Nick Petrangelo co-founded GTO Lab. This piece is the framework he hands students who have a Cash NLHE game and want a structural translation into Squid Classic before they sit down.

§0 — What Squid Classic is, and the framework

Squid Classic is 6-max No-Limit Hold'em (NLHE) with one structural rule added: each time you win a hand, you collect a squid. At game end, every player who has not won a single hand pays a penalty in chips. A session runs long enough that most players collect multiple squids; the players at risk are the ones who never show down a winner.

That rule changes the math. In Cash NLHE you optimize for chips now. In Squid you're also playing for the squid attached to the pot: a fold saves chips but gives up the chance at the squid that comes with winning the hand. Folding has two costs at once.

The weight of the second cost is dialed by val (the penalty multiplier). At val=1 the squid is a mild nudge per hand; at val=10 missing the squid at game-end costs a significant fraction of your stack. The scan covers four val values — 1, 3, 5, and 10 — to show how the mechanic ramps. val=3 is the canonical anchor used throughout.

When I think about the cash → Squid translation, the framework I want students to carry is simple to state. Cash NLHE plays one ledger — chip equity. Every fold saves chips; every call invests chips; the action with the highest chip-EV wins the node. Squid plays two ledgers at once. Chip equity is still there. On top of it sits a second term: the chance of winning this hand and collecting the squid that comes with the pot. The action the solver prefers is the one whose chip-EV plus squid-earning term sums highest — not chip-EV alone.

BB preflop defense — Cash vs Squid val 3
Toggle the regime. The cash defenders (broadways, suited connectors, pairs, premium offsuit) are unchanged. The widening is offsuit junk Cash folds.
BB defense · Cash NLHE 50.6% Aggregate defense vs CO 2.5bb open. Concentrated on the standard cash-defending core.
Cash defense
50.6%
Three representative junk hands · Cash fold %
K5o Offsuit-junk representative — folds nearly always. 99.0%
J2s Suited-junk representative — folds at near-saturation. 91.7%
72o The floor — absolute worst combo, near-pure fold. ~100%
BB defense · Squid val 3 95.8% +45.2pp vs Cash. Cash defenders unchanged; the widening is the offsuit and suited junk Cash folds.
Squid v3 defense
95.8%
Same three hands · Squid val 3 fold %
K5o Now defends close to 100% — squid-earning chance flips the fold call. ~0%
J2s Defends at val=3 — the +45pp widening absorbs hands like this. ~0%
72o Still folds 86.1% of the time — the floor that doesn't get pulled in. 86.1%
Numbers: V2 strategy_grid post-fix rail (2026-04-23). Base: CO opens 2.5bb · 6-max · 100bb · aggregate fresh-state defense. Source: ../dan-cbet-wrong-in-squid/solver-output/bb-defense-squid-overlay.md.

That second term is what reshapes the tree. K5o folds 99% of the time in Cash; in Squid val=3 it defends close to 100%. That isn't a loose call. It's the same calculus, with a second term added that reverses which side of zero the fold-vs-call comparison falls on. At val=1 the per-hand EV gap is small — a single hand's defend-vs-fold decision shifts by a few percentage points of EV. But enough marginal hands cross the defend threshold that BB's aggregate range widens by ~30pp from Cash to val=1 (50.6% → 81.4%). Cash instincts aren't off by much on any single hand at val=1; they're off on which hands make the cut at all.

Two ledgers: how val scales the second cost of folding
Cash plays one ledger — chip equity. Squid plays two — chip equity plus the squid-earning chance attached to the pot. Slide through val to see how the second cost grows while the first stays roughly the same.
BB aggregate defense 50.6%
Chip cost
of folding
small
Squid-earning
chance forfeited
0
Cash NLHE — one ledger. Chip-EV is the only term. Folding a hand like K5o is easy: it loses chips to call, so it folds 99% of the time. There is no second ledger to weigh against the chip cost.
BB aggregate defense: facts.yaml canonical (Cash 50.6% · v1 81.4% · v3 95.8% · v5 99.4% · v10 100.0%). Bar lengths are illustrative of relative scaling, not calibrated to a specific hand. Source: ../dan-cbet-wrong-in-squid/solver-output/bb-defense-squid-overlay.md.

