You sit down in a Squid Classic game. The deck is the same. The positions are the same. Blinds are the same. An open-raise to 2.5bb still looks like an open-raise to 2.5bb. And if you play the first hand the way a decade of NLHE muscle memory trained you to, the solver looks at your action log and flags it in red.
Squid Classic is 6-max No-Limit Hold'em with a single rule change: across the game, the player who never wins a main pot pays the penalty. In a 6-handed game, five "squids" get awarded — one to each hand's main-pot winner — and the unlucky sixth player, the one who ends the game with zero, settles up. The penalty is set by a parameter called val, trained at 1, 2, 3, 5, and 10 big blinds. At val = 3, the loser pays 15 BB and each of the five squid-holders receives 3 BB.
The solver doesn't model this as a per-hand tax. It models it as a continuous forward-looking equity term bolted onto chip EV. Every hand is one more chance to win a squid; every fold gives away that chance. That single overlay cascades through preflop, postflop, and river play — and it rewards the opposite of what NLHE has been drilling into you for a decade at a long list of specific spots.
Here are seven of them. Seven reflexes I had to unlearn. If you're coming in from a cash background, you will too.
Reflex 1. "Fold the big blind when you're card-dead"
The NLHE habit. Cash BB defends 51.8% against a CO 2.5bb open. The remaining 48% — the weakest offsuit junk, hands like K4o, J6o, 83o — gets folded. Modern solvers even underdefend MDF against narrow openers by 7–13 percentage points, and most regs run a touch looser than that.
The Squid Classic number. Against the same CO 2.5bb open at val = 3, the solver defends 95.8% of hands. That's a +44.0pp shift. At val = 10 it saturates at 100%. The hands added to BB's defense are overwhelmingly the junk the Cash range folded — 82% of the Cash→val=1 additions are offsuit junk (Part 3 §3.2).
Source: book-2 Part 3 §3.1 (squid-deltas.md Table 2 lines 78–86). See Methodology & caveats.
Why it flips. In Cash, folding costs you zero — the hand ends and the next one starts clean. In Squid, folding forgoes the chance to win this pot and collect a squid, which is a forward-looking equity cost the chip-EV calculation doesn't capture. BB already has the best pot odds at the table because the blind is committed; the extra chip cost to continue a junk hand is small, and the squid-equity overlay pushes the expected value above zero for almost everything.
What to do instead. At val = 3, almost never fold the big blind to a single raise. Not against UTG, not against anyone. The hands that felt like auto-folds in Cash — K4o, J6o, low offsuit gappers — become mandatory calls. Bad equity preflop is fine; you make it back on the structural flop edge in Reflex 5.
Reflex 2. "Limping is a leak"
The NLHE habit. Modern NLHE solvers don't limp. Limping is strictly dominated by raising or folding — raise-first-in has been the default for every preflop spot in open-raise ranges since the discipline became a discipline. Cash limp rates are 0.0% from every position except SB.
The Squid Classic number. At val = 3, BTN limps 30.2% of the time. CO limps 15.5%. SB limps 98.3%. At val = 10, CO limps 75.5% — three out of four entered hands are limps. These are equilibrium frequencies, not mistakes.
Source: book-2 Part 2 §2.2 (squid-deltas.md Publisher-gap lift table lines 315–319). See Methodology & caveats.
Why it flips. Two things changed at once. BB now defends so wide that raising generates much less fold equity — the extra chip commitment doesn't buy what it used to. Meanwhile, the squid-equity reward is identical whether you entered by limp or raise; both give you a shot at the main pot. That makes limping the minimum-cost way to keep squid equity alive on hands that are too weak to raise profitably but too valuable to fold. Marginal hands slot into a new zone the Cash solver doesn't have.
What to do instead. When a BTN or CO limps in Squid, don't assume weakness — it's the middle of a polarized strategy. Counter from BB with a wider 3-bet range against the limped ranges, not by flat-completing and hoping to hit. And from your own seat, learn the limping zone; the hands you used to fold for 2.5bb are the ones you now enter for 1bb.
