What Is ICM in Poker?

Chips are not money

In a cash game, a $100 stack is worth $100. You can stand up and take it to the cashier. Doubling it makes you exactly $100 richer, and a coinflip for stacks is a break-even proposition.

A tournament chip cannot be cashed. It is only worth whatever it does to your chance of finishing in each paid position, and the payout ladder is not proportional to chips. If four players have equal stacks and the payouts are $500, $300, $200 and nothing, each player’s seat is worth $250 — but the chip leader with double everyone else is not worth $500, because there is no way to win more than $500.

That asymmetry is the whole of ICM. The chips you can win are worth less than the chips you can lose, and the gap grows as the payouts flatten.

How the model works

Malmuth-Harville, the standard formulation, makes one assumption: your probability of finishing next among the remaining players is your share of the remaining chips. Apply that recursively — who comes first, then who comes first among the rest, and so on down the paid places — and you get a full distribution of finishing positions for every player. Multiply by the payouts, and you have each seat’s dollar equity.

The model ignores skill, position, and blind levels entirely. It is a deliberate simplification, and it is still by far the most useful single number available at a final table.

The risk premium

The practical consequence is that a call which breaks even in chips loses money in dollars. At a four-handed final table with even stacks and a 40/30/20/10 payout ladder, a coinflip for your entire stack is a real loss even at exactly 50% equity — you need something closer to 55% before it becomes correct.

That gap is the risk premium, and it explains most of what looks like excessive tightness in good tournament players. They are not scared; they are correctly pricing the chips.

When ICM matters most

• The money bubble. One place from a payout, the difference between busting and folding into the money is enormous relative to the chips involved.
• Final tables with steep ladders. The bigger the gaps between payouts, the more each one is worth protecting.
• Satellites. The extreme case: every qualifying seat pays identically, so chips above the qualifying threshold are worth almost nothing. Correct satellite play looks absurd by cash-game standards, and it is correct.

ICM does not apply to cash games at all. Cash chips are worth their face value because you can leave with them at any time.

What ICM leaves out

The model treats every player as equally skilled and ignores position entirely, which means it undervalues a good player’s stack and overvalues a bad one’s. It also ignores blind levels — a 20 big blind stack and a 20 big blind stack at a much higher level are not the same asset.

These are real limitations, and they mostly push in the same direction: a strong player should be slightly more willing to take marginal spots than raw ICM suggests, because their post-flop edge compounds. But the adjustment is small compared with the effect ICM itself is capturing, which is why the model remains the standard.

Where to go next

Run your own final table through the ICM calculator to see the numbers for the stacks in front of you. The other half of the decision is how often you actually win the hand, which is what the equity calculator is for — ICM tells you the price, equity tells you whether you can pay it.