Cash NLHE plays one ledger. Squid plays two. That is the whole frame.

§1 — Preflop: ranges widen, especially BB

The first place Squid bites is the BB chair. The val ladder for BB aggregate defense:

Regime BB defense Δ from Cash
Cash 50.6% (baseline)
val=1 81.4% +30.8pp
val=3 95.8% +45.2pp
val=5 99.4% +48.8pp
val=10 100.0% +49.4pp

The +45.2pp jump from Cash to val=3 is the single largest range shift Squid produces. Almost every postflop story in this article rests on top of it. When I teach cash regs the format, my working assumption is that the BB chair changes the most.

The +45.2pp is composed almost entirely of hands Cash folds. The cash defenders — suited broadways, suited connectors, pairs, premium offsuit — are unchanged; they were already defending. The widening is offsuit junk and suited-junk that the chip-EV calculation rejects in Cash. K5o is the canonical case: 99% fold in Cash, ~100% defense at val=3. J2s sits in the same zone — 91.7% fold in Cash, defends at val=3. The floor is 72o, which still folds 86.1% of the time even at val=3. The absolute worst combo doesn't get pulled in; everything between K5o-grade and 72o-grade does.

The pattern I want students to internalize: each junk hand carries a small but nonzero squid-earning chance. Multiplied by val=3, that chance more than offsets the chip-EV loss of the call.

Three preflop changes, in order:

  1. BB defense widens by ~45pp at val=3. The headline; the rest scales off it.
  2. CO opens widen by a few percentage points — less than half the BB shift. The opener's marginal squid-earning chance per added hand is smaller than the defender's, so the open-range expansion is real but mild.
  3. 3-bet and squeeze ranges shift toward polarization. The polarization is in the preflop raising range — but because the 3-bet range itself filters out all the Squid-added junk, postflop play in 3-bet pots still tracks closer to Cash (§5). The hero-last state in §4 is where this is most visible.

If the only number you carry away is BB defends 96% at val=3, you have the dominant Squid effect, and every postflop section here lives on top of it.

§2 — Flop c-bet: three patterns, one framework

Three flop patterns — pick a texture
Each texture shows a single fingerprint stat and a one-line behavior label. For the deep walk-through hand-by-hand, see Dvoress's c-bet article.
Pattern 1 fingerprint Both levers fire Frequency saturates and sizing climbs together — the cash tree's "more often, bigger" instinct still recognizes the board.
C-bet % (Cash → val 3)
67.1%→98.2%
Avg bet (Cash → val 10)
2.1→4.0bb
Frequency saturates, sizing grows. BB's added junk on A94r misses cleanly — K5o is high-card, J2s is high-card with backdoor at best, 72o folds anyway. CO keeps its nut-advantage edge with AA flopped top set. Both levers can fire because the added-junk fold equity is large and CO's value range still extracts from BB's calling half.
Pattern 2 fingerprint +5.6bb sizing climb Frequency never clears 65% across the entire val ladder; sizing does the whole job. 765 / 654 / 876 in single-raised pots.
C-bet % (Cash → val 10)
48.3%→61.4%
Avg bet (Cash → val 10)
2.2→7.8bb
Frequency stalls, sizing polarizes. BB's added combos on 765 aren't junk — 98s flopped a straight, 65s and 76s are two-pair hands, 87s is top pair plus an open-ended straight draw. The non-junk half of BB's range alone is stronger than CO's whole range. CO can't push frequency because the range advantage has flipped; what's left is a narrow polarized betting range that wants bigger bets, not more frequent ones.
Pattern 3 fingerprint +0.1bb sizing pinned Frequency saturates textbook-style; sizing stays in a 2.3–2.4bb band across the val ladder. The outlier in the cross-board sizing scan.
C-bet % (Cash → val 10)
58.1%→93.6%
Avg bet (Cash → val 10)
2.3→2.4bb
Frequency saturates, sizing pinned. CO holds the strong hands here — AA as overpair, KK, QQ — but going bigger only folds BB's medium pairs that would have called the small bet. Going larger earns no extra folds from the junk that was already folding. The +0.1bb Cash→val 10 delta is the smallest in the cross-board scan; every other texture opens materially larger.
Numbers: V2 strategy_grid post-fix rail (2026-04-23). Base: CO vs BB SRP · 6-max · 100bb. Sources: ../dan-cbet-wrong-in-squid/solver-output/texture-a-high-dry.md, texture-mid-connected.md, texture-paired-low.md.

