder for hero=s=0 (canonical fresh state):

Three regimes, with val=3 sitting in the limp-pure middle. The val=5 cell isn't in Slice 16 directly but Pull 4 confirms intermediate behavior (raise% climbs back up from val=3's near-zero).

§5.3 — The transition signature replicates across positions

The seven-state shape from Part 4 isn't a CO-specific phenomenon. The same dead-zone-then-cliff-2-peak structure shows up at UTG and BTN too.

UTG and BTN sequences (Pull 3a — H-Σ at val=3, opps fresh):

All three positions show: - The s=1 dead zone (lowest raise rate) - The cliff-1-imminent rise at s=2 - The cliff-2-imminent peak at s=4 - The grinding climb past s=5

Magnitudes scale with positional aggression baseline (BTN > CO > UTG), but the shape is invariant. The cliff structure is structural to Hunt Progressive, not position-specific.

§5.4 — Z gradient on opens

Setup: hero=s=3, val=3, 100bb stacks. Z varies from 5 (fresh) to 2 (cliff-before-game-end). (Data: Pull 4 row-by-row at val=3.)

Modest VPIP variation (~7pp swing) and modest raise variation (~14pp). The implicit ante quadruples from Z=5 to Z=2, but hero's open shape barely shifts. Compare to defense (§6.6) where the same Z swing produces a 45pp raise-rate change — opens are much less Z-sensitive than defense.

Why? On opens hero is choosing whether to initiate a pot. Hero's downside is committing chips with no information about opponents' hands. As Z drops and the implicit ante grows, initiating becomes more rewarding — but also more risky if an opponent re-raises. The two effects partially cancel; the open shape stays close to constant. On defense, hero already knows an opponent is in — the implicit ante advantage doesn't come with the same downside, so hero can shift much more dramatically.

§5.5 — Val asymmetry on opens at fixed Z=3

Setup: hero state varies, Z=3 fixed, val varies. (Data: Phase 5 Slice 16 — paradigm-clean H-Z opens × val ladder.)

The full surface (raise%):

Three things to read off:

1. The cliff-2 peak at hero=s=4 is val-stable in direction. Highest raise rate at every val (47%, 49%, 66%). The +12 cliff dominates everywhere.
2. The val=3 limp-pure shape lives at hero=s=0,1,2. Only val=3 produces the near-zero raise rates at the early hero states. At val=1 and val=10, hero opens raise-heavy at every state.
3. At val=10, hero=s=4 raises 66%. The biggest raise rate in the whole surface. Cliff-2-imminent at high val is a max-aggression state — the squid math is so big that hero plays nearly every hand for a raise.

§5.6 — The "raise to size of squid" heuristic — Hunt Progressive caveat

Uri's "raise to size of squid" heuristic predicts UTG opens to ~3.6 BB at val=3 in Stand-up Game (the implicit ante per pot at fresh state). The intuition: open large enough that you make the implicit ante part of the pot, capturing it for the eventual winner.

In Hunt Progressive, the solver opens larger than Uri's heuristic predicts. Pull 13 shows hero=s=0 raising at 7.2 bb when raising; UTG-position opens (Pull 3a) at 6.7-6.8 bb. These are larger than Stand-up's ~3.6 bb predicted from Uri's formula.

Why? Hunt Progressive's cliff structure adds a forward jackpot premium. The implicit ante for this pot is roughly Stand-up's value, but hero is also opening to set up the cliff approach — every pot won is a step toward s=3 or s=5. Larger opens get more chips into pots the cliffs will multiply if hero crosses.

The heuristic captures the spirit (open larger than chip-EV would suggest, because squid value adds to the pot) but undershoots the magnitude in HP. Use Uri's formula as a lower bound on Hunt Progressive open sizing, not as a target. Cite: uri-squid-invisible-ante §3b.

§5.7 — The opener decision tree

The chapter compresses to: read your tier, then your val, then your position.

At val=3 (canonical training value): - s=0 / s=1 / s=2 (Tier 1): limp wide. Raise rare (under 10%) at UTG/MP/CO. BTN raises 3–12%. - s=3 (post-cliff-1): start raising more — 22–32% across positions. - s=4 (cliff-2-imminent): peak aggression. Raise 32% at UTG, 49% at BTN. Sizing 5.7–6.5 bb. - s=5+ (past-peak): narrow up; raise rates climb slowly (37–60% at BTN by s=7).

At val=1 (chip-EV regime): - Open raise-heavy at every hero state. The limp-pure shape doesn't apply. - Cliff-2 peak still at s=4 (47% raise) — direction holds.

At val=10 (cliff-math regime): - Open raise-heavy AND with bigger commitment. Cliff-2 peak hits 66% at hero=s=4. - The squid math is so big that hero plays nearly every hand for action.

Position-by-position default opens at val=3 fresh state:

⚠️ Val-conditioning: this whole chapter is anchored at val=3. The widening (§5.1) and limp-pure peak (§5.2) are val=3 phenomena; the cliff-2 raise peak (§5.5) is val-stable. See val-stability-audit.md for per-claim verdicts.

Reader takeaway from Part 5. Opens are wide, limp-heavy at val=3, and centered on the cliff-2-imminent peak at hero=s=4. The seven-state shape replicates across UTG, CO, BTN. The Z gradient is modest on opens (much sharper on defense). Val=3 is the limp-pure regime; val=1 and val=10 are raise-heavy regimes with the same s=4 peak.

Draft · Squid Hunt Progressive Part 5 · Pass 1 · 2026-05-02 · v0.1.0

Part 6 — BB Defense — Reading the Squid Economy

[Pass 3: full chapter drafted. §6.1 through §6.9 all complete with paradigm-clean data. Voice consistent with §6.3 revision. Length runs ~3500 words vs the 1800 target — over due to thorough data coverage. Tighten in editorial pass.]

§6.1 — BB defends ~100% in-distribution

Setup: BB defends a 2.5x CO open. Hero squid count varies across s ∈ {0..7}. Opps fresh, val=3, 100bb stacks. (Data: Pull 6 — H-Σ paradigm, opps held fresh; Z drifts as hero crosses safety line.)

In NLHE Cash, BB folds something on the order of 30–40% to a CO open at 100bb. In Hunt Progressive, BB folds 0% at every hero state we measured in-distribution. The fold-call-raise mix becomes a call-vs-raise mix.

(Pull 6 — opps held fresh, so Z drifts from 6 at s=0 to 5 at s≥1. Z drift flagged in PARADIGM_AUDIT.)

The structural reason: BB closes the action heads-up with great pot odds, AND the implicit ante adds squid-equity on top of those pot odds. Even hero=s=1 (the dead zone) defends 100% — folding gives up too much when the squid value is on the line.

This generalizes Uri's "BB defense doubles" finding from Stand-up. Stand-up showed BB's defense rate roughly doubling vs cash NLHE; Hunt Progressive saturates — BB defends every in-distribution state (the only exception being val=1, covered in §6.7). Cite: uri-squid-invisible-ante §3d.

The practical translation: as BB in Hunt Progressive, your decision is rarely fold/play; it's call/raise. Treat fold as a non-option unless you see a state-specific reason to consider it (val=1, past-peak opener — both covered downstream).

§6.2 — The transition signature on defense — peak raise at hero=s=2

The same data, viewed for the raise-rate shape across hero states. The seven-state structure from Part 4 reads through to defense, but the peak is in a different place than on opens.

On opens (Part 3 §3.4), peak raise was at hero=s=4 (cliff-2-imminent) — the +12 cliff in front. On defense, peak raise is at hero=s=2 (cliff-1-imminent) — 82.9% raise.

Why the shift? BB defense is reactive — hero is responding to a price the opener set. The decision is whether to commit chips against an opener who already showed willingness to play. At hero=s=4, hero's own +12 cliff is in front, but committing chips against an opener who's also incentivized to keep going up (because they bet first) gets expensive. At hero=s=2 the +4 cliff is small enough that hero can take a pure pressure approach — raise wide, take down most pots, accumulate squids cheaply.

The shape is: raise mass concentrated at the small-cliff-imminent state, more call mass at higher hero states where chip commitment grows risky. Folding doesn't appear — see §6.1.

⚠️ Val-conditioning note: this section uses Pull 6, which is H-Σ paradigm with Z drift. The s=2 peak is robust at val=3; cross-val behavior is partially covered by §6.4's Slice 15a/b but not at the §6.2-style "opps fresh" paradigm. The peak-at-s=2 finding is a val=3 reading; treat it as the central tendency, not as a universal claim.

§6.3 — The three zones

Setup: hero=BB at 3 squids throughout this section (post-cliff-1, mid-Tier-2). Z=2 (two desperates), val=3, 100bb stacks. The only thing varying cell to cell is the opener's squid count. (Data: data/pull_chapter6_slice_c.out.json. Methodology framework in appendix.)

You're in the BB with 3 squids. You've just crossed the first cliff, so every pot you win now pays double weight. The +12 cliff at s=5 is two pots away. Someone opens from CO. Fold, call, or raise?

In NLHE you'd start with: do I have equity here?

In Hunt Progressive, ask a different question first: how many squids does the opener have? That number reshapes your decision more than your hand does, in three distinct ways.

Read top to bottom. Your action mix isn't sliding along one axis ("raise less, fold more"). There are three different things going on at three different points, plus two reference states above and below. Walk through them in order.

CO has 2 squids — one pot from the small cliff (+4). You raise 80%. Pressure works here because CO's forward stake is modest — about 12 BB on the line at val=3. When you 3-bet, CO folds the marginal stuff rather than fight. The cliff is small enough that denying it is cheap. Take the pot.

CO has 4 squids — one pot from the big cliff (+12). Now you call 84% and raise just 16%. The +12 conversion is the largest single jump in any Squid variant — about 36 BB of forward stake on CO's side. Pressure backfires here. CO's math is too good to fold to a 3-bet, so raising just costs you chips against an opponent who'll escalate. Calling realizes your equity at lower chip cost. Don't try to deny the cliff — just be in the pot when CO crosses it.

CO has 5+ squids — past both cliffs. No more cliffs to fight over. Each new pot adds modest weight (~12 BB at val=3, comparable to the s=2 case), but with nothing to deny on CO's side and no big jackpot to commit to, your marginal hands lose their reason to be in the pot. Folding shows up here for the first time (~14%) — the only state in this paradigm where it does.

What ties the three zones together: how big is CO's forward stake, and is there a cliff to fight over? Small stake + cliff to deny → you raise. Big stake + cliff to fight over → you call. Small stake + no cliff → start folding marginals. Your hand matters less than CO's squid count.

⚑ OQ-7 cross-mode comparison — resolved 2026-05-02 (Phase 6 pull). The total magnitude of this adaptation isn't unique to Hunt Progressive. Blood Battle shows a similar ~75pp spread in BB defense across CO's accumulated squid count; both are accumulation modes, and both adapt sharply. NLHE Cash spreads about 17pp across opener positions; Stand-up about 11pp across CO's safe-vs-desperate states. So the big magnitude is a property of accumulation modes generally, not of HP specifically.

What IS distinctively Hunt Progressive: the three-zone shape. Blood Battle at the same paradigm shows pressure-plateau-drop (raise 76% / 75% / 74% / 9% across s_CO=2/4/5) — high pressure across the middle, then a drop at s_CO=5, but no commit-mode at s_CO=4. The +12 discrete cliff in HP creates the commit zone; Blood Battle's smooth quadratic ramp doesn't have that discontinuity (only +9 marginal at s_CO=4 → s=5). So the chapter's central claim, corrected: the three-zone shape (specifically the commit zone at cliff-2-imminent) is HP's signature. The total magnitude is just what accumulation modes do.

What we just walked through is hero at s=3 specifically. The three-zone shape is sharpest there — raise stepping 80% → 16% → 6% across CO's tier states. Why s=3 in particular?

Look at how much hero's raise rate moves between cliff-1-imminent CO and cliff-2-imminent CO at each hero state, holding val=3:

The reading: at s=0, hero is desperate enough that your own urgency dominates the call (1pp swing — barely moves with CO's state). At s=4, hero is one pot from a +12 cliff of your own — that pulls focus too (17pp swing). s=3 sits in the sweet spot: past cliff 1 already, still two pots from cliff 2, your own forward stake moderate enough that CO's state does the steering. Goldilocks.

That's the val=3 picture. The full surface — raise-rate swing across (hero, val):

(s=5+ skipped — squid totals exceed T=10. Negative swings mean hero raises MORE vs cliff-2-imminent CO than cliff-1-imminent — anomalous behavior at desperate hero in extreme val tiers.)

At val=1, val=3, and val=5 the same story holds: hero=s=3 is the peak responsiveness state, sharpest at val=3. At val=10 the rule breaks — the peak shifts to s=1 and s=3 drops to a 14pp swing. We don't fully trust this. val=10 sits at the edge of the model's training range, and edges can produce artifacts; treat it as a footnote, not the lead. Open work logged as OQ-8.

So the chapter's claim, with discipline: at hero=s=3 across canonical training vals (1, 3, 5), the three zones repeat — sharpest at val=3, softer at vals 1 and 5, but always pointing the same way. §6.4 takes the framework off the s=3 anchor and shows how each hero state plays the same situation differently.

