What ICM is
ICM, the Independent Chip Model, turns a tournament stack into a share of the prize pool. It asks one question: if the tournament stopped now and every place were paid by the schedule, how much of the money would each stack take home on average?
The model behind it, Malmuth-Harville, is simple. Your chance of finishing first is your share of the chips in play. Your chance of finishing second is worked out the same way among whoever is left once the winner is set aside, and so on down the payouts. Each place's prize times your chance of landing there, added up, is your equity.
It knows nothing about the blinds, your position or who plays well. That is the point: it is the baseline every tournament decision is measured against.
Why chips and money come apart
Three players left, the pool paid 50 / 30 / 20. The chip leader holds half the chips but only 38.4% of the money. The short stack holds a fifth of the chips and 28.9% of the money.
The reason is the payout curve. Third place is already locked in: 20% of the pool belongs to whoever busts next, no matter how. Chips only compete for the gaps between places, so the first chip you win is worth more than the last.
Win a flip for all your chips and your stack is twice as big. Your equity is not. Part of what you were playing for was already yours, and the losing half of the flip costs you everything.
| Stack | Chips | Money | vs chips |
|---|---|---|---|
| 5,000 | 50.0% | 38.4% | -11.6 |
| 3,000 | 30.0% | 32.8% | +2.8 |
| 2,000 | 20.0% | 28.9% | +8.9 |
The bubble factor
Another way to say the same thing. The bubble factor is the money you lose per money you win in an all-in against one opponent. In chips it is always 1. Under ICM it is above 1, and it depends on who you face.
Against a bigger stack it is highest: losing ends your tournament. Against a shorter stack it is lowest: losing hurts, but you stay in. That asymmetry is why big stacks shove into medium stacks on the bubble, and why medium stacks fold hands they would never fold in a cash game.
Four left, three paid. Read the table by row: the factor each stack faces against each opponent.
| Stack | against12,000 | against9,000 | against6,000 | against3,000 |
|---|---|---|---|---|
| 12,000 | — | 1.71 | 1.31 | 1.13 |
| 9,000 | 2.69 | — | 1.48 | 1.16 |
| 6,000 | 2.35 | 2.17 | — | 1.18 |
| 3,000 | 1.57 | 1.48 | 1.30 | — |
The bubble and the pay jumps
Pressure peaks wherever the next bust changes the money most: the bubble, where one more bust puts everyone in the money, and the final table, where every place is a jump. Between jumps, when the next few places pay the same, ICM goes quiet and the game looks like chip EV again.
Stack sizes matter more than cards there. The shortest stack is the one everyone is waiting on, so the stacks just above it fold the most: they have the most to lose from a confrontation and the least to gain. In the spot above, the 3,000 stack is already worth 13.3% of the pool. Folding until someone else busts is often worth more than any hand.
Short stacks and push or fold
Under about 15 big blinds the first decision is the whole hand: shove or fold. Shoving ranges stay wide, because the blinds and antes are a large part of the stack and folding bleeds. Calling ranges tighten, because the caller risks their stack against a range they did not choose, and ICM adds the risk premium on top.
The practical rule: the shover gets to be reckless, the caller does not. Facing a short stack's shove, call with hands that beat their range by the risk premium, not by a hair. As the shover, use the risk premium against them: hands you would fold in chips become shoves when the caller has to fold more.
Heads-up, ICM disappears. With two prizes left, every chip is worth the same slice of the difference between first and second, and the game is pure chip EV.
| Stack | against7,000 | against3,000 |
|---|---|---|
| 7,000 | — | 1.00 |
| 3,000 | 1.00 | — |
Where the model stops
ICM freezes the tournament. It ignores the blinds about to hit the short stacks, the position you hold next orbit, and any edge you have over the table. A strong player is right to take slightly worse spots than the model allows when the alternative is watching the blinds eat that edge. A short stack posting the big blind next hand should discount the model's patience.
It also treats every stack as equally likely to bust per chip. In a real field, a player who refuses to fold busts sooner than the model expects, and a player who folds to the money lasts longer.
Use the calculator for the structure of a spot, then adjust. The number tells you how big a mistake you can afford.
Common mistakes
- Calling off as a medium stack because the hand is ahead. Ahead in chips is not ahead in money.
- Treating ICM as a reason to fold everything. The chip leader is supposed to raise more, not less.
- Forgetting the field. On a twenty-table bubble your table is not the tournament: the other stacks count as much as yours.
- Relaxing at the money. The jumps at the final table are bigger than the bubble's.
- Using the risk premium only as a caller. It is a weapon as the shover.