Squid Classic's Last Two: The Endgame Where the Button Stops Being the Best Seat

Nick Petrangelo coaches at GTO Lab and is one of the most-decorated tournament players of the modern solver era. The article you're reading walks the most concentrated game state in Squid Classic — when only two players still need to win a pot, the implicit ante peaks, and the cash-NLHE intuition for "what to do at this seat" stops mapping cleanly. This is Article 2 of the Squid Foundations series; Dan Dvoress walks the implicit-ante mental models in Article 1.

§0 — Squid Classic in sixty seconds + the down-to-two state

If you know the format, skim. If you don't, read once and you're caught up.

Squid Classic is 6-max No-Limit Hold'em with one extra rule. Every time you win a main pot, you collect a "squid" — a win token, capped at one per player. In a 6-handed game there are exactly five squids to win. The game ends the moment five of the six players hold one. The player who never won a main pot pays everyone else: 5 × val BB, where val ∈ {1, 2, 3, 5, 10}. This article uses val = 3 throughout (the standard training default). At val = 3, the loser pays 15 BB.

Two labels: safe (already holds a squid; can't be the loser) and desperate (no squid yet; still in the race).

The Squid penalty pays an implicit ante into every pot — equal to your share of the loser-risk × the penalty. A desperate player at the start of a fresh game has 1/6 of the loser-risk; at val = 3 the implicit ante per pot is 1/6 × 15 = 2.5 BB. As opponents become safe, your share concentrates. At "down to two desperate" — when 4 players have already won pots and only 2 are still in the loser race — your share is 1/2. The implicit ante peaks at 1/2 × 15 = 7.5 BB per pot. Bigger than most pots people fight for.

This article studies that game state. Three pairings are possible at down-to-two depending on which seats happen to be the two desperates: BvB (SB+BB), BTN+SB, and CO+BTN. They share the same squid math (loser-risk = 1/2, ante = 7.5 BB) but differ structurally in blind composition — how much of the visible pot is desperate-money vs safe-player dead money before any voluntary action begins.

The progression — desperate-money in the pot 1.5 → 0.5 → 0 BB; safe-player dead money 0 → 1 → 1.5 BB — drives the equilibrium differences across pairings. We'll see why in §§2-4.

§1 — Hook: a 70-point swing in one game-state transition

You're in the SB with K♠ J♣. CO opens 2.5 BB. BTN folds. UTG and MP have already folded preflop — they're safe. You're desperate. BB is safe (no implicit ante to widen with) and folds at NLHE rate. Action's on you.

In NLHE, the SB faces a CO open with a structured calling/3-betting range. KJo is on the bubble — sometimes a call, sometimes a fold, depending on opener tendencies. The solver's verdict: SB defends 20.2% of all hands facing this open. KJo makes the cut, but most of the SB's range folds.

In Squid Classic at val = 3, all six players still desperate, the SB's defending range widens to 43.9% — the implicit ante closes the pot-odds gap on a swathe of marginal hands.

Now consider the same spot in the down-to-two state. UTG, MP, CO, BB are all already safe. BTN and SB are the only two desperates left, with one squid still to award. The penalty hangs over them like a guillotine — whichever of them wins the next significant pot becomes safe; the other one pays 15 BB.

The SB's defending range jumps to 90.1%.

Cash → all-desperate Squid → down-to-two: 20.2% → 43.9% → 90.1%. The +46-point jump from all-desperate to down-to-two is the largest single-step widening in the dataset. Half of SB's range is fold in NLHE, fold in all-desperate Squid, defend in down-to-two. Those are the hook hands — hands the implicit ante was sized too small for at all-desperate but big enough to flip at down-to-two.

By §7 you'll see the same shape across three pairings, plus three different reasons each pairing distorts equilibrium in its own way.

