ICM Explained: Independent Chip Model for Tournament Players

Every tournament player eventually meets a hand where the maths says call and every instinct says fold. Four players left, a pay jump ahead, and someone shoves into your marginally-ahead hand. Chip EV says take it. Folding still feels right.

The instinct is correct, and ICM is the model that explains why.

Chips are not money

In a cash game, a $100 stack is worth $100. Stand up, rack it, cash it. The chips and the money are the same object.

A tournament chip cannot be cashed. It has no value at all except as a means of surviving to a paid finish. The only thing that pays is where you end up in the ladder, which means the value of a chip depends entirely on how much it changes your chances of finishing higher.

That relationship is not linear, and the reason is structural: you cannot win more than first place. Doubling from 10% of the chips to 20% roughly doubles your chance of winning. Doubling from 50% to 100% takes you from a good chance of first to a certainty of first — a much smaller improvement for the same number of chips.

The consequence, which is the whole of ICM in a sentence:

The chips you can win are worth less than the chips you can lose.

Every ICM adjustment that follows is a restatement of that line.

What the model does

The Independent Chip Model turns a set of stacks and a payout structure into a dollar figure for each player. Its assumption is deliberately simple: your probability of finishing first equals your share of the chips in play. Once someone wins, the model reruns the same logic on the remaining stacks for second place, and so on down the ladder.

That is all it is. No skill, no position, no blind level — just stacks and payouts. The simplicity is the point: it makes the model computable, and it turns out to be accurate enough to reshape strategy correctly even though it ignores a great deal.

The ICM calculator runs this for any set of stacks and payouts. What follows is what it is doing underneath.

A worked example

Three players remain in a tournament with a $1,000 prize pool paying $500, $300 and $200. The stacks:

• Player A — 6,000 chips (60%)
• Player B — 3,000 chips (30%)
• Player C — 1,000 chips (10%)

If chips were money, those stacks would be worth $600, $300 and $100. They are not.

Player A wins outright 60% of the time. To finish second, someone else must win first and A must beat whoever is left — which works out at about 32%. That leaves roughly 8% for third.

That gives an expected prize of 0.60 × $500 plus 0.32 × $300 plus 0.08 × $200, which comes to $412.

Six hundred dollars of chips, worth $412.

Player B wins 30% of the time, finishes second about 48%, and third about 22%, giving roughly $338 — slightly more than the $300 the naive chip count suggests.

Player C, with a tenth of the chips, wins only 10% of the time but is guaranteed at least $200 simply by being present. That floor is worth an enormous amount relative to the stack:

Running the same sum — 0.10 × $500 plus 0.19 × $300 plus 0.71 × $200 — gives $249.

A thousand chips, worth $249. Two and a half times the naive figure.

The three numbers add to exactly $1,000, as they must. And the redistribution is the whole lesson: the chip leader's stack is worth substantially less than face value, and the short stack's is worth substantially more. Value flows down the ladder, because the prize pool is flatter than the chip distribution.

The risk premium

The strategic consequence of all this is the risk premium: the extra equity you need before a call that is break-even in chips becomes break-even in dollars.

Consider a coinflip for your tournament life at a final table. In chips it is exactly neutral — you double or you bust, with equal probability. In dollars it is a clear loss, because the chips you gain are worth less per chip than the chips you lose. You have taken a fair bet in one currency and a losing one in the currency that actually pays.

So the equity you need to call rises. Where chip EV says 50%, ICM might demand 55%, 58%, or more, depending on how steep the pay jumps are and how the other stacks are arranged. That gap is the risk premium, and it is the single most useful number ICM produces.

Two things make it larger:

Steeper pay jumps. The bigger the gap between the position you would fall to and the one you would climb to, the more surviving is worth.

More stacks shorter than yours. If two players are about to bust, folding for a few orbits can be worth more than any hand you could pick up. Every additional short stack behind you increases the value of simply not busting.

And one thing makes it smaller: being the short stack yourself. If you are about to be blinded out, you have little to protect. The player with the most to lose is not the one closest to elimination; it is the middle stack watching a shorter one hang on.

That asymmetry is why the big stack can apply so much pressure. Their risk premium is small, everyone else's is large, and they can raise into a table full of players who all need a premium hand to fight back. Working through that dynamic is the substance of final table strategy under ICM.

Where ICM bites hardest

The money bubble. The most extreme ICM spot in poker. One player from a payout, the difference between busting and surviving is the entire minimum cash, and folding hands that are clearly ahead in chips is routinely correct.

Final table pay jumps. Every elimination moves everyone up. Immediately before a large jump, the short and middle stacks should tighten sharply and the big stack should attack relentlessly.

Satellites. ICM taken to its logical extreme. When every seat pays identically, chips above the amount needed to secure a seat are worth nothing at all. Folding aces preflop can be mathematically correct in a satellite — the only genuinely correct fold of aces in poker.

Where ICM does not apply: early tournament levels with a hundred players left, where the model's effect is negligible and near-chip-EV play is fine. And cash games, ever.

A short worked adjustment

The clearest way to feel the effect is to price the same hand twice.

You are second in chips at a five-handed final table. The player to your left shoves for a stack that would cost you a third of yours to call. You hold pocket nines, which is roughly a coinflip against their shoving range — call it 52%.

In chips, this is a clear call. You are ahead, and chips gained are chips gained.

In dollars, it depends on the two players behind you. If both are very short and about to be blinded out, folding banks two near-certain ladder climbs while calling risks your entire second-place equity on a coinflip. With steep jumps, the equity you would need for that call can rise into the high fifties — above the 52% you actually hold.

Same cards, same opponent, opposite correct answer, decided entirely by two stacks that are not even in the hand. That is the part of ICM that no amount of hand-reading skill substitutes for: the correct decision depends on the table, not just on the pot.

What the model gets wrong

ICM is a useful approximation, not a description of reality. Three assumptions are visibly false:

It ignores skill. A strong player's chips are worth more than the model says, because they will convert them into finishes more often than their stack share implies.

It ignores position and blinds. Being about to post the big blind matters; being on the button matters. The model sees only the number of chips.

It understates the big stack. Finishing probability is not really proportional to stack size, because a big stack can pressure opponents into folding, compounding its advantage in a way the model does not capture.

None of this makes ICM useless. It makes it a floor rather than a ceiling: the adjustments it prescribes are directionally right and usually conservative. Players who ignore ICM entirely lose far more than players who apply it slightly too rigidly.

Using it in practice

You cannot compute ICM at the table. What you can do is build the intuitions it produces, and check them away from the table until they are automatic:

1. Run your real final tables through a calculator. Enter the stacks and payouts you actually faced and see what your chips were worth. A few of these permanently changes how a pay jump feels.
2. Learn the pattern, not the numbers. Short stacks gain relative value, big stacks lose it, and the middle stacks are the ones most constrained.
3. Ask who is under the most pressure. It is rarely you and rarely the shortest stack. It is usually the player just above the short stack, who has the most to lose by getting involved.
4. Widen when nobody can punish you. If you have the big stack on a bubble, your opponents' risk premiums are your profit.

The mechanics of the model itself, without the strategy layer, are covered on the page explaining what ICM means in poker. For the raw hand-versus-range equity you feed into these decisions, the poker equity calculator supplies the other half of the comparison — ICM tells you what equity you need, and that tells you what you have.

If the tournament format itself is newer to you than the maths, the complete beginner's guide to poker covers the ground underneath all of this. ICM is a late refinement; it only pays once the fundamentals are already in place.