Reflex 3. "Check back AA on a wet connected flop to pot-control"
The NLHE habit. Cash solver theory on boards like 8h6d4h is unambiguous: betting AA is overvalued, and checking is strictly higher EV. AA bets 0.3% of the time. Protection betting is a classic overbet-your-hand-strength mistake.
The Squid Classic number. At val = 3, AA on 8h6d4h bets 47.4% of the time. At val = 5 it's 83.4%. At val = 10 it's 98.9%. The Cash theory reverses cleanly.
Source: book-2 Part 4 §4.5 (squid-deltas.md lines 589–596). See Methodology & caveats.
Why it flips. Two forces. First, BB's wider defense adds offsuit junk that doesn't connect with 864-type boards — fold equity that didn't exist in Cash shows up in Squid. Second, the cost of giving a free card isn't just chip-EV anymore; if a draw gets there and BB wins the pot, you've lost the squid-equity upside of winning the hand too. In Cash, free-card cost is one-dimensional. In Squid, it's two-dimensional, and the protection math flips.
What to do instead. Bet your overpairs on low and middle connected dry boards (8h6d4h, 7h5h3h, 9s7s5c) at val = 3 and up. Cash says check; Squid says bet, and the bet gets larger — more like 60% pot than a thin blocker — as val climbs.
Reflex 4. "Use MDF as your BB-defense anchor"
The NLHE habit. Minimum Defense Frequency — how often you need to continue so the opener's bluffs break even — is the Cash BB's defense anchor. Against a 2.5bb open, MDF is 50.0%. Cash BB typically underdefends by 7–13pp against narrow openers; theory purists correct upward toward 50%.
The Squid Classic number. At val = 3 against a CO 2.5bb open, BB defends 93.5% — which is +43.5pp above MDF. Against smaller opens (2.0bb) it's +39.2pp above MDF. Against larger ones (3.0bb) it's +41.8pp. The direction is completely reversed.
Source: book-2 Part 3 §3.3 (squid-deltas.md Table 18 lines 693–712). See Methodology & caveats.
Why it flips. MDF is a chip-only formula. It asks how often you must call to make the opener's bluffs break even on chips. In Squid, folding also costs squid equity — the forward-looking component that MDF doesn't see. The real break-even point sits far above the chip-only threshold, and the gap grows linearly with val.
What to do instead. Forget MDF in Squid. If you're using it to cap your defense rate, you're folding 40+ percentage points too much. The reference point isn't MDF anymore; it's 95%-ish at val = 3, rising to saturation at higher val.
Reflex 5. "Slow down on monotone boards — flush draws protect BB"
The NLHE habit. Monotone boards feel dangerous from the preflop raiser's seat. BB defends with suited hands that can flop a flush draw, so the intuition says c-bet frequency should drop. Cash numbers confirm it: CO c-bets K94ss only 32.2% of the time — the lowest c-bet frequency on any dry or A-high texture in the dataset.
The Squid Classic number. At val = 3, CO c-bets K94ss 86.9% of the time — a +54.7pp shift, the largest positive delta in the entire flop dataset. On 652ss the shift is +45.7pp.
Source: book-2 Part 4 §4.3 (squid-deltas.md Table 3 lines 106–107). See Methodology & caveats.
Why it flips. The flush-draw hands in BB's defense range were already defending in Cash — they don't grow from Cash to Squid. What grew is BB's offsuit junk. On a monotone board, 82-87% of BB's Squid-added hands have no spade. They have no pair, no draw, no backdoor — they fold to a c-bet. The "flush draws protect BB" intuition is right for the Cash range. For the range BB actually defends with in Squid, it's wrong in the opposite direction.
What to do instead. C-bet monotone flops aggressively at val = 3 and up. K94ss, 652ss, Q73ss — they behave more like dry rainbows than wet boards. One caveat: this is heads-up-only. In multiway pots additional defenders dilute the effect because someone eventually does have flush equity.
Reflex 6. "Use blocker logic for pocket pairs on A-high boards"
The NLHE habit. On A-high boards, Cash pocket pair bet frequencies are non-monotonic because blocker logic overrides raw hand strength. KK bets 2.1% of the time — it blocks top pair and thus villain's calling range. 99 (the set) bets 98%. 88 bets 15.8%. 77 bets 24.0%. The pattern zigzags across the pair ranks.