§2 — BvB: blind-vs-blind, both desperates already in the pot

The first pairing to walk is the simplest blind structure. UTG, MP, CO, BTN are all already safe and fold preflop. SB and BB are the only two desperates. Before any voluntary action, both desperates already have skin in the pot — SB has 0.5 BB posted; BB has 1 BB posted. 1.5 BB of the visible pot is desperate-money. There is no safe-player dead money in this pairing.

Yes, BvB has positional asymmetry too — BB has position postflop, SB acts first preflop. The point is not symmetry of position; it's that both desperates are already invested before any voluntary decision happens. This makes BvB the saturation case — neither side has to actively decide whether to enter the pot; they're already in.

Preflop equilibrium across regimes:

By down-to-two BvB, both seats play 100% any-two-cards. The range plateau is hit. The only thing that grows from all-desperate to down-to-two is sizing — BB's raises trend toward 15 BB on average. Range work has nowhere wider to go.

The §2 takeaway: BvB is the saturation pairing. Both desperates are pre-committed and play any-two-cards. There's no marginal range left to widen. The information BvB gives us is the ceiling of what range-widening can do — useful as a baseline against the asymmetric pairings in §§3-4.

§3 — BTN+SB desperate: BTN opens, SB defends with the blind discount

Now the desperate pair includes one blind and one non-blind. UTG, MP, CO are safe and play their NLHE chip-EV ranges. BTN (desperate) opens or folds. SB (desperate, with 0.5 BB already posted) defends or folds. BB is safe — no implicit ante — and plays their NLHE defending range (NOT auto-folding); facing BTN open + SB action, BB folds at NLHE rates roughly 60% of the time and defends in the remaining ~40% (multiway scenarios out of scope).

Preflop equilibrium data is at each desperate's decision node — BTN's open rate and SB's defense rate are computed accounting for the full game tree, including BB's possible defense. They are not conditioned on safe blinds folding. The postflop data in §5 (and the "dead money" framing) conditions on the heads-up branch. On that HU branch, the pot starts with 0.5 BB of desperate-money (SB's blind), ~1 BB of dead money (BB's posted blind that ends up folded), and BTN's voluntary 2.5 BB open.

Preflop equilibrium:

The cleanest two-step pairing in the dataset. BTN opens cash 43% → all-desp 67% → down-to-two 99%; SB defends cash 20% → all-desp 44% → down-to-two 90%. Each step adds ~25-50 points of range. Both seats show large jumps at the down-to-two transition.

Two structural notes worth pulling out:

BTN's open range adds limps progressively. In NLHE, BTN doesn't limp here. By all-desperate, 30% of BTN's playable range limps. By down-to-two, half of BTN's any-two-cards range limps. Limping is the cheapest way to claim the implicit ante across the long run — and at down-to-two, when nearly every hand has the ante attached, half of those hands prefer to limp than to commit chips by raising. (For why limping specifically rewards low-EQR hands, see Article 1's discussion of equity realization and limping.)

SB's defense gets a 0.5 BB discount on calls. This matters at down-to-two because the SB's pot odds for calling already include the discount; once the implicit ante is added, the SB defends almost any-two-cards. The discount is small (0.5 BB) but compounding with the 7.5 BB ante, it makes most marginal SB hands clearly +EV to play.

The §3 takeaway: BTN+SB is the cleanest two-step. Both seats show large jumps cash → down-to-two, and the all-desperate-to-down-to-two transition adds the biggest single-step widening (SB +46 points). This is the article's hook spot for a reason.

⚑ Editor's note — model-bias caveat on absolute sizes and widths (review with Dan + Nick before final): The current solver model differs from popular reference solvers (PioSolver, GTO Wizard) in two related ways. (1) Absolute raise/bet sizes — BB avg raise 9 → 15 BB in §2, postflop avg bet sizes in §5 — are inflated against reader intuition calibrated to older sizings. (2) Absolute opening / defense widths — CO 28.1% open, BTN 16.0% defense vs CO 2.5x — are tighter than reference solvers typically show by ~5-15 percentage points. In both cases, the relative shifts (NLHE → all-desp → d2t widening; sizing escalation) reflect real Squid mechanisms and survive the model bias; only the absolute baselines should be cross-checked against your training tool of choice.