§6.4 — The same situation at every other hero state

Setup: same defense decision as §6.3 (BB facing CO open) but now hero's squid count varies. Hold val=3, Z=2, 100bb stacks. Two paradigm-clean slices: hero state varies vs CO at cliff-1-imminent (Slice 15a), vs CO at cliff-2-imminent (Slice 15b).

§6.3 showed that at hero=s=3 the three zones produce a sharp deny→commit swap. Does that pattern survive when hero is at a different squid count? The answer is mostly yes in direction, but the sharpness depends on hero's own forward stake.

Hero's response to the deny→commit swap, side by side:

Walk through what each hero state does:

Hero at s=0 — your own urgency dominates. You're desperate, fighting for your first squid. CO's tier doesn't change your call: you raise 83% vs cliff-1-imminent CO and 84% vs cliff-2-imminent CO. The pattern collapses entirely — when your own forward stake is the safety line itself, CO's state is noise.

Hero at s=1 — the dead zone defers. You raise 30% vs cliff-1-imminent CO (modest) but call 90% vs cliff-2-imminent CO. Direction holds — you adapt to CO's state — but you're playing softer than the s=2/3 hero would. The 24pp swing is the smallest "real" version of the pattern.

Hero at s=2 — joint cliff-1-imminent. Both you and CO are about to cross cliff 1. The deny zone is sharp (raise 73%) — joint pressure. Vs cliff-2-imminent CO, you swing to call 73%. Direction holds with 45pp magnitude, somewhat softer than s=3.

Hero at s=3 — Goldilocks. Recap from §6.3. The +64pp swing is the sharpest expression of the pattern. Past cliff 1 already, still two pots from cliff 2 — your own forward stake is moderate enough that CO's state does the steering.

Hero at s=4 — your own cliff is competing. You're one pot from a +12 cliff yourself. Vs cliff-1-imminent CO you raise 58% (lower than at s=3 because pressure is split between yours and CO's stakes). Vs cliff-2-imminent CO you call 59% — but you also raise 41%, much more than at s=3, because your own commit pulls toward "be in the pot AND threaten." The 17pp swing is the muted version: direction holds but magnitudes converge.

The rule: as your own forward stake grows, CO's state matters less. The peak responsiveness is at hero=s=3 because that's where your own urgency is dampest.

Mutual cliff-imminent (hero=s=4) — does the three-zone shape still appear when hero ALSO has a cliff in front?

Slice 15d holds hero at s=4 and varies CO across 0..5:

Yes — the three zones reappear, just with smaller swings. Vs cliff-1-imminent CO you raise (58%); vs cliff-2-imminent CO you tilt toward call (59%); vs past-peak CO you start folding (5%). All directions match §6.3. What's different at s=4 is that your own commit pulls call% down vs cliff-2-imminent CO and pulls raise% up — pressure and call coexist when both players have big cliffs in front.

Takeaway for §6.4: the three-zone direction holds at every hero state from s=1 to s=5. The sharpness is hero-state-dependent — Goldilocks at s=3, muted at s=2 and s=4, collapsed entirely at s=0 (where own urgency dominates). The dead zone at s=1 produces the softest version of the real pattern. ⚠️ Val-conditioning: this is the val=3 picture; from val=10 we know own-urgency-dominates extends differently. See val-stability-audit.md.

§6.5 — Does the pattern hold against other openers?

Setup: hero=BB at 3 squids, val=3, Z held; opener position varies between CO (canonical), SB, and BTN. (Data: Phase 4 δ, paradigm-clean O-Z at hero=s=3.)

§6.3 only tested CO openers. If the three-zone shape is structural to Hunt Progressive's mechanics, it should reproduce when the opener is in a different seat. If it's a CO-specific quirk, it won't.

Hero's response across opener position:

Three zones, three positions:

The commit zone is rock-solid across all three openers. When the opener is one pot from the +12 cliff, hero calls ≥83% regardless of opener position. The call-dominant response to cliff-2-imminent opener is structural — wherever the opener sits, the +12 forward stake on their side is too big to pressure.

The deny zone holds at CO and BTN, weaker at SB. Vs cliff-1-imminent BTN, hero raises 65% (similar shape to CO's 80%). Vs cliff-1-imminent SB, hero raises just 35% — the deny pattern softens significantly. Why SB? The SB-vs-BB dynamic underneath has hero defending wide and calling-leaning by default (BB closes action heads-up with great pot odds). That underlying texture dampens the deny response when CO's cliff-1 stake adds onto it. Pressure costs more chips relative to the pot in heads-up, and heads-up vs SB hero already realizes equity well by calling.

The fold zone shows up at every position. When the opener is past-peak, hero introduces fold mass — 14% vs CO, 11% vs BTN, 8% vs SB. The shape is the same; magnitude tracks the underlying calling threshold of each matchup.

Reading: the three-zone shape is a property of Hunt Progressive's mechanics, not a CO-specific finding. The commit zone reproduces almost exactly. The deny zone direction reproduces but magnitude varies with the underlying position dynamic (sharpest in CO and BTN where pre-flop is more contested; muted in SB where BB defaults to call-heavy). The fold zone reproduces with magnitudes that track each position's calling threshold.

⚠️ Val-conditioning: same caveat as §6.3 / §6.4 — tested at val=3 only. The val-stability of the deny→commit swap was confirmed across canonical vals at the CO-opener anchor; the cross-position version is open work. Logged as part of OQ-3.

§6.5b — Off-pot amplification: the pattern generalizes to any cliff-imminent player

Setup: hero=BB s=3, opener is a baseline desperate (s_CO=0). Z=3, val=3, 100bb. The off-pot player BTN's tier varies. (Data: Phase 8 Slice 18a — paradigm-clean O-table, Z asserted at runtime.)

§6.3 showed hero's defense responding to the opener's amplification potential. §6.5 confirmed the pattern reproduces across opener positions. Now the bigger question: does hero respond to any cliff-imminent player at the table — even one not in the pot?

Same hero, same opener, same Z. The only thing varying is BTN sitting at the table outside the pot. Hero's raise rate drops 39pp (88% → 49%) as BTN's tier climbs.

The §6.3 deny→commit pattern is NOT specific to the opener. It generalizes to any player whose accumulation produces a big amplification share or potential. BTN at s=5 is sitting on weight 20 (the maximum) — every future loser owes BTN's full 20-weight share. Hero shifts from pressure-mode to call-mode to be in pots that contribute to BTN's amplification (pots BTN won would have transferred chips out of the loser pool to BTN; pots hero wins keep chips on hero's side, which doesn't directly amplify but does keep hero away from the loser pool).

The cleanest possible test — concentrated vs spread holders, same opener. Slice 18c holds opener=CO=0 and Z=3, varies the holder distribution:

Same total holders, different concentration. Hero plays a completely different game — pressure-dominant when the holder side is spread (no big amplification potential anywhere), call-leaning when concentrated (BTN's amplification share dominates the read).

Refinement: hero responds to amplification potential, not current amplification. Slice 18d — the down-to-two state at Z=2 — showed the inverse pattern of Slice 18c, because at Z=2 hero faces holders who are already past-peak (BTN at s=5, no further cliff coming). When the amplification is in the bag for the high-tier holder, hero shifts back toward chip-pressure mode (raise 67% vs 41% in spread). This is the same finding from a different angle: hero positions around amplification that hasn't happened yet — cliffs in front, not cliffs already crossed.

Practical: when you defend BB in Hunt Progressive, scan the table for high-amplification-potential players — not just the opener. A non-opener BTN sitting at s=4 (cliff-2-imminent, +12 amplification one pot away) is a reason to call rather than raise, because the downstream amplification event is the dominant signal. A non-opener BTN sitting at s=5 (past cliff 2) is a reason to raise — they've already amplified, no further amplification ahead. The s=2 case (cliff-1-imminent, +4 amplification) wasn't directly tested off-pot — theoretically it should produce a smaller pull toward call than the s=4 case (smaller amplification event). Treat as a third gradient between baseline and the s=4 finding.

⚠️ Val-conditioning: tested at val=3 only. Direction is theoretically grounded in Part 2's settlement formula (val-independent); magnitude is val=3-specific.

⚠️ Coverage caveat: Slice 18a tests one off-pot seat (BTN). Whether the magnitude of off-pot amplification scales with the off-pot seat's position relative to hero (e.g., MP vs BTN vs UTG) is open work. Plausibly position-invariant since hero's defense decision is hero-vs-opener and the off-pot player isn't in the pot — but unverified.

§6.6 — Z gradient on defense — much sharper than on opens

Setup: hero=BB at s=3, val=3, 100bb. Z varies from 5 (fresh game) to 2 (cliff before game-end). (Data: Pull 7 row-by-row at val=3.)

Z is the co-desperate count — how many players still have zero squids. Part 2's swing formula tells you the implicit ante grows as Z shrinks (6 × val / (N − 1)). On opens, this produces a modest ~10pp swing. On defense, the swing is much sharper.

A 45pp swing in raise rate as Z drops from 5 to 2 — call rate moves the inverse. Same hero state, same val, same opener; the only thing changing is how much implicit ante is on the line.

Direction: as Z drops, hero shifts from raise-heavy to call-heavy. Why? Higher implicit ante means the pot is more valuable to be IN (calling realizes equity at lower chip cost), and more dangerous to commit chips against (raising into an opener who is also incentivized by the bigger ante invites escalation). The same logic that drove §6.3's deny→commit swap — but here it's driven by Z, not by opp's tier.

❓ Val-conditioning: this measurement is at val=3 only. Phase 7 mapped val × hero × s_CO but did not sweep Z. Whether the ~45pp swing magnitude holds at other vals is open work (see val-stability-audit.md). Direction is structurally predicted by Part 2's framework — should hold across vals — but magnitude is unverified.

Practical: when you walk into a game and the table is fresh (Z high), defend by raising wide. As the game progresses and Z drops, shift toward call-heavy defense. Same hand, different decision.

§6.7 — Val asymmetry on defense — val=1 is the fold zone

Setup: hero=BB at s=3, Z=2, 100bb stacks. Val varies across {1, 3, 5, 10}. (Data: Pull 7 row-by-row at Z=2.)

Val=1 is the only val where BB shows fold mass at hero=s=3 in-distribution. Phase 7 confirmed this across the full hero × val × s_CO surface — fold mass appears at val=1 and is essentially zero everywhere else.

Why val=1 specifically? At val=1, each squid weight unit is worth 1 BB at settlement. Hero's per-pot squid-equity at s=3 (Tier 2, +2 weight per pot) is just 2 BB worth of forward stake — small enough that chip-EV considerations dominate, and chip-EV says some marginal hands should fold to a 2.5x open even closing the action.

At val=3 and above, the squid-equity per pot exceeds the chip cost of defending wide enough that fold mass disappears. The crossover is between val=1 and val=3: the jump from 1 BB to 3 BB per squid weight unit is enough to flip "some folds OK" into "always defend."

The flip side at val=1: raise drops to 6.8%. With small implicit ante, hero doesn't have a big reason to pressure the opener either. The mix becomes call-heavy with some folds — much closer to a cash-NLHE-like response.

✅ Val-stable in direction — Phase 7 confirms only val=1 produces fold mass at any hero state (up to 18.4% at hero=s=4 vs cliff-2-imminent CO). The "val=1 is the fold zone" finding is robust.

§6.8 — The "less zero-sum" corollary applied to defense

The cash reflex on defense: when an opponent opens and you're in BB, you're either ahead of their range (defend) or behind it (fold). Bluff equity comes from making the opener fold to a 3-bet.

In Hunt Progressive, that frame is partially wrong. Per Part 2 §2.3 and Uri §6, your bluff equity comes from a settlement event with two sides (pool + amplification) — not from the calling opener's stack. So you can have profitable 3-bet bluffs even if the opener never folds, because the squid you collect comes from the loser pool, sized by holder amplification.

The practical consequence on defense: the high raise rates you see in §6.1 and §6.2 (BB raising 70-83% of opens at canonical states) aren't all "value raises" in the cash sense. Some of that mass is squid-driven aggression — playing to take the pot AND the squid, where the squid value fills in for the chip-EV gap when your opponent doesn't fold. Cite: uri-squid-invisible-ante §6.

The same multiway pot can host multiple profitable plays at once: a desperate caller defending for the safety squid (the pool side — exit the loser pool), a cliff-imminent hero raising for the cliff conversion (the amplification side — grow your share of the loser payment), and a safe mid-tier player bluffing for the future-loser draw. None of those plays are zero-sum against each other in squid-equity terms — they all source value from the same settlement event but capture different parts of it.

This is why §6.3's deny→commit swap and §6.5b's off-pot amplification finding both exist. In a per-hand zero-sum world, hero would either always raise or always call vs a single opener. In Hunt Progressive's less-zero-sum world, the choice depends on which strategy captures more of the upcoming settlement event — and that depends on the full table's pool + amplification structure, not just the opener's hand.

§6.9 — The defense decision tree

The chapter compresses to a four-cell read at hero=BB s=3 val=3 — the canonical BB defense state.

