The Ante You Can't See: Mental Models for Squid Value

Daniel Dvoress co-founded GTO Lab. The article you're reading walks one mental model: where Squid Classic's value comes from, how to spot it at the table, and why three of Squid's most-cited "different rules" are actually the same machine viewed from three angles.

§0 — Squid Classic in sixty seconds

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

Squid Classic is 6-max No-Limit Hold'em. Same blinds, same positions, same deck. There is one extra rule. Every time you win a main pot, you collect a "squid" — a win token. Each player can hold at most one. In a 6-handed game there are exactly five squids to win. The game ends the moment five of the six players each hold one. The player who never won a main pot pays everyone else.

The penalty has one knob, called val. The loser pays 5 × val big blinds, split evenly across the five holders. The model was trained at five settings: val = 1, 2, 3, 5, and 10. Most of this article uses val = 3 — the most common training default. At val = 3, the loser pays 15 BB and each holder receives 3 BB.

Two labels run through the rest of the piece: - Safe — you already hold a squid. You can no longer be the loser. - Desperate — you don't yet hold a squid. You're still in the race.

That's the entire format change. Five squids to win. One loser pays. Everything else is just Hold'em.

§1 — A hand that costs more than it shows you

Pick this up at the table.

It's six-handed. UTG and MP fold. CO opens to 2.5 BB. You're on the button with J♠ 8♠.

Cash habit takes about a second. J8s is a marginal hand on the button against an early-position opener — dominated by most of CO's opening range, vulnerable to a squeeze from the blinds, sets up reverse implied odds when it does flop something (you make second-best hands often). The solver agrees: in cash NLHE at 100 BB, J8s on the button vs a 2.5x CO open folds 96.6% of the time. Near-universal fold.

Now ask the same question at a Squid Classic table at val = 3. Same hand. Same opener. Same blinds. Same stacks.

The solver defends 100% of the time. Mostly call. Some 3-bet. Zero fold mass.

Same J8s. Same 2.5 BB to call. Same reverse-implied-odds problem against CO's range. The price didn't change. The board hasn't been dealt. Nothing about the chips has moved.

Where did the EV come from?

I'll spend the next two thousand words answering that question. By the end of §5 you'll be able to spot it at the table without doing the full math.

§2 — The cEV machine you already own

Before I add something to the picture, take a second to look at what you already use.

Standard NLHE EV has one term. Chip EV. The expected number of big blinds your decision wins or loses against villain's range, weighted by the probability of each outcome. Pot odds, equity, immediate profit — that's the whole machine.

Run it on the J8s spot from §1. CO opened to 2.5 BB. The pot is now about 4 BB (CO's 2.5 + SB's 0.5 + BB's 1). You'd put in 2.5 BB to call. Your immediate pot odds are 2.5 / (4 + 2.5) ≈ 38%. But that's the price of the call against the pot, not the price of the hand. To compute chip EV, you'd weight your equity vs CO's opening range across all the post-flop scenarios, factor in implied/reverse-implied odds, and net it out.

You don't have to do that math at the table — the solver does. The answer for J8s on the button vs CO 2.5x in standard NLHE: chip EV is slightly negative. Not by a lot. Just enough that the solver folds the hand the vast majority of the time.

That's the engine you've used on every hand you've played. There is no other term. There is no "future cost" or "next-pot bonus" — every hand is independent and the only thing that survives across hands is your stack.

If standard NLHE cash is the game you know, this is the only EV term you need.

Now hold that thought, because Squid breaks it.

§3 — Naming the missing term: squid equity

Go back to the J8s spot from §1. The cash math hasn't changed. J8s loses 0.X BB on average if you call — chip-EV negative. The solver still calls 100% of the time at val = 3. Something is paying for that loss.

That something is what book-2 calls squid equity.

Here's the formal version. Every Squid decision changes your future-state probability of being the game-end loser. Squid equity is that change, weighted by the penalty.

EV_total = chip_EV + squid_equity

Two cases:

You're desperate (no squid yet). Squid equity is positive when this decision raises your chance of becoming safe — every hand you play has some probability of winning a pot, and any won pot turns you safe forever. The expected gain in squid status, multiplied by the val-scaled penalty you'd otherwise pay, is the squid-equity term.

You're safe (already have one). Squid equity is zero. You can't be the loser. The penalty doesn't apply to you. Your decision tree collapses to chip EV — same as cash.

That's the whole machine.

The reason squid equity is invisible to a standard NLHE eye is that it's not in the pot. The pot has chip EV in it — your call is wagering chips for a chance at chips. Squid equity sits one layer back, in the future-state-of-the-game level. The 2.5 BB you're putting in to call isn't just buying a chance at this pot's chips. It's buying a ticket — a small probability of winning the hand, becoming safe, and dropping the future-penalty term to zero.

The size of that term scales with two things, and only two: 1. val — the penalty multiplier. At val = 1 the squid penalty is 5 BB total, the term is small, and Squid plays close to cash. At val = 10 the penalty is 50 BB, the term is huge, and Squid barely resembles cash. 2. How many opponents are still desperate. If four others are desperate when you are too, the loser-risk is split five ways and your share is small. If one other is desperate when you are too, the loser-risk is binary — one of you pays. As opponents become safe, your share of the remaining-desperation pool concentrates. The term grows.

