ICM: Why Tournament Chips Lie

The dollar value of a chip falls as you stack them

The Independent Chip Model (ICM) converts tournament chip stacks into shares of the prize pool. Its core finding sounds impossible until you see it: each additional chip you win is worth less than the one before. A concrete, computed example: three players remain, prizes 50/30/20 (percent of pool). If stacks are equal, symmetry says each player's equity is 33.3%. Now give one player half the chips (50%) and the others 25% each: the Malmuth-Harville calculation puts the chip leader's equity at just 38.3% of the pool — not 50% — while each short stack still holds 30.8%. The leader owns twice the chips of either rival but nowhere near twice the money. Why? Because no matter how huge your stack, you can only win first place once — the excess chips mostly pad a win you were already favored for.

Bubble pressure, explained

The 'bubble' is the last finishing position that pays nothing — one elimination from the money. Picture four players, three paid, and one player covers you. Calling an all-in here risks your entire *tournament life*: lose and you win $0 while three players cash. The chips you might win are worth less than face value (previous section); the chips you'd lose include the ones keeping you alive, worth more. So near the bubble, calling ranges must tighten dramatically — far beyond what pure card equity suggests. Meanwhile shorter stacks at the table are under even more pressure than you, and everyone knows it.

Why folding AQ on the bubble can be right

Here's the concept that separates tournament players from cash players wearing tournament costumes. On the bubble, a mid-stack open-shoves and covers you. You look down at AQ — a clearly profitable call by chip expected value. And folding can still be correct, because ICM converts those chips to dollars asymmetrically: the dollars you lose by busting (falling from near-certain cash to zero) outweigh the dollars you gain by doubling (which mostly pile up chips of shrinking marginal value). Run spots like this in the ICM calculator and you'll watch 'obviously call' hands become clear folds as the bubble approaches. Chip EV and dollar EV are different currencies, and tournaments pay in dollars.

Big-stack bullying is math, not style

Flip the seat: when *you* have the covering stack on the bubble, everyone else's calling ranges are ICM-crushed — they can't call you without risking tournament death, and they know it. So the big stack should attack relentlessly: open more pots, shove over mid-stack raises, steal every orbit. This isn't a personality choice or 'aggressive style' — it's the same asymmetry from the last section, pointed in your favor. The chips you risk are your cheap surplus ones; the chips they must risk are their expensive last ones. The bubble is the one moment in poker where the rich legitimately get richer for free.

ICM at deal time

ICM's most concrete application: final-table deals. When the last few players agree to split the remaining prize pool, splitting by raw chip percentage (a 'chip chop') systematically overpays the big stack — as our 38.3%-vs-50% example showed. An ICM chop pays each player their Malmuth-Harville equity instead, which is the fair-value baseline. If you ever negotiate a deal, know both numbers before you nod: the difference on a real final table is routinely a full buy-in or more, and the player who computed it usually collects it from the player who didn't.

How the numbers are computed (honesty note)

Every ICM figure in this lesson — and in our calculator — comes from the Malmuth-Harville model: it takes each player's chance of finishing 1st as their share of chips, then estimates 2nd, 3rd, and so on by the same logic applied to the remaining players. It's the standard model used across the industry, but it's a model: it assumes equal skill, ignores position and who's about to pay the blinds, and knows nothing about your opponents. Treat its outputs as a strong baseline, not gospel. Drill: open the ICM calculator, enter a 3-handed spot with a 50/25/25 stack split and 50/30/20 payouts, and confirm the 38.3% number yourself — then move chips around and watch how equity actually flows.