For other hero states, apply §6.4's adjustments: - Hero s=0: own urgency dominates — raise vs cliff-1-imminent CO AND vs cliff-2-imminent CO (the deny/commit shape collapses) - Hero s=1 (dead zone): defer everywhere — call-heavy across all CO states - Hero s=2 (joint cliff-1): sharp deny zone, softer commit zone - Hero s=4 (own cliff in front): muted three-zone shape; raise + call coexist vs cliff-2-imminent CO

For other opener positions (§6.5): - vs SB: call-leaning by default — deny zone muted (35% raise vs cliff-1-imminent SB), commit zone reinforced (92% call vs cliff-2-imminent SB) - vs BTN: intermediate between CO and SB; deny zone holds at 65% raise, commit zone 83% call

For off-pot amplification (§6.5b): - A non-opener at s=4 (cliff-2-imminent): modest pull toward call — minor adjustment (~7pp at val=3 Z=3 baseline opener) - A non-opener at s=5 (Tier 3, weight 20): large pull toward call — raise drops 39pp (88% → 49%) - A non-opener at s=2 (cliff-1-imminent): untested off-pot; theoretically a small pull toward call (smaller amplification event than s=4) - Multiple non-openers near cliffs vs one past-peak: the spread-vs-concentrated test (Slice 18c). Concentrated (one past-peak holder) flips hero's response significantly.

⚠️ Val-conditioning: the canonical numbers are at val=3. At val=1 the entire defense distribution shifts — fold becomes a real option (§6.7), raise rates drop, and the deny/commit pattern weakens to about 33pp swing (vs 64pp at val=3 — see Phase 7 surface in §6.3). At val=10 the pattern weakens further but doesn't invert. Treat val=3 numbers as the central tendency; consult §6.7 for the val ladder.

The whole chapter compresses to: read opener's tier, then read hero's tier, then read off-pot amplification, then read Z, then read val. The product gives you the right action.

Draft · Squid Hunt Progressive Part 6 · Pass 1 (§6.3 only) · 2026-05-02 · v0.2.0 in flight

Part 7 — Postflop — Texture-Dependent Cliff Expression

The cliff structure DOES express postflop — but only on connected/middling boards. Dry boards saturate (c-bet at near-100% across all hero states); wet boards converge (modest variation); connected boards show the full signature with a 30+ pp swing in c-bet frequency between pre-cliff-1 and post-cliff-1 hero states.

This chapter is the texture-dependent finding from Phase 5 Slice 17. It supersedes the "postflop is largely tier-insensitive" framing from earlier research notes — the original framing was correct on dry boards and partially correct on wet boards but missed the connected-board cliff signature.

§7.1 — Dry boards saturate

Setup: Ah9c4d. Hero opened CO at val=3 Z=3, BB=s=1 calls, single-raised pot heads-up to flop. Hero c-bet decision.

C-bet frequency saturates at 97-100% across every hero state. Range advantage on a dry ace-high board is so dominant that hero auto-bets regardless of squid state.

What DOES vary with hero state is sizing. s=1 dead zone bets 4.1 bb (smallest); cliff-2-imminent s=4 bets 10.9 bb (largest — almost 3× the dead-zone sizing). The cliff structure expresses through bet size, not bet frequency, on dry boards. Frequency is dictated by the board; sizing reads hero's tier urgency.

Practical: on dry boards, c-bet. Use bigger sizing when you're cliff-2-imminent (s=4) — this is where you want maximum chips in the pot if the cliff is going to upgrade your accumulated weight.

§7.2 — Wet boards converge with modest variation

Setup: Ts9h8h (monotone-flush + connected). Same configuration otherwise.

Wet boards have built-in equity equalization — both players' ranges connect with the board, so neither has the structural range advantage that dry boards give the preflop aggressor. C-bet rate stays in the 84-91% band across hero states. The cliff effect is dampened to 4-7pp variation.

Sizing scales with hero state again: s=1 dead zone 6.7 bb, cliff-2-imminent s=4 13.4 bb. Same pattern as on dry boards but with frequency variation (modest rather than negligible) added in.

Practical: on wet boards, c-bet most of the time but check more at s=1 and s=0 where own urgency is low. Size up materially when cliff-imminent.

§7.3 — Connected boards show the cliff

Setup: 8h7h6c (rainbow, connected, mid-card). Same configuration otherwise.

This is the connected-board cliff signature. C-bet frequency moves from 55% at hero=s=2 (the lowest) to 92% at hero=s=4 (the highest) — a 37pp swing. The entire pre-cliff-1 cohort (s=0,1,2) c-bets less than the post-cliff-1 cohort (s=3,4,5).

Why connected boards specifically? On dry boards range advantage dominates (auto c-bet). On wet boards equity equalizes (modest variation). On connected boards the range-vs-range dynamic is nuanced enough that hero's forward stake (proximity to next conversion) can tip the balance between betting and checking — and the math reads through.

Pre-cliff-1 hero on a connected board takes the cautious line — checking down marginal hands, saving aggression for the next pot. After crossing cliff 1 (s=3), hero's per-pot weight is 2× and the math flips toward c-betting more of the range.

The s=2 trough (55% c-bet) is the cliff-1-imminent hero playing especially passive on this board — preserving the ammunition for the cliff-2 push two pots out. Compare to s=4 (cliff-2-imminent) where hero c-bets 92% — when the +12 cliff is in front, hero pushes hard regardless of texture.

Framework attribution: Nick Petrangelo (idea-source). Nick's hero-last texture-inversion intuition from Stand-up predicted exactly this kind of texture-dependent cliff expression — the more nuanced the board, the more hero's tier urgency reads through.

§7.4 — Why texture matters

Three texture classes, three different cliff expressions. The pattern:

• Dry boards (auto-c-bet textures): range advantage dominates → frequency saturates → cliff expresses through sizing only.
• Wet boards (equity-equalized textures): equity dynamics dominate → modest frequency variation → cliff expresses through sizing + small frequency shift.
• Connected boards (nuanced range-vs-range): neither structural force dominates → cliff effect tips the balance → cliff expresses through both frequency and sizing.

The general principle: the cliff structure expresses postflop wherever the underlying chip-EV decision is close enough that squid-equity considerations can move it. On lopsided boards (dry, hero auto-bets; wet, hero defaults close to NLHE) the squid mechanic is overwhelmed by the chip dynamics. On boards where the chip decision is genuinely close, hero's tier becomes the tiebreaker.

§7.5 — Sizing patterns by hero state

Across all three textures, sizing scales with hero's accumulated weight. The dead zone at s=1 produces the smallest c-bet sizing on every texture; cliff-2-imminent s=4 produces the largest.

The dead-zone effect persists into postflop sizing. Even on dry boards where frequency saturates, hero's sizing differs by 2-3 bb between s=1 and s=4. Tier urgency is encoded in the chips-in-pot decision even when the play-or-check decision is forced.

§7.6 — Practical: read your tier first, then the board

The decision tree for c-betting in Hunt Progressive:

1. Read the board first.
2. Dry → auto c-bet, scale sizing by tier
3. Wet → c-bet most of the time, modest tier sensitivity
4. Connected → tier-conditioned: pre-cliff-1 plays cautious, post-cliff-1 pushes hard
5. Read your tier second.
6. Dead zone (s=1) → smallest sizing, highest check rate on connected boards
7. Cliff-2-imminent (s=4) → largest sizing, highest c-bet rate on every board
8. Post-cliff (s=3, s=5) → moderate sizing, full c-bet on most textures

⚠️ Val-conditioning: this entire chapter is at val=3. The texture taxonomy at val=1 and val=10 has not been tested. Direction is plausibly val-stable (the dry/wet/connected ordering should hold across vals because it's driven by board structure, not squid math), but magnitudes are open work. Logged in val-stability-audit.md as a Phase 8 follow-up.

⚠️ Coverage caveat: Slice 17 tested one board per texture class (Ah9c4d for dry, Ts9h8h for wet, 8h7h6c for connected). Robustness within texture class — does the cliff signature appear on other connected boards (754, 987, etc.)? — is open work. Pull 5 + Pull 9 supplement with vary-Z paradigm but don't add board diversity within texture.

Reader takeaway from Part 7. Postflop expresses the cliff structure on connected boards (30+ pp c-bet swing between pre-cliff-1 and post-cliff-1 hero), saturates on dry boards (frequency near 100%, sizing varies by tier), converges on wet boards (modest variation). Sizing always reads tier urgency: dead zone smallest, cliff-2-imminent largest. The earlier "postflop is largely tier-insensitive" framing was wrong on connected boards — texture matters.

Draft · Squid Hunt Progressive Part 7 · Pass 1 · 2026-05-02 · v0.1.0

Part 8 — Stack Depth, Val Asymmetry, and the Limits of the Framework

The earlier chapters built on val=3 and 100bb stacks as the canonical training values. This chapter is the second-order layer — what happens when val changes, when stacks change, and where the framework starts to break down.

§8.1 — The val=3 limp-pure peak (recap from Part 5)

Val=3 sits in the middle of the trained val range (the model trained on val ∈ {1, 2, 3, 5, 10}) and produces the sharpest expression of Hunt Progressive's cliff structure. Part 5 §5.2 walked through the limp-pure peak at val=3 fresh state — UTG/MP/CO all opening at near-100% VPIP with raise rates under 1%. This is val=3-specific.

The val=3 limp-pure regime is the squid mechanic at peak relative-importance vs chip-EV. Val=1 has the squid value too small to dominate (chip-EV regime); val=10 has the squid value so large that it justifies aggressive chip commitment (cliff-math regime). Val=3 sits in the band where cheap wide participation captures the most.

This isn't an accident of training. The val=3 setting was chosen as a canonical training value because it produces strategy that's distinguishably squid-driven without being warped by either edge. It's the "purest" Hunt Progressive value to anchor strategic teaching on.

§8.2 — The val=10 puzzle and its stack-depth resolution

At val=10, hero opens raise-heavy, not limp-pure. CO at val=10 fresh state raises 44% (vs 0.7% at val=3). Why? Two competing explanations:

1. The cliff math at val=10 is so large that hero pressures pots to capture the future cliff conversion. A +12 cliff at val=10 is 144 chip-units of squid-equity per pot — too valuable to leave to limped pots.
2. The 100bb stacks are too shallow for val=10 to play limp-style. With val=10's bigger implicit ante, the squid-equity-to-stack ratio is high — limping doesn't have enough chips behind to set up the squid math properly.

Pull 8 separates these explanations. Hero=s=3, val=10, Z=2 (high-leverage state), stacks varied:

The pattern reverses dramatically with stack depth. At 50bb hero is raise-heavy (60% raise, limp 18%). At 400bb hero is limp-pure (limp 91%, raise 9%).

So the val=10 puzzle is resolved by stack depth: at deep enough stacks, the squid mechanic returns even at val=10. The val=10-at-100bb shape is a stack-conditioned phenomenon, not a fundamental change to the cliff structure. The framework still applies; the chips-to-implicit-ante ratio just isn't right at 100bb.

This is the unifying result of Part 8: Hunt Progressive's strategy depends on (squid_pressure × stack_depth) / chip_pressure. When that ratio is high, the cliff structure dominates and hero plays the squid game. When it's low, chip-EV considerations dominate and hero plays closer to NLHE.

§8.3 — Val asymmetry on opens vs defense

A subtle finding: val asymmetry behaves differently on opens vs defense.

On opens (Part 5 §5.5): the val=3 limp-pure peak. Raise% lowest at val=3, highest at val=10.

On defense (Part 6 §6.7): the val=1 fold zone. Fold% appears only at val=1; raise rate climbs with val.

The two sides are mirror images. As val increases: - On opens: hero shifts from limp-heavy → raise-heavy - On defense: hero shifts from "some folds OK" → "always defend, sometimes raise"

The unifying read: as val increases, hero plays more aggressively in pots they're already in and more chip-committal on opens. The implicit ante grows, hero's reason to fight any pot grows with it, and the limp-pure middle ground (val=3 opens) gives way to chip aggression at val=10.

This is the same finding as Blood Battle — both accumulation modes produce the val=1 fold-zone-on-defense + val=10 raise-heavy-on-opens shape. Hunt Progressive's val asymmetry is a property of accumulation modes, not of HP specifically.

§8.4 — The unifying frame: (squid_pressure × stack_depth) / chip_pressure

Three forces compete in every Hunt Progressive decision:

• Squid pressure = implicit ante per pot, given by Uri's swing formula × hero's tier weight
• Stack depth = chips remaining behind to express the squid math through pot construction
• Chip pressure = standard NLHE pot odds + EV considerations on the chips themselves

When squid pressure × stack depth >> chip pressure, the squid mechanic dominates — limp-pure shapes, wide defense, cliff-imminent aggression. When it's reversed, hero plays close to NLHE Cash.

The val=3 100bb canonical state sits in the squid-dominant regime by design. Val=1 100bb shifts toward chip-dominant (small implicit ante, regular chip math). Val=10 100bb is in a weird hybrid (big implicit ante but not enough stacks to play limp-style — see §8.2). Val=10 400bb returns to squid-dominant (deep enough to construct the squid math even at high val).

The framework breaks down at the regime boundaries. At val=1 100bb, expecting limp-pure is wrong. At val=10 50bb, expecting limp-heavy is wrong. The cliff structure is always there, but its expression depends on whether the chips-to-ante ratio supports it.