The cash c-bet tree predicts one shape: bigger texture range advantage, more frequent c-bet, sizing broadly stable. Squid takes the same tree and produces three distinct behaviors depending on what BB's widened defense looks like on the specific board. My framework for c-bet decisions in Squid is to ask which of three regimes the texture lives in before reaching for a sizing.

Pattern 1 — frequency saturates, sizing grows. Dry A-high boards, A94r the canonical example. C-bet frequency: Cash 67.1% → val=3 98.2% → val=10 99.7%. Avg bet: Cash 2.1bb → val=10 4.0bb (+1.9bb). Both levers fire together. BB's added junk on A94r misses cleanly — K5o is high-card, J2s is high-card with backdoor at best, 72o folds anyway — while CO keeps its nut-advantage edge with AA flopped top set and AK top pair top kicker. The cash tree's "more often, bigger" instinct still recognizes this board.

Pattern 2 — frequency stalls, sizing polarizes. Mid-connected boards in single-raised pots — 765, 654, 876. C-bet frequency: Cash 48.3% → val=3 55.5% → val=10 61.4% (never clears 65%). Avg bet: Cash 2.2bb → val=10 7.8bb (+5.6bb, the largest sizing climb in the scan). BB's added combos on 765 aren't junk: 98s flopped a straight, 65s and 76s are two-pair hands, 87s is top pair plus open-ended straight draw. BB's connected-hand density on 765 shifts the range advantage toward the defender — enough to stall CO's c-bet frequency. CO can't push frequency because the range advantage has flipped; what's left is a narrow polarized betting range that wants bigger bets, not more frequent ones.

Pattern 3 — frequency saturates, sizing pinned. Paired-low boards, 772r the canonical example. C-bet frequency: Cash 58.1% → val=10 93.6% (+35.5pp). Avg bet: Cash 2.3bb → val=10 2.4bb (+0.1bb). Frequency climbs textbook-style; sizing stays pinned in a 2.3-2.4bb band across the val ladder. CO holds the strong hands (AA as overpair, KK, QQ), but going bigger only folds BB's medium pairs that would have called the small bet. There's no profitable larger sizing to grow into. Frequency scales with fold equity; sizing scales with whether a bigger bet has a value range that pays it off.

Which regime a texture lives in collapses to two questions I give students: does BB's added range fold to a c-bet, and does CO have a top-of-range that bigger bets extract from against a range that pays them off? Yes-yes is Pattern 1. No-yes is Pattern 2. Yes-no is Pattern 3. The cash tree never asks the first question.

For the hand-by-hand walk through each pattern, see Daniel Dvoress's deep-dive article on Squid c-bet textures. The polarization the flop starts continues on later streets — but with fewer middle-ground sizings; the same range-shape pressure that made the flop fork three ways propagates. Per-street data lives in my desperation-polarization article.

§3 — Hero-last vs fresh state: the late-session asymmetry

Cash baseline vs Squid hero-last — same hand, different equilibrium
BB facing a CO open (button-vs-big-blind, single-raised pot). Toggle the state. Numbers are BB-last preflop raise % at val=3.
Cash baseline (anchor) No squid term — one ledger The Cash anchor for the comparison. Fresh-state Squid sits between this and the hero-last numbers on the other tab — the symmetric squid term shifts these upward, but not to the hero-last extreme.
88 Cash baseline raise % from BB-last vs CO open. Mixes downward against position. 0.2%
KQs Cash baseline raise % from BB-last. Sits in the middle of cold-call range. 0.8%
Cash anchors above. Fresh-state Squid sits between the Cash baseline and the hero-last numbers shown on the other tab. Toggle to compare.
Hero-last state · BB vs CO SRP Hero alone has no squid; opponents are safe The squid term loads maximally on Hero. Above a minimum equity threshold, Hero raises at high frequency; below it, Hero folds or limps; the middle that NLHE keeps mixed disappears.
88 Squid val=3 hero-last raise % — BB facing CO open. Above the threshold; raise at high frequency. 74.9%
KQs Squid val=3 hero-last raise % — BB facing CO open. Above the threshold; near-pure raise. 87.6%
The steep transition. The same hand goes from a near-pure fold (Cash 0.2% / 0.8%) to a high-frequency raise (74.9% / 87.6%) once the hero-last asymmetry loads up. For the deep walk through the transition and the postflop texture flip on monotone boards, see Petrangelo's Desperation Polarization article.
Numbers: V2 strategy_grid post-fix rail (2026-04-24). Base: CO opens 2.5bb · BB-last preflop · 6-max · 100bb. Source: ../nick-desperation-polarization/pull_squid_hero_last.out.json.

