KEY TAKEAWAYS
ICM makes tournament chips worth less than their face value when pay jumps are close, so calls and all-ins that look profitable in chips can lose money in prize-pool terms. The Independent Chip Model (ICM) estimates what your stack is worth in prize money, based on every stack at the table and the payouts. It explains why strong players fold hands near the bubble that they would call with at any other time.
This guide assumes a multi-table no-limit hold’em tournament with a standard payout structure, and it keeps the math light. It builds on tournament strategy by stack size by adding the one factor stack depth alone does not capture: what the chips are worth.
Below you will find what ICM is, why chips lose value, a worked bubble example with every number shown, how to adjust at the table, the mistakes that cost the most, and where ICM matters least.
What ICM Is and Why It Matters
ICM is a mathematical model that approximates a player’s overall equity in an unfinished tournament. In the model, current stack sizes determine each player’s chance of every finishing place, and the expected prize money follows from those chances and the payouts. It assumes all players have comparable skill.
The key idea is that tournament chips are not cash. In a cash game, a chip is worth exactly its face value. In a tournament, a chip’s value depends on the prize structure, because doubling your stack does not double your prize money, while losing your stack can take your prize away completely. That gap between chips and prize equity is what the model measures.
Why it matters for decisions: a call is correct when the equity it wins exceeds what it risks, and ICM changes both sides of that comparison. Players who ignore it make calls that are fine in chips and costly in prize money, especially when several players are close to the same pay jump.
Why Chips Lose Value Near the Money
Consider the shape of the payouts. Each place jump pays a fixed amount, regardless of how many chips you hold when you cross it. So the first chips you win beyond survival add less than the last chips you lose before elimination take away. This is called diminishing marginal value, and it makes every all-in a bet with a smaller upside than downside.
The pressure is strongest when many players are near a pay jump. The bubble is the clearest case: the player who finishes one place short earns nothing, so survival itself has a price. Final tables create the same effect, because each elimination moves everyone up a pay step. Short stacks feel this less, since they have little to protect, and large stacks can use the pressure against medium stacks.
The table below shows how stack position changes what a decision costs. The size of the gap, not the label of the position, sets how tight you need to be, and the example after the table puts exact numbers on it.
| Situation | ICM pressure | Typical effect |
|---|---|---|
| Early in the tournament, far from the money | Low | Chips are worth close to their face value, so chip math is a good guide |
| Near the bubble, medium stack | High | Calls and shoves tighten sharply because busting costs the most |
| Near the bubble, chip leader | Lower | Can apply pressure, since busting is rarely a risk |
| Near the bubble, very short stack | Mixed | Little to protect, but still needs to survive to reach a pay jump |
| Heads-up for the title | Low | Only two places remain, so decisions return closer to chip math |
These are tendencies, not rules. Payout structure and the exact stacks decide the details, so the same hand can be a call at one table and a fold at another.
Example Hand: A Bubble Call in Chips vs. in Prize Money
Suppose a multi-table tournament with four players left and three paid. The prize pool is $100, paying $50, $30 and $20, and fourth place gets nothing. Stacks are 4,000, 3,000, 2,000 and 1,000 chips, and blinds and antes are ignored to keep the math clean. You hold the 3,000 stack, and the 2,000 stack shoves all-in. Chip math says you need 50% equity, because you call 2,000 to win a 4,000 pot: 2,000 ÷ (2,000 + 2,000) = 50%.
Now compute prize equity with ICM. Folding keeps your 3,000 chips, worth about $29.49. Calling and winning gives you 5,000 chips, eliminates the shover, and is worth about $38.89. Calling and losing leaves you 1,000 chips, worth about $15.56. The break-even equity is (29.49 − 15.56) ÷ (38.89 − 15.56) = 59.7%, far above 50%.
The gap is easy to see: winning adds about $9.40, while losing costs about $13.93. That ratio, about 1.5, is often called the bubble factor. Suppose the shover’s range gives you about 55% equity with a hand such as A♦ 9♣. In chips that is a clear call, and under ICM it is a fold, because 55% is below 59.7%.
The numbers come from the standard ICM calculation applied to this simplified spot. Real bubbles include blinds, antes, further players acting behind and skill differences, so treat the result as a demonstration of the direction and rough size of the effect, not as a fixed rule.
How to Adjust Your Decisions at the Table
Start by locating the nearest pay jump and counting how many players are close to it. The closer the jump and the more stacks that can bust, the more ICM pressure you face. Then compare your stack with the others: pressure falls on the stacks that could be eliminated, not on the ones that cover everyone.
Then adjust the two types of decision. Calls tighten more than shoves, because a call risks your tournament life without the fold equity a shove has. When you are the one shoving, you sometimes win the blinds and antes without a fight, so a wider range can be correct, especially from a large stack against medium stacks that fold too much.
Finally, use the model as a guide and adjust for the opponent. If an opponent shoves too wide on the bubble, a call becomes more attractive. If an opponent only shoves with strong hands, the call gets worse. To recognize the spot in real play, ask who covers whom, how far the next pay jump is, and how wide the shove is likely to be.
Common ICM Mistakes
The first mistake is calling off by chip math alone near the money, like the 55% call above. The call is profitable in chips and unprofitable in prize equity, and the difference grows with the bubble factor.
The second is the opposite: folding everything because ICM exists. Folding has a cost, since blinds and antes keep taking chips, and a stack that waits too long loses its fold equity. A third is ignoring who covers whom. The chip leader and a stack that just doubled face very different pressure than a medium stack with the same cards.
A fourth is treating ICM numbers as exact. The model ignores skill and positions, and its output depends on every stack being known. Use ICM as a framework for the pressure, and remember that a correct fold can still look bad when the shover turns over a weak hand. Judge the decision, not the single result.
Where ICM Matters Most in Real Play
ICM matters most in the late stages of multi-table tournaments with steep pay jumps, such as the bubble and the final table. Satellites are the extreme case, where all remaining players win the same seat and chips beyond survival have almost no extra value, which makes folding far more important.
It matters least early on, in deep-stacked play far from the money, and in cash games, where chips equal money and the model does not apply. Payout structures vary by event and by site, so check the structure before judging how steep the pressure is.
Tournaments carry real financial risk and high variance. A correct decision can still bust you, and no model removes that. Play buy-ins that fit your bankroll and avoid moving up in stakes to recover from a bad run.
Related Concepts and the Next Step
ICM sits on top of stack depth, so start with what to do at each stack depth. Stack size sets the range of options, and ICM sets how aggressive or cautious each option should be.
The idea of choosing ranges that cannot be exploited is covered in what GTO poker means in practice, and the pressure of facing a big bet is shown in facing a river overbet with top pair. For the shoving side, when to 3-bet or call preflop explains how ranges shift. Combine all of these before trusting any single number.