§8.5 — What the framework doesn't cover

Hunt Progressive in this book is scoped to:

• val ∈ {1, 3, 5, 10} — the trained range. Behavior at val ∉ this range is OOD.
• Stacks ∈ {50, 100, 200, 400}bb — Pull 8 covers four points. Inter-stack behavior interpolated.
• 6-max only. The training and the model are 6-handed. Heads-up, 9-max, MTT-style stack distributions: untrained.
• Z ≥ 2. The model has only learned in-distribution states (Z = 1 is terminal; Z = 0 never arises). Decisions when only one or zero desperates remain: untrained.
• Default tier thresholds (3, 5). The AceSense product spec allows configurable thresholds; the trained model uses the default. Custom-threshold games are out of scope.
• Super Squid OFF, no first-hand double squid. Both are spec'd in the AceSense product but not implemented in training (see Part 1 §"Two known implementation gaps"). Both interact with the cliff thresholds in ways that would change strategy materially when implemented.

⚠️ Val=10 inversion in Phase 7. The Goldilocks finding (Part 6 §6.3) inverted at val=10 — peak responsiveness shifted from hero=s=3 to hero=s=1. We treated this as a possible training-edge artifact rather than a real strategic phenomenon, because val=10 sits at the top of the trained range. If this is artifact rather than feature, the val=10 stack-depth resolution above (§8.2) might also need re-checking. Logged as OQ-8.

⚠️ J-ρ paradigm pulls not yet run. Realistic-flow sampling (where the table state evolves through a sequence of hands rather than being held synthetic) hasn't been applied. Findings here are all from synthetic states (H-Z, O-Z, vary-val, vary-Z). The "down-to-two analog" state from Part 4 §4.7 specifically requires J-ρ or H-T sampling and is not yet covered in data.

Reader takeaway from Part 8. Val asymmetry, stack depth, and the cliff structure interact through (squid_pressure × stack_depth) / chip_pressure. Val=3 100bb is the squid-dominant canonical state. Val=10 100bb is a stack-shallow hybrid; val=10 400bb returns to squid-dominant. The framework is scoped to in-distribution training values; behavior at edges (val=10, Z=1) is partially artifact-prone.

Draft · Squid Hunt Progressive Part 8 · Pass 1 · 2026-05-02 · v0.1.0

Part 9 — Open Questions, Scope, Actionables

This is the closing chapter. Five things: the nine mechanisms recap, the four reads that matter most, the open questions logged for future work, what this book did NOT cover, and a six-rule cheatsheet.

§9.1 — The ten mechanisms

The book identified ten confirmed mechanisms across the campaign. Three are about transitions, two are about squid-economy positioning (pool side + amplification side), and the rest are supporting forces.

Transitions (Part 3): - HP-M1 (transition-imminent shifts). Aggression peaks at cliff-imminent states (s=2 wider, s=4 raise-heavy). Asymmetric across the three transitions (safety line is qualitative pool event, cliff 1 is +4 amplification, cliff 2 is +12 amplification). - HP-M2 (post-transition plateaus). Each transition has a distinct after state: dead zone at s=1 (post-safety-line), coasting at s=3 (post-cliff-1), grinding at s=5+ (post-cliff-2).

Squid-economy positioning (Parts 6 + 4): - HP-M5 (opener-side amplification — deny/commit/fold). Hero's defense response varies sharply with the opener's amplification potential. Cliff-1-imminent opener → deny; cliff-2-imminent opener → commit; past-peak opener → start folding. The three-zone shape is HP's signature, distinct from Blood Battle. - HP-M10 (off-pot amplification + amplification potential). The HP-M5 pattern generalizes beyond the opener. Any cliff-imminent player at the table reshapes hero's defense (Slice 18a: BTN at s=5 drops hero's raise from 88% to 49% even though CO is fresh). Hero responds to amplification potential (cliffs in front), not current amplification (cliffs already crossed) — a holder past s=5 inverts the pull (Slice 18d).

Within-state structure (Part 4): - HP-M3 (three-tier regime + within-tier structure). 1× / 2× / 4× weight tiers, with sub-state variation within each. - HP-M9 (s=1 dead zone — emergent). The lowest-incentive state in HP. Not predicted by the multiplier curve alone; requires the time-pressure framing.

Supporting forces (Parts 5, 6, 7, 8): - HP-M4 (Z gradient — modest on opens, sharp on defense). ~10pp swing on opens, ~45pp swing on defense across Z=2 to Z=5. - HP-M6 (val asymmetry — val=3 limp peak, val=1 fold zone). Open shape varies with val; val=3 is the limp-pure regime. Defense fold mass appears only at val=1. - HP-M7 (stack-depth amplification). Squid mechanic re-emerges at deep stacks even at high val (Pull 8: val=10 limp-pure at 400bb). - HP-M8 (cross-street consistency — texture-dependent). Cliff structure expresses postflop on connected boards (30+ pp swing); saturates on dry; converges on wet.

§9.2 — The six reads that matter most

If you only remember six things, remember these. The product gives you the right play.

1. Read your tier (Part 4). What's hero's squid count, and what tier does that put hero in? Each of the seven states is a different game.

2. Read your transition distance (Part 3). Are you at a transition (s=2, s=4)? Just past one (s=1, s=3, s=5)? In a no-transition zone (s≥6)? Distance to the next cliff drives urgency.

3. Read your opener's tier (Part 6 §6.3). Cliff-1-imminent opener? Deny. Cliff-2-imminent opener? Commit. Past-peak opener? Start folding marginals.

4. Read off-pot amplification potential (Part 6 §6.5b). Scan the whole table, not just the opener. A non-opener sitting at s=4 (cliff-2-imminent) pulls hero's defense toward call. A non-opener at s=5 (already past peak — no further amplification coming) pulls hero toward raise. Same opener, completely different play.

5. Read Z (Parts 5, 6). How many desperates remain? On defense the implicit ante quadruples from Z=5 to Z=2 — the same hand calls vs raises differently. On opens the Z effect is modest.

6. Read val (Parts 5 §5.5, 6 §6.7, 8). The val parameter sets the regime. val=3 is the limp-pure peak; val=1 puts BB defense into the only state where folding shows up in-distribution; val=10 inverts the Goldilocks pattern (peak shifts from hero=s=3 to s=1) and may need a stack-depth correction. Many findings are val-stable in direction across canonical vals (1, 3, 5) but val-conditioned in magnitude. See val-stability-audit.md before generalizing.

The six reads compose: tier × distance × opener's tier × off-pot amplification × Z × val. The full state space is large, but at the table you usually need 2-3 of these reads, not all six. The chapters give you the tools to know which read matters most in each spot.

§9.3 — The five Editor's Qs

Open work logged for future revisions. None are blockers for v1; all sharpen the book on follow-up.

OQ-1 — Pull 11 s=7 Z=2 raise%=90.8% anomaly. Real or out-of-distribution artifact? Want one confirmation pull at a different opp_state to triangulate. Status: open.

OQ-2 — s=1 dead zone interpretation. Need a tighter explanatory frame for Part 4 §4.2b. "Why post-safety-line low-incentive trough specifically" — the framing in this version is "no motives stacked," but the sharper version of the explanation might use Part 2's swing-formula math more directly. Status: open.

OQ-3 — Defense pattern at non-CO openers, fully tested. Phase 4 δ covered SB and BTN openers at hero=BB s=3. The cross-state generalization (other hero states × non-CO openers) hasn't been pulled. Status: partially resolved; full surface needs pulls.

OQ-4 — Cliff aggression × val ladder at hero=s=4 fully tested. Slice 16 covers s=4 across val=1, 3, 10 for opens. Defense-side coverage at s=4 across val ladder is in Phase 7 but not paradigm-clean to the §6.4 standard. Status: partially resolved.

OQ-5 (RESOLVED in Phase 5 Slice 17). Postflop at cliff-imminent hero — does the cliff signature appear postflop? Yes on connected boards; texture-dependent. Part 7 reframes around this finding.

OQ-6 — Triple-check the paradigm framework after data lands. The 5-paradigm taxonomy (H-Z, O-Z, H-Σ, J-ρ, H-T) was formalized 2026-05-01. Phase 5+ data was paradigm-tagged at runtime. A re-read of the framework against the now-complete Phase 7 data is overdue. Status: open.

OQ-7 (RESOLVED in Phase 6). Cross-mode comparison — is the deny/commit/fold pattern HP-distinctive or shared with other accumulation modes? Resolved: total magnitude is shared with Blood Battle (~75pp); the shape (specifically the COMMIT zone at cliff-2-imminent) is HP-distinctive. Documented in §6.3 callout.

OQ-8 — val=10 inversion follow-up. Phase 7 showed the Goldilocks pattern inverts at val=10 (peak shifts from hero=s=3 to s=1). Possible training-edge artifact. Need: a confirmation pull at val ∈ {7, 8, 12, 15} would clarify whether val=10 is the start of a real high-val regime change or just the model's training-distribution edge. Status: open.

OQ-9 (PARTIALLY RESOLVED in Phase 8). Off-pot amplification — does hero respond to non-opener players' tier? Resolved: yes, dramatically. Slice 18a shows BTN at s=5 drops hero's raise from 88% to 49% even though CO is fresh. Slice 18c shows concentrated-vs-spread holders produce opposite play. What's still open: does the magnitude scale with the off-pot seat's position relative to hero (MP vs BTN vs UTG) — currently only BTN tested. Plausibly position-invariant since hero's decision is hero-vs-opener and the off-pot player isn't in the pot, but unverified.

OQ-10 (RESOLVED in Phase 8 Slice 18c). Tier-asymmetric tables — does hero distinguish concentrated vs spread holder configurations? Yes — same opener, same Z, same hero produces 51% call vs 86% raise depending on holder distribution. This is the cleanest demonstration that amplification reading is load-bearing.

OQ-11 (PARTIALLY RESOLVED in Phase 8 Slice 18d). Down-to-two amplification — what does hero do at Z=2 with extreme tier asymmetry? Resolved: hero responds to amplification potential, not current amplification. A holder past s=5 inverts the pull (hero raises 67% vs 41% in spread). What's still open: intermediate configurations (e.g., two cliff-imminent holders, mixed past-peak + cliff-imminent) and the full surface across the holder configurations are not tested.

§9.4 — What this book did NOT cover

Scope acknowledgments — what's outside the v1 work and why.

Squid Hunt Regular (linear weight = s, codename SquidType::REGULAR). Engineering ticket pending; not yet trained. Will get its own book once trained. The "intermediate" Squid variant in the canonical pedagogical order — would sit between Stand-up (Book 2) and Hunt Progressive (this book).

Blood Battle (smooth quadratic, codename SquidType::BLOOD_BATTLE). Trained, separate scope, separate book (Book 7). Cross-mode comparisons in this book are pointer-only — see Book 7 for full Blood Battle treatment.

Stand-up Game (binary cap = 1, codename SquidType::CLASSIC). Book 2 covers it. Cited extensively for Uri's framework but the strategy details are in that book.

Super Squid + first-hand double squid. Both in the AceSense product spec, both not implemented in training (see Part 1). Both interact with cliff thresholds in ways that would materially change strategy when implemented. The book's findings are scoped to: Super Squid OFF, no first-hand bonus.

Configurable tier thresholds. Book covers default (3, 5) only. Different threshold configurations would shift the cliffs and reorganize the seven-state structure.

MTT/SNG Hunt Progressive. Out of scope — the trained model is for cash-style 6-max games.

PLO Hunt Progressive. Out of scope.

Z=0 / Z=1 strategic states. OOD — Z=1 is the game-end trigger, Z=0 never arises. Strategy at and below Z=1 isn't trained.

J-ρ paradigm sampling. The "down-to-two analog" in Part 4 §4.7 specifically requires realistic-flow sampling (state evolves through a sequence of hands). Not in current campaign. Logged for Phase 8.

§9.5 — The seven-rule cheatsheet

The whole book compresses into seven rules. Use this as the at-the-table reference.

1. Read your tier and distance to the next cliff first. Hero's squid count + position relative to s=3 and s=5 determines posture. (Parts 3, 4.)

2. At s=4 (cliff-2-imminent), play maximum aggression. Raise wide on opens (~48% at val=3, higher at val=10); commit through pots; sizing 5-7 bb. The +12 cliff is the largest single amplification event in any Squid variant. (Parts 3, 4, 5.)

3. At s=1 (dead zone), expect to do almost nothing. Lowest aggression state in HP. VPIP drops, raise rate near zero at val=3. Don't manufacture aggression here. (Part 4 §4.2b.)

4. On BB defense, read opener's tier before opener's hand. - Cliff-1-imminent opener (s=2)? Raise to deny — small cliff is cheap to fight. - Cliff-2-imminent opener (s=4)? Call to commit — big cliff is too expensive to fight. - Past-peak opener (s≥5)? Marginal hands lose value, start folding. (Part 6 §6.3.)

5. Scan the whole table for amplification potential, not just the opener. A non-opener at s=4 (cliff-2-imminent) pulls hero's defense toward call. A non-opener at s=5 (past peak — amplification already in the bag) pulls hero toward raise. Same opener, different play. (Part 6 §6.5b.)

6. Z drives the implicit ante. Defend tighter when Z is high (fresh game), commit more when Z is low (mid-to-late game). On opens the Z effect is modest; on defense it's a 45pp raise-rate swing. (Parts 5 §5.4, 6 §6.6.)

7. Postflop: dry boards auto-c-bet (sizing reads tier). Connected boards check the cliff first — pre-cliff-1 hero plays cautious, post-cliff-1 hero pushes hard. Wet boards are intermediate. (Part 7.)