§4 — CO+BTN desperate: dead money becomes the prize fund

Last pairing. UTG, MP are safe and play NLHE chip-EV ranges (often folding). CO (desperate) opens. BTN (desperate) defends. SB and BB are safe — no implicit ante — and play their NLHE defending ranges (they are NOT auto-folders; they defend their blinds at standard NLHE rates against the action they face). At d2t in this pairing, the safe blinds fold ~70-85% of the time (SB facing CO+BTN action defends ~5-15% from OOP; BB facing CO+BTN+SB action defends ~20-35%). The remainder of the time, one or both blinds defend and the pot becomes multiway — a scenario this article does not analyze.

Two layers of conditioning to keep separate. The preflop equilibrium below (CO open rate, BTN defense rate) is at the actor's decision node — the solver computes BTN's optimal play accounting for all subsequent branches, including SB/BB defending. Preflop data is not conditioned on safe blinds folding. The postflop data in §5 (and the "dead money" EV framing) explicitly assumes safe blinds folded — the HU postflop scenario is what those numbers describe. Multiway postflop scenarios are out of scope.

So the framing splits: preflop, the safe blinds may or may not defend, and the desperate pair's strategy already prices that. Postflop, the data is HU-only — and on that HU branch, the visible pot is 0 BB of desperate-money (neither desperate has posted), ~1.5 BB of dead money (the safe blinds' posted-then-folded chips), and CO's voluntary 2.5 BB open.

Preflop equilibrium (BTN's response is at BTN's decision node — it is the same regardless of whether SB/BB later defend or fold; the multiway branches affect what happens after BTN acts, not BTN's choice itself):

CO's open rate triples cash → down-to-two; BTN's defense rate roughly quintuples. Both seats widen aggressively because each has the implicit ante on every winnable pot, and on the heads-up branch the 1.5 BB safe-blind dead money pads the pot odds for both.

The §4 takeaway: CO+BTN is the dead-money fight on the heads-up branch. When the safe blinds fold (~70-85% of the time), the posted blinds become a prize fund the desperates fight over; that's the scenario this article walks through §5. When the safe blinds defend, the dynamics shift to multiway and warrant a separate analysis.

Preflop ranges across all three pairings — the widget below visualizes opener and defender ranges for each pairing across cash, all-desperate, and down-to-two. The range structures back the prose: BvB collapses to identical at d2t; BTN+SB stays asymmetric (BTN keeps high-card structure, SB widens to nearly any-two); CO+BTN both wide but neither saturated.

§5 — Postflop: range advantage drives frequency, nut advantage drives sizing

Standard poker theory: c-bet frequency tracks range advantage (whose range has more equity overall on the board); sizing tracks nut advantage (whose range has more of the very strongest hands). Apply the lens to the three pairings at down-to-two and the postflop direction split falls out cleanly — different range structures produce different outcomes.

Equity realization (EQR) is also load-bearing here. When ranges widen at down-to-two, players don't realize all their preflop equity. A hand that defends in Squid has the same raw equity as in NLHE but realizes far less postflop — gets pushed off, makes weak pairs, fails to continue against pressure. Lower EQR means the aggressor extracts more per BB invested.

The data on K72r dry (NLHE → all-desp → d2t):

All three rise on dry boards. Magnitudes: BvB +20.5pp, BTN+SB +40.5pp, CO+BTN +16.5pp. The same mechanism drives all three — defenders widen on K-high dry, but the widening adds offsuit junk that has nothing on this texture (no flush draws, no straight draws, just dominated kickers and underpairs). Aggressor's range advantage GROWS as defender widens. C-bet frequency UP. Sizing also climbs, modestly.