There are two squid states at the table, and the same hand at the same val plays differently in each one.

In the fresh state, no one has a squid. The squid term applies symmetrically to everyone; equilibrium looks roughly like Cash with an additional but uniformly-priced second term. Wider, but smooth.

In the hero-last state, Hero is the only player still without a squid; everyone else is safe. The squid term loads maximally on Hero's side, while opponents have no forward-looking pressure left to defend marginal hands. I treat hero-last as splitting the equilibrium into a different shape entirely.

In the hero-last state, BB facing a CO open (button-vs-big-blind, single-raised pot): above a minimum equity threshold, Hero raises at high frequency. Below it, Hero folds or limps. The middle that NLHE keeps mixed disappears. Pocket eights at BB-last raises 0.2% of the time in Cash — and 74.9% of the time in Squid val=3 hero-last. KQs goes from 0.8% to 87.6% on the same comparison. The cash equilibrium for those hands maps to neither figure cleanly.

I flag hero-last as the highest-leverage moment in any Squid session. The equilibrium is most distorted from Cash there. The asymmetry is largest, and the EV cost of playing a Cash-shaped strategy in a hero-last spot is the steepest leak the framework predicts. Counting squids before every preflop decision is what students who play the cash tree well miss when they sit down at val=3 — the table-state read is part of the strategy.

For the deep dive on the steep transition, the texture flip on monotone boards, and the per-combo data behind the asymmetry, see my desperation-polarization article. The on-ramp here just sets up the two states so the deep dive's mechanism is legible.

§4 — What still works (NLHE habits that survive Squid)

Four guardrails — tap to flip
The parts of the cash tree that survive Squid's penalty pressure. Each front face names the condition; the back gives the takeaway.
Guardrail 1
3-bet pots
When the preflop tree gets to 3-bet, BB's defending range no longer carries the wide junk piece that reshapes single-raised pots.
Tap to flip →
Takeaway
Cash logic mostly survives. BB's 3-bet range is AA-TT / AK / AQs. No widened junk to reshape postflop ranges — the 3-bet pot equilibrium is much closer to Cash than the SRP equilibrium is.
← Tap to flip back
Guardrail 2
Multiway pots
When more than two players see the flop, the equity-realization math against multiple ranges changes the call calculus.
Tap to flip →
Takeaway
Cash logic mostly survives. Junk that defends BvCO at val=3 doesn't defend in 3-way pots. The realization math punishes it. Multiway looks more like Cash than HU does — tighter, narrower defends.
← Tap to flip back
Guardrail 3
Low val (val < 3)
The penalty multiplier dials the weight of the second equity term. At low val, per-hand EV shifts are small even as aggregate ranges widen.
Tap to flip →
Takeaway
Cash sizing survives; Cash range-construction does not. At val=1 the per-hand EV shifts are small — but the aggregate range can still widen materially (BB defense 50.6% → 81.4%, +30.8pp). Cash sizing logic mostly survives; Cash range-construction does not.
← Tap to flip back
Guardrail 4
Turn / river on blank runouts
By the turn, BB's Squid-added junk has mostly folded; the hands left in the pot are core defenders.
Tap to flip →
Takeaway
Cash logic mostly survives on BB's calling range. Squid-added junk is mostly out by the turn. Polarization on later streets is in CO's betting range, not BB's calling range — BB plays close to its Cash equilibrium.
← Tap to flip back
Four guardrails for cash regs new to Squid. Focus study on button-vs-big-blind single-raised pots at val ≥ 3 — that's where the framework deviates most.

