On a tournament bubble with four players left and three paid, a player holding 2,000 chips faces an all-in from the chip leader. A cash-game player would call with anything that wins half the time. Under the Independent Chip Model, the same player needs to win about 70% of the time to make the call worth it. That gap of twenty points is ICM in one hand.
The model explains a fact every tournament player feels but few can put into numbers: chips do not equal money. (For the vocabulary used below, from “bubble” to “chop,” the site’s poker terms glossary has short definitions.) If you double your stack in a tournament, your share of the prize pool does not double. If you lose all your chips, your share drops to nothing.
The asymmetry is the whole story.
Where the model came from
The method is older than modern tournament poker. David Harville, a statistician, published it in 1973 as a way to estimate how often horses finish second and third, given how often each is expected to win. Mason Malmuth, co-founder of Two Plus Two Publishing, applied the same logic to poker tournaments in the 1980s, which is why the method is also known as the Malmuth-Harville model.
The name “independent” comes from its core assumption. The model treats every player as equally skilled and ignores position, blinds and who is likely to bust whom. It looks at nothing but the chip counts and the payout table. That makes it crude, but it also makes it something you can calculate.
How an ICM calculation works
The model rests on one rule: a player’s chance of finishing first equals their share of the chips in play. With 40% of the chips, you win 40% of the time. For second place, the model removes the winner and applies the same rule to the remaining chips, and so on down the payout list. Adding up every possible finishing order, weighted by its probability, gives each player an expected payout in dollars.
Take a small example. Four players remain in a sit-and-go with a $1,000 prize pool, paying $500, $300 and $200. Fourth place gets nothing. The stacks are 4,000, 3,000, 2,000 and 1,000 chips, so 10,000 in total.
Swipe the table
| Player | Chips | Share of chips | ICM equity |
|---|---|---|---|
| A | 4,000 | 40% | about $336 |
| B | 3,000 | 30% | about $295 |
| C | 2,000 | 20% | about $236 |
| D | 1,000 | 10% | about $133 |
Look at the extremes. Player A holds 40% of the chips but only 33.6% of the money. Player D holds 10% of the chips and 13.3% of the money. Big stacks are worth less than their chip count suggests, and short stacks are worth more, because the short stack is still guaranteed a chance to climb the pay ladder every time someone else busts.
Doing this by hand gets heavy fast. With nine players and nine paid places, the number of finishing orders runs into the hundreds of thousands. Software does the work: PokerStove offered an early ICM calculator, and modern tools such as ICMIZER, HoldemResources Calculator and GTO Wizard combine ICM with push-fold and postflop solving.
The bubble hand, worked through
Back to the opening spot. Player A (4,000) shoves, and Player C (2,000) has to decide. Right now C’s equity is about $236.
If C calls and wins, C doubles to 4,000 and A drops to 2,000. C’s equity rises to about $336. That is a gain of roughly $100.
If C calls and loses, C is out in fourth place with nothing. That is a loss of about $236.
So C risks $236 to gain $100. To break even, C must win 236 out of every 336 times, roughly 70%. Ignoring the blinds already in the pot, a chip-EV view would say 50% is enough. The extra twenty points are often called the risk premium.
Player D, sitting out of the hand, is the quiet winner. If C busts, D moves into the money with a guaranteed $200 and an ICM value of about $249, up from $133. Every all-in between two other players on the bubble is good news for the shortest stack, as long as it does not involve them.
Now flip it. When A risks only 1,000 chips calling D’s all-in, A’s equity moves from about $336 to about $384 on a win and down to about $282 on a loss. A needs a little under 53% to call. The chip leader can call far wider than the medium stacks can, and the medium stacks cannot fight back. This is the mechanism behind the big stack bullying a bubble.
What ICM changes in your strategy
The first lesson: tighten your calls, not your shoves. A player who moves all-in can win the pot without a showdown, and fold equity is worth even more under ICM because nobody wants to bust. A player who calls has no fold equity at all. On a bubble, ranges for calling all-ins shrink much more than ranges for shoving.
The second lesson is to watch who covers whom. A stack that can knock you out carries real ICM pressure. A stack you cover can only cost you chips. Most of the aggression on a bubble should go toward players who cannot eliminate you and who have their own reasons to fold. The relationship between stack sizes and pressure is covered from the early-level side in the site’s piece on big versus short stack strategy.
Third, pay attention to the pay jumps. A bubble with a min-cash of a few hundred dollars in a $10,000 event carries less pressure for a deep stack than a final table where each place adds tens of thousands. If you play satellites, the effect is extreme: when every remaining seat wins the same package, a player with enough chips to lock up a seat can be correct to fold pocket aces.
Where the model breaks down
ICM ignores skill. A strong player who expects to outplay a final table has good reason to take less risk early, since their edge is worth more than the model allows. A weak player facing better opponents may prefer more variance.
It also ignores the future. The model freezes the tournament at one moment and says nothing about the blinds coming around or the fact that a short stack in the big blind next hand is about to be forced in. Tools that try to fix this run “future game simulations,” modelling a few hands ahead before applying ICM at the end.
And it struggles with big fields. Early in a large multi-table tournament, such as the WSOP Main Event with thousands of entrants, the money is far away and the model’s effect on a single decision is small. The pressure builds through the last levels before the money, when hand-for-hand play begins and tables slow down to a crawl.
At final tables, ICM also leaves the table and enters the negotiation. When players agree to split the prize pool, an “ICM chop” gives each player their modelled equity from the current stacks. The chip leader often argues for more than the model gives, pointing to the very skill and position effects ICM leaves out. The shorter stacks answer that the model already rewards them for nothing more than survival, and the haggling over a few hundred dollars can take longer than the hand that set up the deal.