The Squid Classic number. At val = 3, every pocket pair from KK through 77 bets between 70% and 100%. KK jumps from 2.1% to 70.4% (+68.3pp). 88 from 15.8% to 96.0% (+80.2pp). The non-monotonic pattern completely flattens.
Source: book-2 Part 4 §4.6 (squid-deltas.md lines 705–716). See Methodology & caveats.
Why it flips. Blocker logic is a tiebreaker between marginal chip-EV lines. In Cash, the right call on KK on A-high comes down to "I block your top pair, so you fold more often than raw equity says you should, but the pot's small and it doesn't matter much either way." In Squid, the decision isn't a tiebreaker anymore. BB's calling range is wide and junk-heavy, and every pocket pair has enough equity against that range to profitably bet. The blocker-driven Cash subtlety disappears under squid-equity pressure.
What to do instead. On A-high boards in Squid, bet your pocket pairs between 70% and 100% of the time regardless of which pair it is. The KK-checks / 99-bets / 88-checks Cash ordering doesn't apply. If you find yourself checking back KK on an A-high flop because "it blocks TP," you're playing Cash theory in a Squid game.
Reflex 7. "Wider flop c-betting means wider turn barreling"
The NLHE habit. Range continuation says: once you c-bet wide, you barrel wide. The c-bet range is the turn range, roughly, with adjustments for board texture and opponent actions. Cash K72r blank-turn barrel frequency is 58.2%.
The Squid Classic number. At val = 3, CO barrels that same K72r blank turn only 49.0% of the time — −9.2pp lower than Cash, even though the flop c-bet range was wider. On an ace turn the drop is steeper: Cash 74.5% → Squid v3 61.1% (−13.4pp).
Source: book-2 Part 5 §5.1 (squid-deltas.md lines 338–342). See Methodology & caveats.
Why it flips. The squid-equity overlay is a forward-looking term that settles at pot entry, not per-street. The extra hands CO c-bets in Squid — marginal bluffs with no turn equity — captured their squid-equity reward when they entered the flop. By the turn, that component is already cashed; the decision reverts to chip-EV, and chip-EV says the weak bluffs give up. The value hands barrel normally. The measured barrel rate averages over a wider population that now includes more give-ups, so the rate drops.
What to do instead. Don't extend Squid's wider flop aggression into the turn. Most of the Squid-added bluffs should surrender on a blank card; keep firing the value hands and the real semi-bluffs. And the flip side: if you checked the flop instead of c-betting, BB's range didn't get filtered, so the delayed c-bet plays much bigger than in Cash (Cash 65.9% → Squid v3 82.7% on the same K72r blank turn).
What doesn't actually change
The hook sells "NLHE muscle memory lies at seven spots." The rest of the game still plays mostly the way you remember. Calibration matters here:
- Position still dominates. Every seat widens in Squid, but the UTG → SB ordering survives intact. Late position is still worth more than early position — it's worth more more.
- The rules of equity don't change. Equities, blockers, combo counts — the underlying game tree is NLHE. The solver is solving NLHE with one new term.
- Late streets revert toward Cash. The squid-equity overlay has mostly done its work by the flop c-bet call. Turn and river decisions in contested pots approach Cash-like chip-EV logic (with specific exceptions — river polarization on dry blank runouts is wilder than Cash, see book-2 Part 5 §5.6).
- Premium hands still want chips in the middle. AA is still the best hand. The slow-play exceptions in Squid are narrow and structural, not the default.
Squid Classic is not a new game. It's NLHE with a new equity term. The seven reflexes above are exactly the places that term materially shifts the optimal line. Everywhere else, the old reads hold.
Where to go deeper
This article is the reader's door. The deep reference lives in book-2 — Squid Classic, co-authored with the QuintAce research team:
- Part 3 — BB Defense — full BB playbook: the 9-category composition table, the MDF reversal in detail, position-dependent Cash caveats, check-raise and check-call trees.
- Part 9 — Actionables — 25-point cheat sheet compressing Parts 2–7 into one reference page.
- Parts 1–8 — the full mechanism spine, scope map, known issues, and multiway findings.
Read book-2 on internals.quintace.ai.