Closing. Hunt Progressive's strategy lives at the intersection of NLHE chip math and a two-sided squid settlement event. The implicit ante (Part 2) is the unifying mental model; the seven states (Part 4) are the vocabulary; the three transitions (Part 3) are the moments of change; the squid economy — the future-loser pool (who pays) plus holder amplification (how much each loser pays) — is the source of value; reading amplification potential (cliffs in front, not cliffs already crossed) is the dominant skill; the cliffs at s=3 and s=5 are the structural facts that make Hunt Progressive Hunt Progressive.

Draft · Squid Hunt Progressive Part 9 · Pass 1 · 2026-05-02 · v0.1.0

Methodology Appendix — How the Data Was Collected and Compared

This appendix is for the reader who wants to know: how was each strategic claim in this book backed by data, and what does "comparing strategies across game states" actually mean in a game where multiple things vary at once?

If you're reading the book to learn how to play Hunt Progressive, you can skip this — the chapter sections each open with a plain-English setup line that tells you what's varying and what's held. This appendix unpacks the why behind those conventions.

The comparison problem

When a book says "hero plays X at state Y," it's holding several things implicit: - Hero's seat (BB, CO, etc.) - Hero's squid count - Opponents' squid counts (5 separate values) - Z = number of desperate players at the table - Val (per-squid chip value at game-end) - Stack depth - Action context (preflop open, defense, c-bet)

When we compare "hero at state Y vs state Y'," one of these axes varies. But the comparison only means something if we're explicit about which axis varies and what's held — otherwise the data point at Y' is about a different table situation than Y, and the comparison conflates two effects.

The default — hero state varies, table state held

The book's default convention: when a section says nothing about its setup, the comparison is varying hero's squid count while holding the table state at Z=3 fixed (three desperate opponents + safe-fillers), val=3, 100bb stacks, CO opens, BB defends.

This is the workhorse paradigm — it answers the question "how does hero's strategy change as hero crosses different squid states?" — and most chapters use it.

Named deviations

Five chapters ask different questions; each names what's varying explicitly in the setup line.

1. "Hero state varies, opponents in their starting state." Used in Part 5 sections that show real-game progression. Opponents are held at all-zero squids (the fresh state at game start); Z drifts as hero crosses the safety line. Tests "what does hero face at each squid count in a natural game arc" rather than "what's the isolated hero-state effect." Slight Z drift is part of the comparison — hero crossing s=0 → s=1 reduces Z by one as part of the natural trajectory.

2. "Opponent's tier varies, hero held." Used in Part 6 (the marquee defense chapter) where the question is "how does hero respond when the opener changes tier?" Hero is fixed at one squid state (s=3 in §6.3); the opener's squid count sweeps from 0 to 5; Z held at 2; the other opponents' composition holds to maintain Z. This is the only chapter where hero is the pinned variable and an opponent does the varying.

3. "Z varies, everything else held." Used in Part 5 §5.4 (Z gradient on opens) and Part 6 §6.6 (Z gradient on defense). Hero, val, stacks, opener position all held; Z varies from 2 to 5. Tests "how does the table-level desperate-count affect strategy."

4. "Val varies, everything else held." Used in Part 5 §5.5 and Part 6 §6.7. Hero, Z, stacks, opener position all held; val sweeps {1, 3, 5, 10}. Tests "how does the per-squid chip value affect strategy."

5. "Stack depth varies, everything else held." Used in Part 8. Hero, Z, val, opener position all held; stack depth sweeps {50, 100, 200, 400} bb. Tests "how does the squid-vs-chip pressure ratio shift with stack depth."

A sixth (realistic-flow joint sampling) is named for completeness but not yet used in the book — see "Open methodology questions" below.

Why the framework matters

Without paradigm discipline, two cells of "hero at state Y" can describe completely different table contexts and silently diverge. A coach reading "BB raises 80% vs cliff-1-imminent CO" needs to know: was the rest of the table held the same way that "BB calls 84% vs cliff-2-imminent CO" was measured? If yes, it's a clean comparison. If no, the comparison conflates hero-response with whatever else was changing.

The framework's job is to make that audit-able. Each pull that produced data for the book is paradigm-tagged at runtime — what's held, what varies, expected Z value — and Z-asserted per cell to catch silent drift. Pulls that don't conform get flagged in the audit (pulls/PARADIGM_AUDIT.md) and aren't quoted as primary evidence.

Pulls used in this book

The full audit (which cells of which pulls are safe, which need re-pull) lives in pulls/PARADIGM_AUDIT.md.

Open methodology questions (Editor's Q register)

A 4-agent methodology review on 2026-05-01 surfaced five concerns the book should triple-check before v1 ships:

1. Symmetry of safety-line transition. When hero crosses s=0 → s=1, the "fixed Z" paradigm requires shifting one opponent from desperate to safe to preserve Z. This produces a small asymmetry: at hero=s=0 the opp composition is slightly different from hero=s=1. The patterns we see are robust to this, but the methodology should formally check whether any chapter's claims depend on the asymmetry.

2. Realistic-flow sanity check. All current pulls hold Z, opp counts, or some abstraction "constant" — none sample from realistic game-progression distributions where Z, hero state, and opp composition co-move naturally. A future Phase batch could test whether the isolated-effect findings survive when realistic correlations are restored.

3. Z definition consistency across pulls. Some pull scripts count hero in Z; others don't. The current convention (Z = number of zero-squid players including hero if applicable) is enforced via runtime assertion in Phase 5+ pulls. Older pulls' Z labels need verification before being quoted.

4. The s=1 dead zone interpretation. Hero=s=1 has the lowest defense-response sensitivity to opp tier across all hero states. This is well-replicated in data but not derived from the mental model — it's an emergent finding. We should frame it as observed not explained until further data settles whether it's an in-distribution model property or an OOD artifact.

5. Per-position vs per-pair generalization. Most cross-state findings tested at hero=BB vs CO opener. Phase 4 δ verified the marquee finding at hero=BB vs SB and BTN openers (pattern holds, magnitudes shift). Other position pairs (CO opener × SB defender, etc.) untested. The book scopes claims to "BB defending CO" by default; cross-position claims get caveat.

These are tracked in research-base.md as Editor's Qs OQ-1 through OQ-6.

How to verify a specific claim against the data

For a sophisticated reader who wants to audit a claim:

1. Find the section in the book. Note its setup line — what's varying, what's held.
2. Find the corresponding pull in data/. The book's section frontmatter cites the source JSON file.
3. Find the cell(s) in the JSON corresponding to the claim's specific (hero_s, s_CO, Z, val, stacks) coordinates.
4. Check pulls/PARADIGM_AUDIT.md for that pull's cleanliness verdict. If ✅, the claim is paradigm-clean. If ⚠️, the claim may need a paradigm-respecting re-interpretation or a fresh pull to re-verify.

This audit chain is the book's substitute for citation; we don't link external solver outputs because the data was produced specifically for this book against the engineering team's preview rail. The cited JSON files plus the paradigm audit together let any coach or researcher verify the claims independently.

Methodology appendix · Squid Hunt Progressive · 2026-05-02 · v0.2.0 in flight

Book 8 — Squid Hunt Progressive

Working title: Squid Hunt Progressive: A Strategy Manual Subtitle (working): the implicit ante, the safety line, and two cliffs Estimated total length: ~13,800 words across 9 parts Authoring mode: QuintAce-led, with idea-source attribution to coaches (Dan + Uri + Nick); Uri's article cited directly, Dan's + Nick's articles not cited (in rebuild) Source data: data/pull_phase_2_full.out.json (13 axes) + data/pull_phase_3_bilateral.out.json (bilateral test) + data/pull_phase_4_lens_validation.out.json + data/pull_phase_5_paradigm_aligned.out.json (paradigm-clean) + data/pull_phase_6_validation.out.json (cross-mode + Goldilocks val-test) + data/pull_phase_7_responsiveness_surface.out.json (full hero × val × s_CO surface). See pulls/PARADIGM_AUDIT.md AND val-stability-audit.md before quoting. Codename / source folder: SquidType::DOUBLE / engineering-department/.../squid-double/ Mechanisms catalogued: 9 confirmed (HP-M1 through HP-M9) — see research-base.md

Structural choice — breaking parity with Blood Battle Book 7

Decision (2026-05-01): Hunt Progressive uses a 9-part shape, not Book 7's 8-part shape. The break is deliberate.

Blood Battle's smooth quadratic ramp doesn't have discrete strategic states — every s value is a slightly different point on a continuous curve. Per-tier content is less load-bearing in Book 7, and a single mental-model chapter would feel forced.

Hunt Progressive's structure is qualitatively different along three axes:

1. Three discrete transitions — the safety line (s=0 → s=1, status change: desperate → safe) and two weight-function cliffs (s=2 → s=3, +4 conversion; s=4 → s=5, +12 conversion). Each transition reshapes hero's strategy in stepped ways.
2. Within-tier internal structure — the seven hero states (s=0 head start, s=1 dead zone, s=2 cliff-1-imminent, s=3 post-cliff-1, s=4 cliff-2-imminent, s=5 post-cliff-2, s≥6 grinding) need vocabulary BEFORE chapter-level decisions can be discussed.
3. Coach framings transfer directly. Uri's "less zero-sum than NLHE" (his §6) and Uri's swing formula 6 × val / (N − 1) extend cleanly to Hunt Progressive. Capturing these as a standalone mental-model chapter (Part 2) lets the rest of the book cite back to it.

So the 9-part shape adds one new chapter vs Blood Battle's 8: a dedicated mental-model chapter (Part 2) before the structural-feature content (Part 3 — the three transitions). The per-state content (Part 4 — "The Seven States") moves earlier in the sequence so decision chapters (Parts 5-7) can reference within-tier vocabulary without re-introducing.

Parts (9-part shape)

Part 1 — What Is Squid Hunt Progressive (the primer)

Status: drafted (primer.md) ✅ Length: ~1,600 words Scope: Rules, end-conditions, tiered multiplier curve with two cliffs, payout math (with worked example reproducing AceSense product spec §III.1), terminology, three weight-function tiers with within-tier nuance preview (s=0 head-start + insurance, s=1 dead zone, s=2 cliff-1-imminent, s=3 post-cliff-1, s=4 cliff-2-imminent, s≥5 past-peak), implementation gaps. Axis 1 framing: as Z decreases, the implicit ante per pot grows (no "chip lead" language).

Part 2 — The Implicit Ante and the Future-Loser Pool ★ NEW (the mental-model chapter)

Mechanisms: none directly — establishes the framing every other chapter cites Length target: 1,800 words Scope: The thesis of the book. Hunt Progressive's strategy follows from one unified mental model: every pot pays an implicit ante — and that ante doesn't come from your opponent on this hand. It comes from whoever ends up the lone-loser at game-end (Uri §6). The squid game is "less zero-sum than NLHE" in a per-hand sense, with practical consequences.

Sections: - 2.1 The implicit ante per pot — Uri's swing formula. Swing = 6 × val / (N − 1), where N = number of desperates remaining. At fresh state (N=6) val=3: 3.6 BB per pot. At down-to-two (N=2) val=3: 18 BB per pot. The per-pot quantity is the SWING — what changes if hero wins this pot vs doesn't — not "current expected loss" (a different number that doesn't drive the decision). Cite: uri-squid-invisible-ante §1. - 2.2 Where the squid value comes from. Uri's §6 reframed: the squid prize doesn't come from your opponent's stack on this hand. It comes from the eventual game-loser. Per-hand chip transfers ARE zero-sum (chips conserved in the pot), but per-hand squid-equity transfer is NOT (the squid pays at game-end from a shared loser pool). This is the structural break from cash NLHE. Cite: uri-squid-invisible-ante §6. - 2.3 The "less zero-sum" corollary, applied to Hunt Progressive. In the same multi-way pot, a desperate player can profitably defend, a cliff-imminent player can profitably call wide, and a safe mid-tier player can profitably bluff — all simultaneously. Bluff equity comes from the future-loser pool, not from the calling opponent's stack. This contradicts the cash reflex of "deny equity from the calling station." Cite: uri-squid-invisible-ante §6. - 2.4 Hunt Progressive extension: the swing scales with hero's tier. At s<3 the per-pot swing for a desperate hero matches Stand-up's (Uri's formula). At s≥3 the swing scales because winning a pot at the 2× or 4× tier re-mints the entire accumulated stash, not just the new squid. The implicit ante grows with hero's accumulated weight. (Foreshadows Part 3: the cliffs are when the multiplier flips, redefining the per-pot swing for everyone holding squids.) - 2.5 What this mental model predicts (preview of the rest of the book). Reading the future-loser pool is the dominant skill. The transitions reshape the pool. Hero's tier and hero's opponents' tiers determine where the pool is forming. - 2.6 What the mental model does NOT predict. Hunt Progressive has within-tier internal structure (s=0 head start vs s=1 dead zone vs s=2 cliff-imminent are very different even though all are 1× tier) that's emergent in the data — not derivable from the mental model alone. Part 4 unpacks the seven states. - 2.7 How to read the data in this book — what we hold constant, what we vary. (Brief signpost.) Each chapter section opens with a one-line setup that says what's held and what varies — e.g., "hero=BB at s=3 throughout this section; Z=2 fixed; opener's squid count varies." The default if a section doesn't say is: hero state varies, table held at Z=3 fixed, val=3, 100bb. Anything else (varying Z, val, stacks, opp tier) gets called out in the setup line. The full methodology framework — the why behind these conventions, the formal naming, the paradigm-cleanliness audit of every pull — lives in the methodology appendix at the end of the book for readers who want the rigor. (See appendix; OQ-6 pending triple-check.)