The magnitude differences across pairings track range-structure asymmetry. BvB: ranges become identical at d2t (both 100% any-two-cards). Range advantage = 0; the rise is driven by the implicit ante on every won pot, not range advantage. Modest +20.5pp. BTN+SB: BTN's opening range stays more high-card-structured than SB's defending range. Range advantage holds. Biggest rise (+40.5pp). CO+BTN: both ranges are wide (CO 85% open, BTN 86% defense); CO retains some structural edge but less than BTN does over SB in the BTN+SB pairing. Smallest rise (+16.5pp).

The data on 7♠6♠5♦ wet two-tone (NLHE → all-desp → d2t):

Wet board: all three pairings DROP at d2t. Net cash → d2t: BvB -11.6pp, BTN+SB +1.6pp (basically flat), CO+BTN -44.7pp (the largest drop). Same mechanism in reverse: on connected wet textures, defenders catch the texture across the entire equity spectrum (sets, two-pair, straights, draws) — the widening adds hands that DO have equity. Aggressor's range advantage SHRINKS. C-bet frequency DOWN.

But when c-bets do come, the betting range is heavily polarized — and a polarized range carries a specific bluff-to-value ratio (B:V) at a given size: roughly 1:3 at half-pot, 1:2 at pot-size. The combination of "frequency down + B:V structure" pushes sizes up dramatically. Avg bet sizes balloon at d2t on 765two: BvB SB 2.2 → 8.7 BB; BTN+SB BTN 4.4 → 9.2 BB; CO+BTN CO 2.6 → 8.3 BB — all roughly 4× the NLHE baseline. Sizing scales with polarization.

CO+BTN's largest drop (-44.7pp) reflects the highest cash baseline (77.0% in NLHE — CO c-bets very wide on 765 in NLHE because BTN's narrow NLHE defense keeps the texture friendly to CO's range). When BTN's defense widens to 86% at d2t, CO's range advantage on this texture evaporates faster than for the other pairings — hence the bigger drop.

The unified explanation, in canonical theory: range advantage drives c-bet frequency; polarization (B:V structure) drives sizing; equity realization (low for the wide defender) is what lets aggressors size up. Direction is texture-driven, not pairing-driven — wet textures shrink range advantage; dry textures grow it. Magnitude varies across pairings because each pairing's cash baseline and range structure differ, but all three respond to the same mechanism in the same direction within a given texture.

MDF and bluff-frequency math — the standard "defenders over-defend MDF in Squid → bluffs decrease in optimal frequency" story — is real but downstream. Yes, defenders over-defend MDF because the implicit ante closes their pot-odds gap. But that's the consequence of range advantage, not the cause. Read range first.

(The model-bias caveat from §3 applies to all postflop sizings cited here — the absolute BB numbers are bigger than the previous-generation model would have shown.)

§6 — Per-seat EV: which seat is the most valuable?

Natural question after walking the three pairings: if you could pick which seat to be desperate at, which would you pick?

The cash-NLHE habit is clear: the BTN. Position pays. Closes the action postflop. Has the most information per decision. In Squid down-to-two, the answer is more interesting.

Per-seat per-hand EV at equilibrium, in big blinds:

How to read this table. Each cell is the seat's forward-looking expected value (EV) per hand dealt, in big blinds. The opener's EV is the solver's avg EV at the opener's first decision (which already integrates over all subsequent play). The defender's EV is computed by integrating across the opener's strategy mix: P(opener raises) × def-vs-raise EV + P(opener limps) × def-vs-limp EV + P(opener folds) × defender's sub-game EV.

Convention caveat — read before using these numbers. Each cell is EV as the solver reports it at each seat's first decision node. Two things to know about that:

1. The EV is forward-looking from each seat's first decision. Sunk blind costs are not subtracted (the blind is already in the pot when the seat makes their first decision).
2. The EV is NOT literal "chip flow per hand." A pure chip-flow metric in HU NLHE would sum to ~0 (zero-sum minus rake); these numbers don't. The solver's EV reporting includes the squid-equity term plus an implicit baseline reference (e.g., relative to "always fold" or some passive default), so absolute sums across two seats reflect "value of optimal play" rather than "expected chips at hand end."