Methodology and caveats
Review-stage verbose methodology per article pipeline convention. Compressed for final publish.
Data source. All numeric claims in this article trace to book-2 (Squid Classic), v1.8.0, and its underlying solver research in engineering-department/gameplay-ai/projects/llm-verifier-game-expansion/squid-classic/ (source versions frozen at build). Each reflex section's inline caption cites the specific book part, section, and source-data line range.
Base conditions. Unless stated otherwise, all findings are:
- Classic mode (one of three Squid variants; the only mode the research covers)
- 6-max, 100bb effective
- Fresh state — all players start with zero squids at the time of the decision
- Heads-up postflop (CO vs BB SRP) for Reflexes 3, 5, 6, 7
- val = 3 as the canonical anchor; val-scaling noted where informative (R3, R4)
- Standard 2.5bb open, 3bb 3-bet where applicable
Scope limits. This article stays inside the intersection of tested positions, boards, and val levels. Per book-2 Part 8, applying findings outside that intersection is extrapolation. In particular:
- MP postflop is thin — the "CO" numbers in Reflexes 3 / 5 / 6 / 7 should not be applied to MP without caution.
- Multiway at val ≠ 3 is thin — all flop c-bet findings are heads-up; multiway dilutes them.
- R5 monotone is explicitly heads-up only.
- Blood Battle and Double mode are declared but not tested — Classic-only scope.
State dependence. Every finding above measures "fresh state" (nobody has a squid yet). Once the squid distribution shifts — you win one, an opponent wins one — ranges move dramatically. A safe CO with three desperate opponents plays 12.9% VPIP; a desperate CO with three safe opponents plays 88.8%. That's a 75.9pp spread in the same position at the same val. Count the squids before every decision.
Reach (§6c compliance). This article cites predominantly aggregate (class-level, by-position, by-val) frequencies from book-2's §6a aggregate-scan data pattern. Per-combo mixed-strategy reach is the reliability lens for hero-specific outputs (§6b pattern) and doesn't apply to class-level aggregates here. The one specific-combo claim — AA on 8h6d4h in R3 — uses a canonical high-weight teaching combo with no out-of-distribution concern; the source book cites the combo explicitly at squid-deltas.md lines 589–596.
Known issues. Three issues in the llm-verifier-game-expansion/squid-classic project register (KI-1, KI-2, KI-4) affect raw EV field reliability in Squid. This article cites policy/frequency outputs only (VPIP, defense %, c-bet %, bet frequency). No EV values are reported. The Squid regression smoothing artifact (KI-2) doesn't propagate to frequency outputs.
Checkpoint drift (R7). Specific turn-barrel magnitudes drift across training runs per Part 5 Research notes (KI-6). The direction — barrels decrease — is stable. The specific −9.2pp / −13.4pp numbers in R7 are the v1.0-certified values; on the current model checkpoint the magnitudes differ somewhat while the pattern holds.
Widget selection (§8b compliance). Cycle 1 scan of docs/poco-widget-inventory.md completed 2026-04-21. Three widgets shipped:
1. Reflex Toggle — custom; no library widget fits "N-way parallel NLHE-vs-Squid binary flips." Ad-hoc build justified and documented.
2. MDF Deviation Gauge — adapted from DeadMoneyScaleV5 / MixFrequencyDial conceptual template.
3. Board-Type Classifier — adapted from WetOrDryQuizV2 + CbetBoardTextureV2 shapes.
Three distinct interaction shapes (binary toggle, slider-guess-then-reveal, tap-to-classify). No widget duplicates a static chart. Widget count = 3, inside the §8b [2, 6] range.
Voice and author attribution. This v1 runs in third-person narration. Uri Peleg is this article's author and the Squid Classic domain authority; his first-person voice enters in the final.md pass after he reviews the deployed page. Per METHODOLOGY §1b, author review runs against the live URL, not a doc draft. The brand register is analytical peer — solver is the tool, Uri is the voice.
Cross-reference to book-2. Every claim above has a cross-reference into a book-2 part and section where the full data table, mechanism causal story, and scope caveats live. Readers who want proof, not summary, should click through.
Draft · v1 · 2026-04-21 · Uri Peleg — supported by solver and QuintAI