Reader takeaway: every chapter from here cites this one. If you internalize this part, the rest of the book teaches you how to apply it.

Part 3 — Three Lines to Cross: The Safety Line and Two Cliffs

Mechanisms: HP-M1 (transition-imminent shifts), HP-M2 (post-transition plateaus) Length target: 1,700 words Scope: Hunt Progressive has THREE strategically distinct transitions, each reshaping hero's posture. The first is a status change (the safety line); the other two are multiplier conversions (the cliffs). The asymmetry across the three is the core insight.

Sections: - 3.1 The three transitions, three different mechanisms. Overview table + sequence chart from Pull 2 (raise%: 0.7 → 0.2 → 10.0 → 27.0 → 39.1 → 34.7 → 44.4 → 51.3). - 3.2 The safety line (s=0 → s=1) — status change, not multiplier change. At s=0 hero is the only player who can become the lone-loser; capturing the first squid removes that downside entirely. Two motives stack at s=0: insurance AND the head start to cliff 1. Hero plays max-participation (87.4% VPIP at CO val=3). ⚠️ Val-conditioned: the limp-pure shape (raise% < 1%) is val=3-specific; at val=1 and val=10 hero=s=0 raises 37-44%. The max-participation finding (high VPIP) is val-stable, but the "limp-pure" character is the val=3 form. - 3.3 Cliff 1 (s=2 → s=3) — the +4 conversion. Modest activity rise pre-cliff (raise% climbs from 0.1% at s=1 to 10.0% at s=2 to 27.0% at s=3). Limp-heavy approach because the cliff is small enough that wide cheap participation captures it. - 3.4 Cliff 2 (s=4 → s=5) — the +12 conversion. Peak aggression in the entire game. Raise% peaks at 47% at s=4 Z=2 with 6.5bb sizing. The denomination upgrade for hero's accumulated stash is the largest in any Squid family variant. Hero can't afford NOT to commit when the +12 is one pot away. - 3.5 The post-transition plateaus. s=1 dead zone (post-safety, no immediate cliff in sight — covered in depth in Part 4). s=3 (post-cliff-1, coasting toward cliff 2). s=5+ (post-cliff-2, accumulation grinding). - 3.6 Why the three transitions are asymmetric. Marginal-value math: safety line is a status flip (qualitative); cliff 1 is +4 (quantitative, modest); cliff 2 is +12 (quantitative, decisive). The proportional jump on cliff 2 is the largest discrete strategic event in Hunt Progressive. - 3.7 How transitions reshape the future-loser pool (Part 2 reference). The safety line removes hero from the pool. The cliffs re-mint the pool's payout structure: when an opponent crosses a cliff, ALL their accumulated weight upgrades, shifting which players are most likely to be the future-loser.

Part 4 — The Seven States — Hero's Squid Count as a Strategic Variable

Mechanisms: HP-M3 (three-tier regime + within-tier structure), HP-M9 (s=1 dead zone), HP-M2 (post-transition plateaus) Length target: 1,800 words Scope: Every hero squid count is a different game. Walks through s=0,1,2,3,4,5,6+ with their distinct incentives. Establishes the within-tier vocabulary that decision chapters reuse.

Sections: - 4.1 Three weight-function tiers — the math foundation. 1× / 2× / 4× weight multipliers, recap from Part 1. - 4.2 Within Tier 1: three sub-states. - 4.2a s=0 (desperate / safety-line-imminent): two motives stack — insurance against lone-loser AND head start to cliff 1. Maximum participation. At val=3 (canonical training value) the shape is limp-pure (raise%=0.7%, raise_bb=7.10). ⚠️ Val-conditioned: at val=1 and val=10, hero=s=0 raises 37-44% — not limp-pure. Phase 5 Slice 16 confirms; see val-stability-audit.md. - 4.2b s=1 (the dead zone) — HP-M9. Post-safety, no immediate cliff in sight. The lowest-incentive state in the game: marginal value of the next squid is just +1, and the cliff at s=3 is still two pots away. Phase 3 confirmed: tiny bilateral response (~3pp swing across modes vs 49pp at hero=s=3). Hero coasts. Don't expect aggression here the way you would from s=0 or s=2. - 4.2c s=2 (cliff-1-imminent): one pot from the +4 conversion. Activity rises to 91% VPIP with raise%=10%. - 4.3 Within Tier 2: two sub-states. - 4.3a s=3 (post-cliff-1): limp-heavy participation (62.7% limp), coasting as ammunition for cliff 2. The +2 marginal of the next squid is small; hero saves chip aggression for the next push. - 4.3b s=4 (cliff-2-imminent): peak aggression in the entire game. The +12 cliff is one pot away — raise%=47%, raise_bb=6.5. Hero plays tighter (VPIP 85% vs s=2's 91%) but more aggressive per pot. - 4.4 Tier 3 (s ≥ 5): past-peak grinding. Each additional squid is +4 weight (flat). Most selective. Strategy converges toward chip-EV with a small accumulation bonus. - 4.5 The seven-state cheatsheet. A one-page summary table reader can use at the table. - 4.6 How the seven states map to the future-loser pool. Lower hero count = closer to being the future-loser; higher hero count = further from it. Hero's state directly modulates how aggressively hero positions relative to the pool. (Part 2 reference.) - 4.7 The "down-to-two" analog — Hunt Progressive shape. Framework attribution: Nick Petrangelo (idea-source). In Stand-up, the down-to-two state is when only two players are still desperate. In Hunt Progressive, the analog is when the future-loser pool collapses to one or two candidate players (high accumulated weight on one side, low on the other) — a state that combines low Z with extreme tier asymmetry.

Part 5 — Preflop Opens

Mechanisms: HP-M4 (Z gradient on opens), HP-M6 (val asymmetry — val=3 limp peak) Length target: 1,600 words Scope: Per-position widening, the val=3 limp-pure shape, the cliff signature across positions.

Sections: Default paradigm: H-Z (vary hero state, Z=3 fixed) unless noted. Subsections that vary other axes are labeled.

• 5.1 The widening — Hunt Progressive vs the three baselines. [H-Σ paradigm — fresh state, all opps at 0.] Pull 1 cross-mode table (NLHE / Stand-up / Blood Battle / Hunt Progressive). Hunt Progressive widens every position +42 to +56pp vs NLHE.
• 5.2 The val=3 limp-pure peak. [H-Σ at fresh state.] Distinguishing finding from Blood Battle: at val=3 fresh state, raise% < 1% at UTG/MP/CO. Pure-limp strategy. Why? At val=3 the per-pot swing is at peak relative to chip-EV; cheap participation maximizes capture. Cite: uri-squid-invisible-ante §3a (the limp-return finding extends here from Stand-up).
• 5.3 The transition signature replicates across positions. [H-Z default at Z=5, opps fresh.] Pull 2 + 3a UTG/CO/BTN sequences — same s=4 raise-peak shape across seats.
• 5.4 Z gradient on opens. [Vary-Z paradigm — Z varies explicitly from 5 down to 2; hero=s=3, val=3 held.] Modest ~10pp swing from Z=5 to Z=2. The implicit ante grows but the open shape stays similar.
• 5.5 Val asymmetry on opens at fixed Z=3. [H-Z paradigm × val ladder — Phase 5 Slice 16.] At Z=3, val=3 the limp-pure peak holds at hero=s=0,1,2; at val=10 hero=s=4 raises 65.8%, the cliff-2-imminent peak across vals.
• 5.6 The "raise to size of squid" heuristic — Hunt Progressive caveat. Cite: uri-squid-invisible-ante §3b. Uri's heuristic predicts UTG opens to ~3.6 BB at val=3 in Stand-up. In Hunt Progressive the solver opens larger across all positions because the cliff structure adds a forward jackpot premium. Caveat the heuristic but keep the spirit.
• 5.7 Practical: the opener seat decision tree. Tier-by-tier, position-by-position default opens.

Part 6 — BB Defense — Reading the Squid Economy ★ MARQUEE CHAPTER

Mechanisms: HP-M5 (opener-side amplification — deny/commit/fold), HP-M10 (off-pot amplification + amplification potential) Length target: 1,800 words Scope: The single most striking finding in the campaign. Hero's defense response varies sharply with opener's tier proximity: - vs cliff-1-imminent opener (s_CO=2): BB raises 79.5% — DENY the cliff - vs cliff-2-imminent opener (s_CO=4): BB CALLS 83.5% — COMMIT to seeing a flop - vs past-peak opener (s_CO=5): BB starts FOLDING (13.8%)

Strategic frame (per Part 2): the squid value doesn't come from your opponent on this hand — it comes from the eventual game-loser. So hero isn't "exploiting CO" when CO is desperate; hero is positioning relative to where the future-loser pool is forming. Small cliff (+4) → deny is cheap and the pool stays open. Big cliff (+12) → commit because the pool reshuffles dramatically if CO converts. Past-peak opener → CO won't be the future-loser, so marginal hands lose value.

Default paradigm: O-Z (vary opener tier, hold hero+Z). This chapter inverts the book's H-Z default because the marquee finding is hero's response to varying opponent tier. Subsections that use H-Z (vary hero) are labeled.

Sections: - 6.1 BB defends 100% in-distribution. [H-Z — Pull 6, hero state varies, opps fresh.] Fold range is essentially empty across in-distribution states. Connects to Uri's §3d (BB defense doubles in Stand-up) — Hunt Progressive saturates further. Cite: uri-squid-invisible-ante §3d. - 6.2 The transition signature on defense — peak raise at hero=s=2 (82.9%). [H-Z — Pull 6.] - 6.3 The deny/commit/fold pattern (Pull 12 — the marquee finding). [O-Z paradigm: hero=s=3 fixed, opener s_CO ∈ {0..5} varies, Z held. The chapter's central finding requires varying opp.] - 6.3a vs CO at s=0: hero pressures (87.8% raise). Both players have strong forward stake in the pool. - 6.3b vs CO at s=2 (cliff 1 imminent): hero denies (79.5% raise). Small cliff is cheap to deny. - 6.3c vs CO at s=4 (cliff 2 imminent): hero commits (83.5% call). Big cliff means hero must be in the pot when CO crosses. - 6.3d vs CO at s=5+ (past-peak): hero folds 13.8%. CO's weight is locked in 4× tier, won't be future-loser. - 6.3e Why the pattern flows from future-loser-pool positioning, NOT from "exploitation" (per Part 2's framing). - 6.4 Hero-state generalization (Phase 4 + 5). [Hybrid: H-Z × O-Z — sweep hero state at fixed s_CO, then sweep s_CO at fixed hero=s=4.] The deny/commit/fold pattern's SHARPNESS is hero=s=3-specific. Vs cliff-2-imminent opener, most hero states (1, 2, 3, 4) shift to call-heavy — only s=0 keeps raising. Hero-state changes the magnitude, not the direction. - 6.5 Cross-position validation. [O-Z at hero=BB s=3, opener position varies.] Pattern holds vs CO and BTN openers; vs SB-opener less aggressive overall but still commits at s_SB=4. - 6.6 Z gradient on defense — much sharper than on opens (~45pp swing). [Vary-Z paradigm — Pull 7, val=3 hero=BB s=3.] - 6.7 Val asymmetry on defense — val=1 is the only state where BB folds in-distribution. [Vary-val paradigm — Pull 7.] - 6.8 The "less zero-sum" corollary applied to defense. Cite: uri-squid-invisible-ante §6. The cash reflex of "deny equity from the calling station" is partially wrong here — bluff equity comes from the future-loser pool. A desperate caller and a safe bluffer can both have profitable plays in the same pot. - 6.9 The defense decision tree — four cells. cliff-1-imminent / cliff-2-imminent / past-peak / standard.

Part 7 — Postflop — Texture-Dependent Cliff Expression

Mechanisms: HP-M8 (refined Phase 5 — texture-dependent, not null-direction) Length target: 1,400 words (upgraded from 1,000) Scope: The cliff structure DOES express postflop — but only on connected/middling boards. Dry boards saturate; wet boards converge with modest variation; connected boards show the full cliff signature with a 30+ pp swing in c-bet frequency between pre-cliff-1 and post-cliff-1 hero states. This is a substantive chapter, not a null-result calibration.

Default paradigm: H-Z (vary hero state, Z=3 fixed, BB pinned at s=1, val=3) × board axis. Three boards tested.