What this means for using the table: within-pairing comparisons (BB vs SB; SB vs BTN; BTN vs CO) are robust — both seats share the convention, so the bias cancels and the d2t edges (+6.21, +2.78, +0.11) are reliable. Absolute magnitudes shouldn't be read as "X BB of chip flow per hand"; they're "X BB of EV as the solver reports it." A clean chip-flow recomputation is on the v2 backlog.

Three findings, each surprising in its own way.

1. At BvB down-to-two, the Big Blind is the most valuable seat by a wide margin. Big Blind EV = 9.99 BB/hand vs Small Blind EV = 3.78 BB/hand — a +6.21 BB/hand advantage to BB. The reason is information and position: BB acts after SB at every decision (preflop after SB's open/limp/raise; postflop closing the IP role). When both players are playing 100% any-two-cards (range plateau), the 1.5 BB invested in blinds redistributes asymmetrically because BB sees SB's action before deciding. The 7.5 BB implicit ante per pot just amplifies whatever positional edge exists.

2. At BTN+SB down-to-two, the Small Blind beats the Button by a clear margin. Small Blind EV = 4.27 BB/hand vs Button EV = 1.49 BB/hand — a +2.78 BB/hand advantage to SB. This is counterintuitive. The standard cash-NLHE assumption is that BTN's positional advantage dominates. The data says otherwise.

The reason: at down-to-two, BTN opens 99% (essentially any-two-cards), splitting roughly 50/50 between raise and limp. SB's response leverages aggression — high 3-bet frequency against BTN's raises (SB avg raise size 17.5 BB vs BTN's 7.2 BB open) and a wide-defending strategy against BTN's limps. SB's preflop aggression extracts heavy value from BTN's wide opening range — BTN can't continue against 3-bets with most of its hands and gets folded out. SB also benefits when BTN folds (which happens ~1% of the time at d2t): SB inherits the BvB-with-safe-BB sub-game and plays it profitably. The positional advantage that helps BTN postflop is overwhelmed by the preflop-aggression dynamic in down-to-two.

3. At CO+BTN down-to-two, the seats are nearly symmetric — the BTN edge is only 0.11 BB/hand. Cutoff EV = 1.20 BB/hand vs Button EV = 1.31 BB/hand. The two pairings are nearly equally positioned: both desperates are non-blinds without skin in the pot before voluntary action; both have to claim the safe blinds' 1.5 BB dead money via aggression. CO opens first, BTN defends with closing position — these advantages cancel. The remaining 0.11 BB edge to BTN is the pure positional residual once the d2t mechanics dominate.

The §6 takeaway: "the BTN is the most valuable seat" is a cash-NLHE habit that doesn't survive Squid down-to-two in two of three pairings. In BvB, the seat that closes the action wins (BB, by +6.2 BB/hand). In BTN+SB, the seat that aggressively responds to the wide opener wins (SB, by +2.8 BB/hand). Only in CO+BTN does the BTN's positional edge come through — and barely (+0.1 BB/hand). Re-grade your seat-quality intuition for desperation-state spots; the BTN is no longer the default best seat.

§7 — The pattern: one mechanism, three positional responses

Three pairings, one mechanism (loser-risk = 1/2, implicit ante = 7.5 BB at val = 3). All three pairings respond to the same concentration of squid-equity, but the magnitude of each response correlates with several positional factors at once — pre-action desperate-money invested, conditional safe-blind dead money, and range-structure asymmetry. With only three data points across these correlated factors, we describe the differences rather than attribute them to a single causal driver.

