Poker Wiki
ICM in Poker: The Independent Chip Model
ICM turns tournament chips into prize money. It shows why the chips you risk are worth more than the chips you win, and how much extra equity a call needs on the bubble and at a final table.
ICM calculator
What every stack is worth in prize money, and the equity a call really needs. Separate numbers with commas.
| Stack | Chips | Prize pool | Value |
|---|---|---|---|
| 4,000 | 40.0% | 29.9% | 298.76 |
| 3,000 | 30.0% | 27.0% | 270.24 |
| 1,500 | 15.0% | 19.6% | 195.61 |
| 1,000 | 10.0% | 15.2% | 151.67 |
| 500 | 5.0% | 8.4% | 83.73 |
- Equity the call needs
- 78.6%
- In chips alone it needs 50%. The extra 28.6 points are the risk premium.
- Stack value after each result
Fold: 270.24
Call and win: 343.74
Call and lose: 0.00
Blinds and antes are left out, so the chip-only number is always 50%; dead money lowers both. Want the answer to become instinct? Try the ICM Trainer
What ICM is
The Independent Chip Model, or ICM, estimates what a tournament stack is worth in prize money. In a cash game a chip is a chip: double your stack and you have doubled your money. In a tournament you cannot cash out your chips, and you get paid by the place you finish, so the link between chips and money bends.
The model works from just two inputs: every stack in play and the prizes still to be paid. Your chance of finishing first is your share of all the chips. Your chance of each lower place follows from the same idea: given who has already placed, you take the next place in proportion to your share of what is left. Multiply each place's chance by its prize, add them up, and you have the value of the stack.
That gives you two ways to measure any decision. Chip EV counts the chips you expect to gain. Dollar EV counts the prize money. In a cash game they are the same thing. Near the money in a tournament they often disagree, and the prize money is what you get to keep.
Chips and prize money on a bubble
Here is a spot we will use throughout. Five players are left, four get paid, and the prizes are $400, $300, $200 and $100. There are 10,000 chips in play. Try it in the calculator above: these are its starting numbers.
| Player | Stack | Share of chips | ICM value | Share of prizes |
|---|---|---|---|---|
| A | 4,000 | 40.0% | $299 | 29.9% |
| B | 3,000 | 30.0% | $270 | 27.0% |
| C | 1,500 | 15.0% | $196 | 19.6% |
| D | 1,000 | 10.0% | $152 | 15.2% |
| E | 500 | 5.0% | $84 | 8.4% |
Player A has 40% of the chips and only 29.9% of the money. Player E has 5% of the chips and 8.4% of the money. Chips move value from big stacks toward short ones, because the big stack cannot win more than first prize, and the short stack has a real chance of slipping into a paid place while others bust.
Why the chips you win are worth less
Start with five equal stacks of 2,000. Each is worth $200. Now one player doubles up through another, and holds 4,000. Doubling their chips took their stack to $300: a gain of $100. The player who lost is out with nothing, and that was a loss of $200.
So that all-in risked $200 to win $100. In chips it was an even bet. In money it was laying two to one. The other three players did nothing, and each went from $200 to $233, because a player busting moved them up the ladder.
This is the core idea behind ICM: the more chips you have, the less each extra chip is worth. It is also why folding has value in a tournament, even though it never does in a cash game. When you fold, you stay in, and other players may bust ahead of you.
Risk premium: what a call really needs
Risk premium is the extra equity you need over a chip-only call. To find it, compare three stack values: folding, calling and winning, and calling and losing. The required equity is the gain from folding over losing, divided by the full gap between winning and losing.
Take Player B facing an all-in from Player A, who covers them. B is worth $270 after a fold. Winning makes B $344. Losing busts B on the bubble, for nothing. So B needs 78.6% equity, where a chip-only call with no dead money needs 50%. The risk premium is 28.6 points.
required equity = (value after fold - value after loss) / (value after win - value after loss)
| Caller | A shoves | B shoves | C shoves | D shoves | E shoves |
|---|---|---|---|---|---|
| A calls | - | 66.5% | 55.7% | 53.5% | 51.7% |
| B calls | 78.6% | - | 57.8% | 54.6% | 52.2% |
| C calls | 73.7% | 72.4% | - | 58.3% | 53.3% |
| D calls | 69.8% | 68.8% | 65.0% | - | 53.5% |
| E calls | 61.0% | 60.4% | 57.7% | 55.2% | - |
Three patterns stand out. Calling a stack that covers you is expensive, because you can bust: for every player except A, the highest number in the row is the one against A. The bigger your stack, the more that hurts: Player B needs 78.6% against A, while the shortest stack, Player E, only needs 61.0% against the same shove, because E has less to protect. And a big stack calling a short one barely pays any premium: A needs just 51.7% to call E.
