Final table of a $200 buy-in. Eight players left. You have 18 big blinds in the small blind. Folds to you, you look down at A-K offsuit. Big blind is the chip leader with 75 BB. The pay jumps from 8th to 1st are $1,200 / $1,800 / $2,800 / $4,200 / $6,000 / $9,000 / $14,000 / $24,000.
Cash-game answer: ship it. A-K is a premium hand, the big blind has to call wide, you're getting your money in good.
ICM answer: it's much closer than you think, and against a chip-leader who's playing well, it might be a fold. Welcome to the part of tournament poker where being right about poker isn't the same as being right about the math.
What ICM actually is
The Independent Chip Model (ICM) is a way of converting tournament chip stacks into expected prize-pool dollar equity. It treats each player as having a probability of finishing in each remaining position, calculates the probability-weighted dollar value of those finishes, and sums them.
You don't have to do the math yourself — every ICM calculator on the web (including ours) does it for you. What matters is what the math means: chip equity and dollar equity are not the same thing, and the gap between them is the entire reason late-tournament strategy is different from cash.
Why $1 of chips ≠ $1 of equity
A simple example. You're at the final table of an $100 buy-in, 9 players left. Pay jumps:
| Place | Prize |
|---|---|
| 1st | $2,500 |
| 2nd | $1,500 |
| 3rd | $1,000 |
| 4th | $700 |
| 5th | $500 |
| 6th | $350 |
| 7th | $250 |
| 8th | $200 |
| 9th | $150 |
Suppose all 9 players have equal stacks. Pure chip equity says each player has 1/9 of the total chips, so 1/9 of the total prize pool ($7,150 / 9 = $794) in expected equity. Wrong. ICM says each player has equal probability of every finish, which sums to a weighted average closer to $794 — but only because stacks are equal here.
Now imagine one player has double the chips of the other eight. Pure chip equity says they have 2/10 of the prize pool ($1,430). ICM says less — maybe $1,200 — because their chip lead can't translate proportionally into dollars (they can't win more than 1st place). The reverse is true for the short stacks: their chip count is small but their dollar floor is the min-cash ($150), so they're worth more in dollars per chip than the leader.
This is the key insight: chips compound less efficiently than they accumulate. Doubling your stack at a final table never doubles your dollar equity. Losing your stack costs you your entire remaining dollar equity, which includes the value of all the lower pay jumps you've already locked in.
When ICM matters
ICM matters most when:
- Pay jumps are large relative to the buy-in. Final tables, satellite bubbles, top-3 finishes.
- Stacks are unequal. Short stacks have inflated $-per-chip value; big stacks have deflated $-per-chip value.
- Few players remain. The math compounds as the number of remaining payouts shrinks.
And it matters less when:
- You're deep in chips with many players left. Early in a tournament, with 200 players remaining and a flat-ish payout curve, ICM is barely distinguishable from chip-EV.
- You're well above average stack with no immediate pay jumps. Building your stack is still good; you just won't be punished as harshly for losing chips.
- You're already short — close to the felt. When you're 5 BB deep, you can't afford ICM-sensitivity; you have to maximize chip-EV or you'll bust on the blinds.
The two big ICM mistakes
Player A gets to the final table and tightens up dramatically. They fold spots that would be auto-shoves in a cash game, hoping to ladder up the pay jumps by surviving. They're playing too tight. ICM rewards survival but doesn't reward passivity — eventually the blinds eat them and they exit in 7th when they could have applied pressure and finished higher.
Player B gets to the final table and plays the same chip-EV strategy they used at 50 BB deep. They 3-bet shove A-Q at 25 BB into a tight big stack. The big stack folds — but they fold because their range includes hands that beat Player B's. When Player B doesn't get folds, they're crushed and exit. They were playing too aggressive.
The correct ICM-aware adjustments:
- Big stacks: Open wider, defend wider, attack short stacks. Their dollar-per-chip is low, so risking chips is cheap.
- Medium stacks: Tighten significantly. Don't get into coinflips with other medium stacks. Wait for spots against short stacks (who have to defend) or shove against the big stack (who has to call tight).
- Short stacks: Stay aggressive. You have the most to gain per chip, and you've lost the least dollar-equity if you bust. Don't fold your way out.
How to start incorporating ICM
You don't need to do ICM math in real time. What you need is to internalize a few specific situations where ICM changes the answer:
1. Final tables with steep pay jumps: Tighten your calling ranges. A coinflip that's clearly +EV in chips might be -EV in dollars.
2. Bubble (about to make the money): Big stacks are bullies; short stacks are desperate; medium stacks must NOT bust. The single most expensive mistake is medium-stack-vs-medium-stack on the bubble.
3. Satellite bubbles: Even more extreme. Satellites pay flat (everyone above the cutoff wins the same seat), so the marginal chip is almost worthless if you're already going to qualify. Fold your way in. This is the only spot in tournament poker where folding A-A is sometimes correct.
For practice, play with our free ICM calculator. Plug in your stack sizes and the prize structure, and run "what if I bust here vs win this hand" comparisons. The numbers stop feeling abstract after about a dozen examples. After that, the right ICM adjustments become second nature in real tournaments.
Track tournament results in PokerCharts with finish position and pay-jump structure, and you can start to see whether your late-tournament play is moving the right direction — laddering up consistently, or bouncing around because of ICM mistakes you didn't notice.