BvB: both desperates already invested in the blinds (1.5 BB desperate-money, 0 BB dead). Saturation case — both play 100% any-two-cards at d2t. Range structures collapse to identical → no range advantage either way. Postflop: c-bet climbs modestly on dry boards (+20.5pp on K72r); drops on wet boards (-11.6pp on 765two) because there's no theoretical reason to bet often when both ranges are identical. When bets do come, sizing is dramatically up (B:V structure of polarized betting range). BB > SB by 5+ BB/hand — closes the action.

BTN+SB: intermediate (0.5 BB desperate-money, conditional 1 BB dead when BB folds at NLHE rate). Range structures asymmetric — BTN preserves opening structure; SB widens to defend nearly any-two. Postflop: biggest rise on dry boards (+40.5pp on K72r); roughly flat on wet boards (+1.6pp on 765two — the inverted-U cash → all-desp → d2t pattern flattens out). SB > BTN by 2.7 BB/hand — aggressive 3-betting captures the dead money + ante, eating BTN's positional edge.

CO+BTN: the no-skin pairing (0 BB desperate-money, conditional 1.5 BB dead when both safe blinds fold at NLHE rate). Range structures both wide (CO 85% / BTN 86%) but BTN defends without a blind discount. Postflop: modest rise on dry boards (+16.5pp on K72r); biggest drop on wet boards (-44.7pp on 765two) — the high cash baseline (77.0% in NLHE, since BTN's narrow NLHE defense keeps 765 friendly to CO's range) gives the most room to fall as BTN widens. CO ≈ BTN — closest to symmetric per-position EV.

The unified picture:

Read across the rows: the same loser-risk concentration produces three different responses; magnitudes correlate with several positional factors that vary together across the pairings. Postflop direction is texture-driven (wet drops, dry rises) across all three pairings — it's not pairing-specific. Magnitude scales with cash baseline and range-structure asymmetry: pairings with higher cash baselines have more room to drop on wet, pairings with sharper range-structure asymmetry have more room to rise on dry.

The "blind structure causes the differences" framing is one descriptor among several correlated factors. With only three pairings and several confounded variables, we describe the correlations and stop short of a single-cause attribution.

§8 — Closer

When the loser race narrows to two, the geometry of cash poker collapses in three different ways depending on who's left.

In BvB, both players have already paid their preflop premium and play any-two-cards — there's nowhere wider to go, the bettor's range polarizes, and the seat that closes the action takes the lion's share of the EV.

In BTN+SB, the SB takes BTN's chips with 3-bets the BTN can't fight — the implicit ante swallows the positional advantage that a cash-NLHE player would have leaned on.

In CO+BTN, the safe players' blinds become a prize fund both desperates fight over preflop — but on wet boards, both desperates' wide ranges catch the texture and the aggressor's range advantage collapses just like in the other pairings. Texture decides the postflop direction; the pairing decides the magnitude.

Three structurally different geometries from one structural change. The chip-EV math you've used your whole career still works — but the geometry is no longer cash. The implicit ante is the same in all three; the blind structure determines what the ante does to the equilibrium.

Read the blind structure first. Range advantage second. The frequency and sizing fall out from those two.

Methodology

⚑ Note on safe-player behavior: Safe players don't auto-fold. They have no implicit ante, so their decision tree collapses to standard NLHE chip-EV — they play NLHE defending ranges (~40% BB defense, ~16% BTN defense, etc.). The pairings this article walks (BvB, BTN+SB, CO+BTN) occur when action folds to the desperate pair, which happens at the cumulative NLHE folding probability for those scenarios. When safe players DO enter the pot, the dynamics are multiway and out of scope here. The article's analysis is conditional on action folding to the desperate pair.

Source: book-2 (Squid Classic) v1.8.0 — squid mechanic, val parameterization, "safe" / "desperate" labels.

Data rail: gameplay-ai range-viewer preview endpoint via strategy_grid_client.py. Defaults: 100 BB stacks, 6-max, 2.5 BB opens. Custom SquidConfig.counts arrays construct each "down to two desperate" pairing scenario (4 of 6 safe; 2 specific seats desperate).