Now go back to J8s.

In cash, J8s on the button is chip-EV negative against CO's range. That's why cash folds 96.6% of the time. In Squid val = 3, the chip-EV is still negative — but J8s is good enough to win some pots when it does flop something, and every pot won is a squid earned. The squid-equity term is positive enough to flip the total. Solver: 100% defend.

Same hand. Same chips. Different machine.

§4 — Three "differences from cash" that are really one mechanic

Squid Classic looks like it has lots of new rules. Wider preflop ranges. Looser BB defense. Different bluff frequencies. Limping is suddenly correct. Position EV is amplified.

It doesn't. It has one new term. The term — squid equity — explains all the visible rule changes. Three sub-claims show how.

§4a — Difference A: Desperate ranges widen

A desperate player gets a positive squid-equity bonus on every hand they could win. The bonus doesn't have to be large. It just has to be larger than the chip-EV gap on the marginal hand.

Worked example. BTN with 8♠ 7♠ at val = 3, facing CO open. Assume chip-EV of calling is −0.3 BB — a clear cash fold. Now add squid equity. Suppose this hand wins the pot 25% of the time when played (a rough approximation for a marginal suited connector). Each squid is worth val = 3 BB to the player who wins one. Your gain in expected squid value from playing this hand is roughly 0.25 × 3 BB = +0.75 BB. Total: −0.3 + 0.75 = +0.45 BB. The hand flips from clear fold in cash to clear call in Squid.

This is exactly what shows up in the data. At BTN facing CO's 2.5x open at val = 3, BTN's defense rate almost doubles — from 16% in cash to 35% in Squid. (Note: this is defense rate against a CO open specifically — not BTN's opening range from an unopened pot, which is much wider in both formats.) Limp frequency on the call portion goes from 5% to 23%. The hands that flip from fold-to-play in Squid sit in the marginal zone — J8s, A9o, T8s, JTo, 22, 97s. Each one is chip-EV negative in cash and chip-EV-plus-squid-equity positive in Squid.

Book-2 calls this "every preflop range widens." The mechanism is squid equity adding a positive term to marginally-negative cash hands.

§4b — Difference B: Safe and desperate play different games

Safe players have a zero squid-equity term. Their decision tree is pure chip EV. They are, at that point, playing cash.

Desperate players have a positive squid-equity term. Their decision tree is chip EV + squid equity. The same hand, the same board, the same action sequence — completely different decision depending on which state you're in.

This shows up most cleanly in BB defense. Take BB facing CO 2.5x open at val = 3. The BB's overall defending range goes from 39.6% in cash to 93.5% in Squid — a 54-point swing, the largest preflop range shift in 6-max. The mechanism: every defended hand is a chance at winning a squid, and squid equity is large from the BB seat (BB is the only seat that closes preflop action and gets a discount on the call).

But that 93.5% applies only to a desperate BB. A safe BB (who already won a hand earlier in the game) plays the BB seat at the cash 39.6% defense rate. Same seat. Same opener. Different state. Different game.

The strategy isn't a function of position alone anymore. It's a function of position and state. Two new dimensions, not one.

§4c — Difference C: Bluff frequencies drop on wet boards

Cash NLHE c-bet theory anchors on minimum defense frequency. The aggressor's bluff frequency is calibrated so that the defender's break-even fold rate is exactly MDF. If villain folds more than MDF, bluffs print money; if villain folds less, bluffs lose money.

In Squid, desperate defenders overdefend MDF. Every defended pot is a chance at a squid, so the defender's optimal call rate is wider than the cash MDF baseline. This shifts the bluff break-even threshold upward. Bluffs that broke even in cash now lose money.

The signal is texture-dependent. On dry boards — A94 rainbow, K72 rainbow, J72 rainbow — overall c-bet rate actually rises in Squid. The reason is a separate effect on those boards: when villain over-defends, your strong made hands face a wider calling range, so betting becomes more profitable for value (and for "protection" — denying equity to weak hands that would call profitably otherwise). The bluff-bend exists in principle on dry boards too, but it's invisible underneath the value-rise. On wet boards — boards where bluffs and value mix — the bluff-bend wins, and overall c-bet rate drops.

The data: CO c-betting on 7♠ 6♠ 5♦ (a wet mid-connected two-tone) after a single-raised pot vs BB. In cash, CO c-bets 69.5% of the time. In Squid val = 3, CO c-bets 55.5% — a 14-point drop. Average bet size when CO does bet goes up (2.6 BB → 3.7 BB), confirming that the bets which survive concentrate on stronger value hands. The bluff-end of CO's range is checking back more, because the best response to a defender who over-defends MDF is to bluff less.

Three differences. One mechanic. Squid equity is doing all the work — and if you can name it, you can spot it at the table.

§5 — Napkin math: spot the squid equity

You're at the table. You don't have time to compute squid-equity terms in BB. You need a heuristic that runs in three seconds.