The numbers are not symmetrical, and that is where pressure comes from. If A shoves on B, B needs 78.6%. If B shoves on A, A needs 66.5%. Same two players and same chips at risk, but A can attack far more freely.
What that means for real hands
Numbers like these are easier to feel against specific hands. Below, you hold the first hand and know your opponent holds the second. That never happens at the table, so treat it as a test of how high the bar is: against a real range your equity is usually lower than a dominating hand gets here.
| Your hand vs theirs | Your equity | A calls E (51.7%) | E calls A (61.0%) | B calls A (78.6%) |
|---|---|---|---|---|
| AA vs KK | 81.3% | Call | Call | Call |
| AKo vs AQo | 74.0% | Call | Call | Fold |
| JJ vs AKo | 57.3% | Call | Fold | Fold |
| QQ vs AKs | 53.8% | Call | Fold | Fold |
| 22 vs AKo | 53.0% | Call | Fold | Fold |
Look at the last column. Player B is folding ace-king against ace-queen, a hand that dominates. Only aces against kings clear the bar. Meanwhile the chip leader can happily get in a small pair against ace-king when the short stack shoves. Same cards, different stacks, opposite decisions. To check other hands, use the equity calculator.
Payouts change everything
Risk premium comes from the jump between prizes, so the payout table matters as much as the stacks. Keep the same five stacks and the same $1,000 pool, and only change how it is split. Here is what Player B needs to call Player A.
| Payout | B's stack value | Equity needed |
|---|---|---|
| Flat: 280, 260, 240, 220 | $247 | 92.1% |
| Standard: 400, 300, 200, 100 | $270 | 78.6% |
| Top-heavy: 700, 200, 70, 30 | $291 | 60.3% |
| Winner takes all: 1,000 | $300 | 50.0% |
The flatter the payouts, the more a bust costs and the less first place adds, so survival is what pays. The more top-heavy they are, the more you need chips to get anywhere near the money that matters. In a winner-takes-all event there is no ladder at all, so every stack is worth exactly its share of the chips, and ICM has nothing to add.
That is why satellites, where every ticket pays the same, are the extreme case for tight play on the bubble, and why a top-heavy final table rewards more aggression than its stacks alone would suggest.
On the bubble
The money bubble is where ICM bites hardest, because busting pays nothing while the next player out gets a prize. Short stacks at the table push this further. Compare two versions of the same call, A shoving on B with the same stacks for those two players. With the other 3,000 chips spread evenly, B needs 78.1%. With two players on just 100 chips each, B needs 82.2%. The players on 100 might bust on their own at any moment, so B has more to gain from waiting.
It does not stop the moment the bubble bursts, either. If Player E goes out and Player C picks up those chips, the stacks are 4,000, 3,000, 2,000, 1,000, and B calling A still needs 72.2%, because every step up the ladder still pays.
- Medium stacks play tightest, especially with shorter stacks still in. They have the most to protect and the most to gain by waiting.
- Big stacks can open and shove far wider into medium stacks, because those stacks have to fold so much.
- Short stacks still have to take their spots. They are not protecting much, and the blinds will take the rest of the stack anyway.
- Skip marginal spots. A play that wins a few chips in chip EV is usually a losing play in prize money.
At a final table
Final tables have the biggest pay jumps, so the same logic applies with more at stake. Here is a nine-handed table with 160,000 chips and a pool of $10,050, paying from $3,000 for first down to $200 for ninth. The chip leader holds 26.3% of the chips and 17.9% of the money. The shortest stack has 2.5% of the chips and 5.2% of the money.
With stacks like these, the second stack needs 71.2% to call the chip leader, and only 52.5% to call the 6,000 stack. The chip leader needs 51.2% to call the shortest stack. So the two biggest stacks should avoid each other, and both should attack the middle of the table.
Risk premium does not only apply to all-ins. It applies to every chip you put in, on every street, for both the player betting and the player calling. A player who can only continue with strong hands will also struggle to fire multiple bets as a bluff. If that player raises into you, you can often defend wider than usual, because they will find it hard to keep barrelling.