Three-board taxonomy (Phase 5 Slice 17 — H-Z × board sweep):

Sections: - 7.1 Dry boards saturate. Ah9c4d: 97-99% c-bet across all hero states. Range advantage dominates; the cliff effect washes out. Sizing varies modestly with hero state (s=1 dead zone bets ~4bb; cliff-2-imminent s=4 bets ~11bb), but frequency is invariant. - 7.2 Wet boards converge with modest variation. Ts9h8h: 84-91%, no clear cliff signature. Built-in equity equalization on monotone-flush boards squashes the hero-state effect. - 7.3 Connected boards show the cliff. 8h7h6c: pre-cliff-1 hero c-bets 55-75%; post-cliff-1 c-bets 87-92%. The s=2 (cliff-1-imminent) state is the LOWEST c-bet at 55% — pre-cliff hero on a connected board takes the more cautious line, saving aggression for the next pot. After crossing cliff 1 (s=3), c-bet jumps to 87%. Framework attribution: Nick Petrangelo (idea-source) — his hero-last texture-inversion intuition from Stand-up predicted exactly this kind of texture-dependent cliff expression. - 7.4 Why texture matters. Dry boards: range advantage so dominant that hero auto-bets regardless. Wet boards: equity equalization. Connected boards: range-vs-range dynamics are nuanced enough that hero's forward-stake (proximity to next conversion) tips the balance between betting and checking. The cliff structure reads through. - 7.5 Sizing patterns by hero state. Across all three textures, sizing scales with hero's accumulated weight: s=1 dead zone bets smallest (4-7bb), cliff-2-imminent s=4 bets largest (11-14bb). The dead zone effect persists into postflop sizing. - 7.6 Practical: read your tier first, then the board. On dry boards, just c-bet. On connected boards, hero state determines frequency materially — pre-cliff-1 hero plays it cheaper. On wet boards, modest hero-state effect; default close to NLHE.

Part 8 — Stack Depth, Val Asymmetry, and the Limits of the Framework

Mechanisms: HP-M6 (val asymmetry), HP-M7 (stack-depth amplification) Length target: 1,400 words Scope: Val and stack-depth as second-order effects on top of the tier framework.

Sections: Default paradigm: vary one axis at a time (val OR stack OR Z), holding others at default (hero=s=3, val=3, 100bb, Z=3). Each subsection is a one-axis sweep around the H-Z baseline.

• 8.1 The val=3 limp-pure peak (recap from Part 5). [Vary-val on opens at hero=s=3 fresh.]
• 8.2 The val=10 puzzle and its stack-depth resolution. [Vary-stack at val=10 hero=s=3 Z=2 — Pull 8.] At val=10 100bb stacks chip-EV dominates (47% raise); at 200-400bb stacks the squid mechanic returns (limp 92%).
• 8.3 Val asymmetry on opens vs defense. [Vary-val × position-pair at hero=s=3.] CO peaks at val=3; BB defense peaks at val=1. Same finding as Blood Battle.
• 8.4 The unifying frame: (squid_pressure × stack_depth) / chip_pressure. Squid mechanic dominates when this ratio is large.
• 8.5 What the framework doesn't cover. val ∈ {1, 3, 5, 10}, stacks ∈ {50, 100, 200, 400}, 6-max only.

Part 9 — Open Questions, Scope, Actionables

Mechanisms: synthesis chapter Length target: 1,300 words Scope: - 9.1 The 9 mechanisms recap. One sentence each. Three transitions come first (HP-M1, HP-M2 — the safety line + two cliffs); future-loser-pool positioning second (HP-M5); within-state structure third (HP-M3, HP-M9 dead zone); supporting forces (HP-M4, M6, M7, M8) last. - 9.2 The four reads that matter most. Read your tier (Part 4). Read your transition distance (Part 3). Read your opponents' tiers (Part 6). Read Z (Part 5). The product gives you the right play. - 9.3 The five Editor's Qs. Pull 11 s=7 anomaly, why s=1 is more passive than s=0, defense pattern at non-CO opener, val ladder at s=4 cliff, postflop on cliff-imminent hero. - 9.4 What this book did NOT cover. Z=0/Z=1 (OOD), Squid Hunt Regular (untrained), Blood Battle (separate book), Super Squid + first-hand (untrained — amplified in Hunt Progressive), MTT/SNG/PLO Hunt Progressive. - 9.5 Cheatsheet — 6-rule action list.

Coach citation map

Attribution policy for Dan + Nick. The currently published Dan and Nick articles cover Stand-up Game (aka Squid Classic) and are being rebuilt to also cover Squid Hunt Progressive. For this book, we attribute Dan's and Nick's framings as idea-sources where they appear and do NOT cite or link the published articles. Once their rebuilds land, citation upgrades happen as a refresh pass.

Uri attribution policy. Uri's article is cited directly. His uri-squid-invisible-ante is Stand-up by content, but the framings (implicit ante, future-loser pool, less-zero-sum, raise-to-size-of-squid) explicitly extend to accumulation modes per his own §6 caveat.

Mechanism → Part mapping

What's NOT in the book

• Squid Hunt Regular (untrained, future book)
• Blood Battle (Book 7, paused)
• Stand-up Game (Book 2)
• Super Squid + first-hand double squid (untrained — amplified in Hunt Progressive)
• Configurable tier thresholds — book covers default (3, 5) only
• MTT/SNG/PLO Hunt Progressive — not in scope

Phase 5 drafting order (recommended)

1. Part 2 first — anchors the unified mental model. Everything else cites it.
2. Part 3 — establishes the three transitions (data-driven counterpart to Part 2's framework).
3. Part 4 — sets up within-tier vocabulary that decision chapters need.
4. Part 6 — the marquee defense chapter. Uses Parts 2 + 3 + 4 fully.
5. Part 5 — preflop opens. Lighter, draws on Part 4 vocabulary.
6. Part 7 — postflop texture-dependent (refined Phase 5). Three-board taxonomy with the connected-board cliff finding. Substantive chapter (1,400w), not the short null-result it was originally framed as.
7. Part 8 — limits.
8. Part 9 — synthesis last (depends on everything above).

Estimated drafting time: 4–5 weeks at ~3-4 days per part for parts 2-6, ~2 days per part for parts 7-9.

Editor's Qs to resolve before / during Phase 5

• OQ-1: Pull 11 s=7 Z=2 raise%=90.8% anomaly. Real or OOD artifact? Want one confirmation pull at different opp_state.
• OQ-2: s=1 dead zone interpretation. Need a one-paragraph explanatory frame for Part 4.2b — why post-safety low-incentive trough specifically.
• OQ-3: Defense pattern at non-CO opener. Pull 12 only tested CO opener. Phase 3 follow-up: SB and BTN opener variants.
• OQ-4: Cliff aggression × val. Pull 11 only at val=3. One val-ladder pull at s=4 would clarify whether the 47% raise rate holds at val=1, 5, 10.
• ~~OQ-5: Postflop at cliff-imminent hero.~~ ✅ RESOLVED in Phase 5 Slice 17. Postflop DOES express the cliff on connected boards (8h7h6c) — pre-cliff-1 hero c-bets 55-75%, post-cliff-1 87-92%. Dry boards saturate; wet boards converge. Part 7 reframed as texture-dependent.

These are NICE TO HAVE — none are blockers for v1 drafting. Defer to Phase 3 follow-up pull batch if/when needed.

Open infrastructure items

1. Engineering ask (Slack later): VERSION file in squid-double/ upstream + production-checkpoint training-coverage confirmation
2. flagship-8.shell.html template (clone from flagship-7) — needed for tools/build.py integration
3. specs/sections/hp-part-N-*.section.md per-section specs — 9 specs needed (one per part)
4. books.yaml Book 8 sections list — updated to 9 entries with new titles

Outline · Squid Hunt Progressive · 2026-05-01 · Phase 3 — restructured 9-part shape, three-transitions framing, idea-source attribution policy · pre-Phase-5 drafting

Squid Hunt Progressive — Research Base

Live document. Findings flagged as CONFIRMED (clean data, replicable) or OBSERVED (one pull, needs corroboration). EMERGENT = mechanism that wasn't in the original predictions and surfaced from the data.

Headline findings

1. The 4→5 cliff produces the sharpest aggression spike in the entire game. At s=4 Z=2 val=3 CO, raise%=47.0% with 6.50bb sizing — the most aggressive state we measured. Drops to 26.2% at s=5 (post-cliff plateau). ✅ Val-stable in direction: Phase 5 Slice 16 confirms cliff-2 peak at hero=s=4 across val=1/3/10 (47% / 49% / 66%); magnitude grows with val.
2. Within each tier, the cliff-imminent state plays distinctly from the post-cliff state. s=2 (about to cross cliff 1) plays differently from s=3 (just crossed). s=4 (about to cross cliff 2) plays differently from s=5 (just crossed). Pull 2 + Pulls 10–11 confirm.
3. EMERGENT — s=1 is the most passive state in Hunt Progressive. Raise%=0.1% at CO val=3 — less active than desperate (s=0). Not predicted from the rules. Likely the "low-leverage waiting state" — too far from cliff 1 (+4) to push, but already safe from being lone loser. ✅ Val-stable in direction: Slice 16 confirms s=1 trough at val=1 (17.7%), val=3 (0.1%), val=10 (17.5%). Sharpness peaks at val=3.
4. EMERGENT — Hero's defense response inverts between cliff-imminent opponents. vs CO at s=2 (cliff 1 imminent): BB raises 79.5% (deny-the-cliff). vs CO at s=4 (cliff 2 imminent): BB CALLS 83.5% (commit-the-call instead — bigger reason to see a flop). vs CO at s=5+ (past-peak): BB even folds 13.8%. Pull 12 — strongest single mechanism in the campaign. ✅ Direction val-stable, magnitude val=3-conditioned: Phase 7 surface shows the deny→call swap holds at val=1/3/5/10 with raise-swing magnitudes +34/+64/+47/+14 pp. Sharpest at val=3; weakens at val=10 but does not invert.
5. Postflop is largely tier-insensitive. Dry A94r: cbet 99-100% across Z. Wet T98ss: cbet 79-86%. The tier framework lives preflop, not postflop. ⚠️ Texture taxonomy NOT TESTED across val — see val-stability-audit.md.

Phase 0 — Sanity pull (2026-05-01)

Confirmed: rail returns visibly cliff-aware behavior; no deployment gap. See Phase 0 history below + headline finding 1.

Phase 2 — Full pull campaign (2026-05-01)

13 axes, 141s runtime. Full results: data/pull_phase_2_full.out.json. ⚠️ Mixed-paradigm pull — see pulls/PARADIGM_AUDIT.md for row-by-row guidance before quoting.

Pull 1 — Per-position × val × mode comparison

CONFIRMED: Hunt Progressive (DOUBLE) plays distinct from Blood Battle (BLOOD_BATTLE) across all positions and vals.

Notable Hunt Progressive vs Blood Battle differences at fresh state val=3:

OBSERVED: At val=3 fresh state, Hunt Progressive collapses to near-pure-limp (raise% < 1% at UTG/MP/CO). Blood Battle still raises 3-10% across positions. The cliff structure pushes hero to maximum cheap participation — every limped pot is a step toward s=3 → s=5 cliffs.

OBSERVED: Val=1 is narrower in Hunt Progressive than Blood Battle. UTG val=1: HP 34.4% vs BB 65.5%. MP val=1: HP 40.7% vs BB 67.1%. The +12 cliff at s=4→5 only matters when val is high enough to make the chip reward worth chasing — at val=1 the cliff payoff (1×12=12 chips) is too small to reorganize strategy.

Pull 2 — Per-hero-squid-count at CO val=3, opps fresh

CONFIRMED: The strategic shape changes dramatically across squid counts. The cliff structure dominates.

CONFIRMED — HP-M1 (cliff-imminent aggression). Raise% sequence: 0.2% (s=1) → 10.0% (s=2) → 27.0% (s=3) → 39.1% (s=4) → drops to 34.7% (s=5). Two distinct climb-and-pause patterns.

CONFIRMED — HP-M2 (post-cliff plateau). s=5 plays calmer than s=4. s=3 plays calmer than s=2 in absolute aggressiveness (raise size drops from 6.60 to 4.40bb).

EMERGENT — HP-M9 (s=1 passivity trap). s=1 plays MORE PASSIVELY than s=0 (desperate). Lowest raise% in the table. Not predicted from rules. Hypothesis: at s=1 hero is already "safe from lone loser" but has nothing to leverage — too far from cliff 1 to push for it.

Pull 3a — Replication at UTG and BTN

CONFIRMED: Cliff signature replicates across positions.

UTG raise% sequence: 0.1 (s=0) → 0.1 (s=1) → 5.3 (s=2) → 21.6 (s=3) → 31.9 (s=4) → 28.5 (s=5) → 34.5 (s=6) → 39.3 (s=7). BTN raise% sequence: 3.2 (s=0) → 1.3 (s=1) → 11.8 (s=2) → 31.9 (s=3) → 48.8 (s=4) → 43.6 (s=5) → 53.7 (s=6) → 60.3 (s=7).

Same shape: s=1 dip, climb to peak at s=4, slight retreat at s=5. BTN's peak at s=4 is 48.8% — even more dramatic than CO's 39.1%.

Pull 4 — Val ladder × Z, hero=s=3 CO

CONFIRMED — HP-M6 (val asymmetry):

• val=1 Z=2: limp 33.8%, raise 37.5%, raise_bb 7.20bb (LOW val: hero plays raise-heavy + big sizing)
• val=3 Z=2: limp 81.0%, raise 13.0% (canonical val: limp-heavy)
• val=10 Z=2: limp 27.5%, raise 47.6% (HIGH val: chip-EV dominates, raise-heavy)

The val=3 limp peak is preserved. Below and above val=3, hero shifts to raise-heavy. This matches Blood Battle's val=3 peak finding.

Pull 5 — Postflop dry (A94r) × Z

CONFIRMED — postflop dry saturates (similar to Blood Battle):

OBSERVED — sizing climbs ~2bb across the Z gradient on dry boards. Smaller effect than preflop, but real.