Three steps:

1. Am I safe or desperate? If safe, your squid-equity term is zero. Play chip EV. You're in cash. Skip steps 2 and 3.
2. How many opponents are still desperate? This is the same as the count of remaining squids to award. At the start of a fresh game, five. As opponents win pots and become safe, the count drops. The lower the count, the more your share of the loser-risk concentrates.
3. Is chip-EV covering my squid-equity term at this val? If chip-EV is comfortably positive, the call is fine in both regimes. If chip-EV is comfortably negative AND val is 1 or 2 AND most opponents are still desperate, fold. If chip-EV is slightly negative AND val ≥ 3, lean toward play — squid equity is probably enough.

Two practice spots, run the heuristic.

Spot 1. UTG opens with 7♣ 2♦. (Just go with it.) Should the solver play 72o under the gun at val = 3?

Step 1: hero is desperate (assume start of game). Step 2: five opponents are desperate. Step 3: chip-EV of opening 72o UTG is deeply, deeply negative. Squid equity from the worst hand in poker at the tightest position in the game is too small to matter — 72o wins almost no pots, contested or uncontested.

Heuristic predicts: fold.

Solver: fold 100% in cash, fold 100% in Squid val = 3. Heuristic correct. Both regimes agree.

Spot 2. You're in BB with K♠ 5♣ facing CO's 2.5x open at val = 3. Should the solver defend?

Step 1: hero is desperate (assume early game). Step 2: five opponents are still desperate. Step 3: chip-EV of defending K5o is negative — K5o is dominated by everything in CO's range, plays poorly post-flop, gets 3-bet off the hand. In cash, K5o folds 91.6% of the time. But K5o does occasionally flop top pair on a king-high board, win the pot, and become safe. The squid-equity term at val = 3 is large enough to flip the total.

Heuristic predicts: play.

Solver: cash fold 91.6%; Squid val = 3 defends 100% (call 64%, 3-bet 36%). Heuristic correct. Cash and Squid disagree, and squid equity is the gap.

The heuristic isn't a substitute for solver work. It's a screen — it gets you to the right answer faster than running the full decomposition, and it focuses your attention on the parameters that actually move the verdict (your state, opponents' states, val).

§6 — Closer

The squid penalty doesn't show up in the pot. It shows up in the games you don't survive long enough to play. Every hand you fold without a squid is a hand you can't win, and every hand you can't win is one less ticket out of the loser seat. That's the whole machine.

Cash trained you to read one number on every decision: chip EV. Squid trained the solver to read two: chip EV plus squid equity. The first is in the pot, in front of you, paid in chips. The second is in the future state of the game, paid in penalty avoided.

If you've ever sat at a Squid table and felt like the solver's calls were absurd — the BB defending K5o, the BTN limping with 22, CO checking back top pair on a wet flop — what you were watching was the second term doing its work. It looks like loose play. It isn't. It's a different machine, computing on inputs you couldn't see.

Now you can see them. Three steps, at the table, in three seconds. Squid equity is a skill, not a mystery — and most of the "weird" Squid decisions you've second-guessed in your own play probably weren't weird at all.

Methodology

Source: book-2 (Squid Classic) v1.8.0 Part 1 — squid mechanic, val parameterization, "safe" / "desperate" labels, the squid-equity decomposition, and the structural observations (preflop ranges widen, BB overdefends MDF, state-dependent strategy).

Data rail: gameplay-ai range-viewer preview endpoint via strategy_grid_client.py. Defaults: 100 BB stacks, 6-max, 2.5 BB opens. Squid configuration via SquidConfig.fresh(val=3.0) for hero-desperate spots, SquidConfig.hero_last("BB", val=3.0) for postflop c-bet spots.

Fresh solver pulls (Phase A, 2026-04-25, 12/12 queries successful): - §1 hook: BTN J8s vs CO 2.5x open at val=3 — custom payload btn_defense_vs_co_open() (stock Defense.vs_open is BB-only). Cash 96.6% fold → Squid 0% fold. - §4a aggregate: BTN defense rate vs CO 2.5x open: cash 16.0% → Squid val=3 34.9%. Limp share within the calling portion: 5.1% → 22.5%. (This is BTN's response to a CO open — not BTN's unopened-pot opening range.) - §4b aggregate: BB defense vs CO open cash 39.6% → Squid val=3 93.5%. - §4c bluff: CO c-bet on 7♠6♠5♦ wet two-tone after BvBB SRP — cash 69.5% c-bet → Squid val=3 55.5% c-bet (-14 pts). Cross-references on dry K72r, A94r, J72r show RISE (protection-bet pattern) for completeness. - §5 practice: UTG 72o open both regimes 100% fold; BB K5o vs CO open cash 91.6% fold → Squid val=3 100% defend.

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

Scope limits: val=3 default throughout. Other val settings discussed at the level of mechanism (val scales the penalty linearly) but not separately verified per spot. 100 BB stacks; deeper or shallower regime breaks are out of scope for this piece. Per-combo solver granularity is class-level / range-level only; no individual combo selection arguments.

Coined terms: none. All vocabulary ("squid equity," "safe," "desperate," "val") is book-2 v1.8.0 canonical terminology.