Most of the cash tree still works in most spots. I tell cash regs new to Squid to focus study on button-vs-big-blind single-raised pots (BvBB SRP) at val ≥ 3 and trust the cash tree elsewhere. Four guardrails:

  1. 3-bet pots play closer to Cash. BB's 3-bet range is AA-TT / AK / AQs — no widened junk to reshape postflop ranges.
  2. Multiway pots play closer to Cash. Junk that defends BB-vs-CO at val=3 doesn't defend 3-way: weak hands can't realize their equity in multiway pots, so the junk folds.
  3. Low val (val<3) — the per-hand EV shifts are small, but the aggregate range can still widen materially. Cash sizing logic mostly survives; Cash range-construction does not.
  4. Turn / river on blank runouts revert toward Cash. Squid-added junk is mostly out by the turn; remaining hands are core defenders. Polarization is in CO's range, not BB's calling range.

For the seven NLHE reflexes that break in Squid, see Uri Peleg's companion piece.

§5 — One framework, five applications

The cash → Squid translation is one framework with five street-level applications. Carry the two-ledger model. Classify each texture with the two questions from §2. Count squids before every preflop decision. That's the first orbit.

Methodology and caveats

Data rail. All numeric claims trace to the gameplay-ai strategy_grid range-viewer preview endpoint (https://preview.rlserv.aceguardianrl.com/api/strategy_grid) via strategy_grid_client.py, following the SquidConfig.hero_last action path for Squid regimes and the Postflop.cbet aggregate for Cash. This article ships zero fresh pulls — every number cites an existing anchor in a sister article. BB preflop defense and per-hand fold deltas are sourced from ../dan-cbet-wrong-in-squid/solver-output/bb-defense-squid-overlay.md. Flop c-bet ladders for A94r, 765, and 772r are sourced from Dan's three texture anchors (texture-a-high-dry.md, texture-mid-connected.md, texture-paired-low.md). Hero-last per-hand raise% data is sourced from ../nick-desperation-polarization/pull_squid_hero_last.out.json.

Rail-fix disclosure. The V2 endpoint received a server-side fix between 2026-04-22 and 2026-04-23 that moved Cash c-bet cells materially on several textures; Squid-regime cells were bit-identical pre- and post-fix. All cells cited here are post-fix. Per methodology §4f the fresh endpoint is the numeric source of truth.

Base conditions. 100bb effective, 6-max, CO opens 2.5bb, BB last-to-act. val=3 is the canonical anchor for the "Squid" prose label; the val=1 / 3 / 5 / 10 ladder is the full scan range. The hero-last vs fresh-state distinction is noted explicitly in §3; aggregate numbers in §§1-2 do not split by state.

Granularity. Range-level (BB defense %, CO c-bet %, avg bet) and class-level (premium-class bet %). Per-combo strategies are referenced through Dan's deep-dive when load-bearing; this on-ramp doesn't re-pull combo-level data.

Scope limits. Single-raised pots only, heads-up only, val ≥ 3 for the "Squid" prose claims unless the val=1 column is named explicitly, CO opener / BB defender as the canonical chair pair. Squid Classic only — Squid MTT and Squid Showdown are not in scope. MP postflop data is research-thin (book-2 Part 4 Research notes); the framework here should not be extrapolated to MP without targeted samples. The research behind this piece doesn't fully cover MP postflop in Squid.

Cross-references. Daniel Dvoress's "The C-Bet You Built for NLHE Is Wrong in Squid" walks each of the three §2 patterns at hand-by-hand depth. My "Desperation Polarization" article walks the §3 hero-last steep transition and the texture flip. Uri Peleg's "Seven NLHE Reflexes to Unlearn in Squid" is the third companion piece. All three sister articles are the deep-dive referents this on-ramp points at; this piece's job is to give the reader a structure those deep-dives slot into.

What this article didn't do. No fresh solver pulls were run; the turn/river polarization claim at the close of §2 is qualitative and cross-referenced to the deep-dive articles, not a per-street numeric anchor. CO open-range widening in §1 is described with directional magnitude ("a few percentage points — less than half the BB shift") rather than a precise aggregate number — the existing pulls don't cover that aggregate cleanly, and the on-ramp scope didn't justify a fresh pull. Both items are flagged in coverage_gaps: for future extension.

Nick Petrangelo — supported by QuintAI.