Pull 6 — BB defense per hero count, val=3 opps fresh

CONFIRMED — HP-M1 expressed on defense, with cliff signature at s=2:

EMERGENT NOTE: the cliff-imminent peak on DEFENSE is at s=2 (not s=4 like on offense). At s=4 the defense is actually CALMER than s=2. This is the inverse of the open frequencies. Possibly because at s=4 hero needs to call to commit to the pot (chase the +12 cliff via showdown), while at s=2 hero needs to raise to deny opponent (the +4 cliff is small enough that deny-the-cliff matters more than commit-to-cliff).

Fold% is 0.0% across every state. BB defends 100% in Hunt Progressive across all in-distribution states.

Pull 7 — BB defense × val × Z (hero=BB s=3)

CONFIRMED — HP-M4 (Z gradient on defense):

As Z increases (more co-desperates), defense shifts call → raise. Strong gradient (~45pp swing) on defense compared to ~10pp on opens.

Pull 8 — Stack depth at val=10, Z=2, hero=s=3

CONFIRMED — HP-M7 (stack-depth amplification):

Same pattern as Blood Battle. At deep stacks, the cliff incentive dominates chip-EV.

Pull 9 — Postflop wet (T98ss) × Z

CONFIRMED — wet boards show modest Z gradient:

cbet 79.4% (Z=2) → 86.4% (Z=5). 7pp range. Sizing climbs 11.96 → 13.20bb across Z.

Pull 10 — Cliff 1 transition (s=2→3)

CONFIRMED — HP-M1 around s=2:

At Z=2 val=3 CO: s=1 raise% 0.1% → s=2 raise% 3.4% → s=3 raise% 13.0%. Modest transition (cliff 1 marginal is +4 — small).

Pull 11 — Cliff 2 transition (s=4→5) — THE BIG CLIFF

CONFIRMED — HP-M1's strongest expression:

At Z=2 val=3 CO: - s=3: VPIP 94.0% raise% 13.0% raise_bb 7.20bb - s=4: VPIP 96.7% raise% 47.0% raise_bb 6.50bb ← peak aggression - s=5: VPIP 83.4% raise% 26.2% raise_bb 5.30bb (post-cliff plateau) - s=6: VPIP 83.6% raise% 38.3% raise_bb 4.90bb - s=7: VPIP 100% raise% 90.8% raise_bb 7.20bb (extreme — flag for verification)

Editor's Q: s=7 at Z=2 raise%=90.8% is anomalous — possible OOD edge case (very rare state in training). Pull 11 row 13 needs cross-check before being cited as a mechanism. May be model artifact rather than real strategy.

Pull 12 — Opponent-cliff-imminent (THE BLOCKBUSTER)

CONFIRMED — HP-M5 (future-loser-pool positioning):

Hero=BB s=3 vs CO opens. Vary CO's squid count s_CO:

The deny/commit/fold pattern. Three distinct responses to opponent's cliff distance: - vs CO at cliff 1 (s=2): hero raises 79.5% — small cliff (+4) is cheap to deny via raise pressure. - vs CO at cliff 2 (s=4): hero CALLS 83.5% — big cliff (+12) means if CO crosses to s=5, the future-loser-pool reshuffles dramatically. Hero needs to be in the pot to capture some of the conversion equity. - vs CO past-peak (s=5+): hero starts folding 13.8% — CO's accumulated weight (4× tier) means CO isn't going to be the future-loser. Marginal hands lose value, hero folds.

Driver (refined per Uri's §6): hero is positioning relative to where the eventual-loser pool is forming, NOT capturing bilateral equity directly from CO on this hand. The squid-equity transfer at game-end comes from whoever ends up lone-loser, not from CO's stack on this specific pot. So hero's strategic adjustment to CO's tier is about reading the future-pool trajectory, not about "exploiting" CO.

This is the single most striking finding in the campaign and the centerpiece of Part 4.

Pull 13 — Tier comparison sweep at Z=3, val=3, CO

CONFIRMED — HP-M3 (three-tier regime), with internal structure:

• 1× tier (s=0,1,2): raise% sequence 3.7 → 0.1 → 6.9 (clear s=1 dip)
• 2× tier (s=3,4): raise% sequence 26.2 → 48.5 (cliff-imminent peak at s=4)
• 4× tier (s=5,6,7): raise% sequence 37.6 → 49.2 → 56.9 (no s=anomaly here, monotonic)

The three tiers are clearly distinct, BUT the within-tier structure is what carries the strategic content — especially s=1 vs s=2 within Tier 1 and s=3 vs s=4 within Tier 2.

Confirmed mechanism catalog

After Phase 2 + Phase 3 bilateral-equity test, the mechanism list is:

9 mechanisms total. Two earlier candidates (HP-M9 renamed/refined, HP-M10 merged into HP-M5) refined after Phase 3 bilateral-equity test (see Phase 3 section below).

Framing decision (2026-05-01, refined)

After Phase 3, we're keeping the two-lens framing PLUS the s=1 dead zone teaching point — but with the per-hand zero-sum claim corrected per Uri's §6.

1. The implicit ante (Dan/Uri's mental model) — squid-equity adds a forward-looking term to every pot. Hero plays wider because the per-pot prize is larger. The squid value doesn't come from your opponent on this hand — it comes from the eventual game-loser (Uri's §6). Per-hand chip transfers are zero-sum (chips conserved in the pot); per-hand squid-equity is not zero-sum (the squid pays at game-end from a shared loser pool, not from a specific opponent on this hand); game-end aggregate is zero-sum.
2. The two cliffs (data-derived) — discrete conversion events at s=3 (1× → 2×) and s=5 (2× → 4×). Cliff 2 is the dominant force; cliff 1 is moderate.
3. The s=1 dead zone (HP-M9) — strategic teaching point, not a separate force. The trained model treats s=1 as the lowest-stakes hero state in the game: own marginal value of next squid is +1 (no cliff), AND response to opponent's count is minimal. Reader cue: don't expect aggression from a player at s=1 the way you would from a player at s=0, s=2, or s=4. The s=1 player has captured insurance and is now coasting until cliff 1 is in reach.

Phase 3 tested whether a separate "bilateral squid-equity" or "exploitation premium" should be a third organizing lens. Strict tests failed (val non-monotonic, hero=s=1 swing tiny, mode comparison weak). Why the bilateral framing was wrong: per Uri's §6, the squid-equity transfer is NOT a per-hand bilateral exchange between hero and a specific opponent. Hero's squid-equity gain doesn't come from villain's stack on this hand — it comes from whoever ends up the lone-loser at game-end. The deny/commit/fold pattern in Pull 12 reflects hero positioning relative to where the future-loser pool is forming, not direct equity capture from CO. Same data, sharper frame.

Useful corollary (from Uri's §6 applied to Hunt Progressive): in the same multi-way pot, a desperate player can profitably defend, a cliff-imminent player can profitably call wide, and a safe mid-tier player can profitably bluff — all simultaneously. The cash-poker reflex of "deny equity from the calling station" is partially wrong in Hunt Progressive, because hero's bluff equity comes from the future-loser pool, not from the calling opponent's stack on this specific hand.

Open questions / Editor's Qs to resolve before Phase 5

• OQ-1: s=7 Z=2 raise%=90.8% anomaly (Pull 11). Is this a real strategic finding or a rare-state OOD model artifact? Want a confirmation pull at s=7 with different opp_state to test.
• OQ-2: Why does s=1 play more passively than s=0? Need a brief interpretive frame for the primer/Part 6.
• OQ-3: Defense-cliff-inversion (HP-M10) at non-CO opener. All Pull 12 data is BB vs CO. Does the inversion hold when SB or BTN opens? Could be a future pull.
• OQ-4: Does s=4 cliff aggression hold across all val tiers? Pull 11 only tested val=3. Worth a small val ladder at s=4.
• OQ-5: Postflop on cliff-imminent vs post-cliff hero. Pull 5 + Pull 9 all hero=s=3. Does cliff structure express postflop at s=4 vs s=5? Could be one focused pull.
• OQ-6: Revisit comparison-paradigm framework after Phase 5 data lands. ⚑ Editor's Q (methodology, double + triple check needed). The current 2-axis taxonomy (H/O/J × Σ/Z/T/ρ → 5 useful paradigms) was distilled from a 4-agent review on 2026-05-01. The reviews surfaced that the original A/B/C framing had a category error (C was vary-opp, not vary-hero). The refined framework is internally consistent but has a known asymmetry at the safety-line transition (hero=s=0 vs s=1+ requires different opp redistribution to maintain fixed Z). After Phase 5 data lands, check (a) whether the safety-line asymmetry produces misleading patterns in any cell, (b) whether to add a stricter H-T paradigm (Z + tier composition both held), and (c) whether to add a Pull 18 batch sampling joint hero/opp distributions from realistic game progression (Paradigm J — flagged as missing by methodology critic). Triple-check before any Phase 5 drafting cites paradigm-labeled data.
• OQ-7: ✅ RESOLVED 2026-05-02 (Phase 6 cross-mode pull). Magnitude is NOT distinctively HP. Blood Battle shows ~75pp spread; HP shows ~74pp; both are accumulating-squid modes. NLHE shows ~17pp (across opener position); Stand-up ~11pp (across CO's safe-vs-desperate state). The dramatic adaptation is a property of accumulation modes generally. Distinctively HP is the SHAPE: the deny/commit/fold three-zone with COMMIT mode at cliff-2-imminent. Blood Battle shows pressure-plateau-drop with no commit-mode at s_CO=4 (raise 76% / 75% / 74% / 9% across s_CO=2/4/5) because BB's smooth quadratic doesn't have the discrete +12 cliff at s=4→5 (only +9 marginal). §6.3 prose updated to reflect this. The "three zones" pattern is HP's signature; total magnitude is a property of accumulation as a class.
• OQ-8: Val-dependence of the inverse-U responsiveness pattern. ⚑ Editor's Q (data + theory). Partially resolved 2026-05-02 (Phase 7 surface). Phase 7 mapped the full hero × val × s_CO grid (56 cells). Goldilocks at canonical training vals (1, 3, 5) is supported: hero=s=3 is the peak responsiveness state at all three. Magnitude peaks at val=3 (~64pp), shrinks at val=5 (~47pp) and val=1 (~34pp). val=10 inverts: hero=s=1 becomes peak (~32pp); hero=s=3 drops to ~14pp. The val=10 inversion is the only val tier where the pattern shifts. Hypothesis: val=10 is the highest val in training distribution; the inversion may be a training-edge artifact rather than a real strategic phenomenon. Remaining open: (a) is val=10's inversion a training-edge effect or a real game-theoretic feature? Test would require pulling at val>10 (OOD by definition, so might not be measurable) or checking whether the inversion persists in a different rail/training run. (b) The hero=s=0 anomaly (negative swings at val=1 and val=10) is unexplained — desperate hero plays MORE aggressively vs cliff-2-imminent CO than cliff-1-imminent at extreme vals. Possibly an own-urgency × val interaction. Logged for v0.3 follow-up.

Outline refinement implications

The data strongly supports the planned 8-part shape, with significant content-level changes:

Part 2 — "The Two Cliffs" needs reframing

Original frame: "two cliffs at s=3 and s=5, hero plays aggressively before each."

Refined frame: the two cliffs produce DIFFERENT strategic behaviors: - Cliff 1 (s=2 → 3): small marginal (+4). Activity rises, but raise% stays modest. Limp-heavy participation. - Cliff 2 (s=4 → 5): big marginal (+12). Activity AND aggression peak here. Both raise% and sizing dominant.

The two cliffs aren't symmetric — they're qualitatively different beasts. Part 2 should be structured around that asymmetry.

Part 6 — "Per-Tier Strategy" needs emergent-mechanism integration

Add HP-M9 (s=1 passivity trap) — make it a named subsection. The passivity at s=1 is counterintuitive and reader-valuable.

Part 4 — "Preflop Defense (BB)" needs major upgrade

Add HP-M10 (defense-cliff-inversion) as the centerpiece finding. This is the most striking single finding in the campaign — the binary defense regime change between cliff-imminent opponents. Current Part 4 is a thin chapter; with this finding it becomes a marquee chapter.

Part 7 (formerly Part 5) — postflop chapter — REVISE FROM NULL-RESULT TO TEXTURE-DEPENDENT (Phase 5)

Earlier framing was "postflop converges across hero states; the cliff lives preflop, not postflop." Phase 5 Slice 17 disproves the strong version of this claim. On dry (A94r) boards, frequencies do converge to 97-99% across hero states. On wet (T98ss), modest variation. But on connected/middling (8h7h6c) boards, the cliff signature expresses clearly: pre-cliff-1 hero (s=0,1,2) cbets 55-75%; post-cliff-1 (s=3,4,5) cbets 87-92%. A 30+ pp swing.

The chapter should be reframed from "null result" to "texture-dependent": - Dry boards: frequency converges, sizing varies modestly - Wet boards: frequency converges (88% range), sizing varies - Connected boards: cliff signature visible in BOTH frequency and sizing

This makes Part 7 a more substantive chapter than the null-result framing suggested. Length target should bump back up from 1,000w to 1,300-1,400w.

New emergent content for Part 8 (synthesis)

Two emergent mechanisms (HP-M9, HP-M10) deserve highlighting in the actionables. The defense-cliff-inversion in particular is the kind of insight that's both counterintuitive and operationally usable.

Research base · Squid Hunt Progressive · 2026-05-01 · Phase 2 complete (13 axes, 10 mechanisms)