Custom payload helpers (built for this article): - btn_defense_vs_co_open() — BTN as the actor against CO's open (stock Defense.vs_open is BB-only) - custom_defense_vs_open(opener_pos, defender_pos, ...) — generic non-BB defense (for SB-vs-BTN-open and similar non-stock action lines) - custom_defense_vs_limp(opener_pos, defender_pos, ...) — generic defender response when opener limps (calls 1 BB) instead of raising - custom_cbet_with_defender(opener_pos, defender_pos, board, ...) — generic non-BB postflop c-bet (for BvB SB-c-bet and BTN-vs-SB BTN-c-bet)

Fresh solver pulls (2026-04-25 + 2026-04-27 refresh, 50+ queries; all successful): - 3 pairings (BvB, BTN+SB, CO+BTN) × 2 spots per pairing (open + defense vs raise) × 3 regimes = 18 preflop pulls - 3 pairings × 1 spot (defense vs limp) × 3 regimes = 9 limp-defense pulls (added 2026-04-27 for §6 integration) - 2 pairings (BTN+SB, CO+BTN) × 3 regimes = 6 sub-game pulls (defender's open EV when opener folds, for §6 fold-branch terminal value) - 3 pairings × 2 boards (K72r dry, 7s6s5d wet) × 3 regimes = 18 postflop pulls (CO+BTN row re-pulled 2026-04-27 with consistent custom defender) - §6 per-seat per-hand EV metric (added 2026-04-27): for each pairing × regime, integrate the defender's EV across the opener's strategy mix. - Opener EV = reach-weighted avg EV at the opener's open node (already accounts for all downstream play). - Defender EV = P(opener raises) × avg EV at "defense vs raise" + P(opener limps) × avg EV at "defense vs limp" + P(opener folds) × defender's sub-game open-EV. The fold-branch sub-game uses Preflop.open(defender) with the same SquidConfig — captures the value of the defender becoming the new opener after the original opener folds. For BvB, the fold branch is terminal (defender wins +0.5 BB exact, no sub-game); for BTN+SB and CO+BTN it's a real sub-game and the EV is queried fresh. - Convention + known limitations of the absolute values. Each cell is EV as the solver reports it at the seat's first decision. The convention is forward-looking from each seat's first decision (sunk blind costs not subtracted), but the absolute values are not literal "chip flow per hand" — pure chip-flow EVs in HU NLHE would sum to ~0 (zero-sum minus rake), and these numbers don't. The solver's reporting includes the squid-equity term against some implicit baseline (likely "always fold" or a passive default), so summing two seats' EVs gives "joint value of optimal play vs baseline," not "joint expected chips at hand end." Within-pairing comparisons (BB vs SB; SB vs BTN; BTN vs CO) are robust — both seats share the convention, so the bias cancels and the d2t edges (+6.21, +2.78, +0.11) are reliable. Cross-pairing absolute comparisons should be read directionally, not as literal chip-flow magnitudes. A clean chip-flow recomputation that closes the zero-sum identity (with the squid term decomposed against a known baseline) is on the v2 backlog. - Strategy mix probabilities (P(raise), P(limp), P(fold)) are extracted from the opener's open-node response. - Result: each row in the §6 table is "expected forward EV per hand dealt, in BB" — directly comparable across rows (no longer conditional on opener entering the pot).

All numbers cited above are fresh against the live endpoint and stored in facts.yaml. No recycled data from other articles.

Scope limits: val = 3 default throughout. Other val settings discussed at the level of mechanism (the implicit ante scales linearly with val) but not separately verified per spot. 100 BB stacks; deeper or shallower regime breaks are out of scope.

Coined terms: none. All vocabulary is either book-2 v1.8.0 canonical ("squid equity," "safe," "desperate," "val") or standard poker theory ("range advantage," "nut advantage," "equity realization," "bluff-to-value ratio").