Using ICM to make deals
Players at the end of a tournament often split the remaining money, and ICM is a common way to set the numbers. Say three players are left with 5,500, 3,000 and 1,500 chips, with $900, $600 and $300 left to pay. ICM gives them $730, $606, $464.
A plain chip chop would give the chip leader 55.0% of the $1,800 pool: $990, more than first prize. That cannot be fair, since first prize is the most they could ever win. On the other hand, ICM tends to be generous to short stacks, so a chip leader may fairly ask for a little more than the model says.
The limits of ICM
ICM is a useful simplification, and it is important to know what it leaves out.
- It looks only at this hand. Rising blinds, and the future value of a big stack that can push others around, are not in the model.
- It treats every player as equally skilled. A much better player has good reasons to avoid a coin flip that the model calls fair, and a much weaker one may prefer it.
- It ignores position. A very short stack about to pay the blinds is in a different spot from one with a whole orbit to wait, but ICM values them the same.
- It only works close to the money. Far from the paid places, there are too many players and too many prizes, and other factors matter more.
- Using it to apply pressure assumes your opponents feel that pressure. A player who ignores ICM and calls too wide cannot be pushed around, and you lose by trying.
Newer methods try to look ahead, simulating the next few hands with ICM at the end of each. They help most when a table is short-handed and stacks are very short, and they are too slow to use outside of study. For most players, plain ICM builds the right instincts.
Common mistakes
- Treating every tournament spot as an ICM spot
- ICM matters near the money and at final tables. Far from the prizes it has little to say. That does not make every early chip equal: a player with a real skill edge still has reason to pass on an even gamble, because they expect better spots later, and a weaker player has less reason to.
- Folding everything on the bubble
- Tight is only right for the stacks it applies to. A big stack that stops attacking gives away its main advantage, and a short stack that waits will get blinded out.
- Shoving into a player who will not fold
- Pressure only works on someone who feels it. Watch for players calling wider than they should, and stop bluffing them.
- Using chip EV to justify a call
- If a hand is only just profitable in chips, it is often a clear losing call in prize money. Close calls in chips are usually folds.
- Forgetting who still has to act
- These numbers are for two players. A third player left to act adds risk, so the call often needs even more than the table says.
Practise it
You cannot run the numbers at the table, so the goal is instinct. Use the calculator above on your own hands from past tournaments, then get reps in the ICM Trainer, which sets you bubble and final table spots and shows you where the chip answer and the ICM answer part ways. For the ranges these adjustments start from, see the preflop charts. When you are deciding whether a call is priced right in the first place, the pot odds calculator gives you the chip-only number to compare with, and minimum defense frequency shows why those numbers no longer tell the whole story here.
More on ICM in Poker
Concepts and tools that cover this topic.
- GlossaryPoker Terms Starting With I3 poker terms beginning with I, defined with examples and linked to the articles that explain them: ICM Pressure, In Position Advantage, Indifference Principle.
- GlossaryPoker Terms Starting With B6 poker terms beginning with B, defined with examples and linked to the articles that explain them: Backdoor Draw, Blind vs Blind Dynamics, Blockers, Bluff Candidate.
- GlossaryPoker Terms Starting With C6 poker terms beginning with C, defined with examples and linked to the articles that explain them: Calling Too Much (Calling Station), Capped Range, Check-Raise, Cold Call.
- GlossaryPoker Terms Starting With D4 poker terms beginning with D, defined with examples and linked to the articles that explain them: Delayed C-Bet, Donk Bet, Donk Bet (Leading Out), Dry Board.
- GlossaryPoker Terms Starting With F2 poker terms beginning with F, defined with examples and linked to the articles that explain them: Flush Draw, Folding Too Much.
- GlossaryPoker Terms Starting With G4 poker terms beginning with G, defined with examples and linked to the articles that explain them: Geometric Bet-Sizing, Geometric Sizing, GTO Deviation, Gutshot Straight Draw.
- GlossaryPoker Terms Starting With H1 poker term beginning with H, defined with examples and linked to the articles that explain them: High SPR: Nuttiness Required.
- GlossaryPoker Terms Starting With M5 poker terms beginning with M, defined with examples and linked to the articles that explain them: MDF Pressure, Merged Betting Strategy, Merged C-Bet, Minimum Defense Frequency (MDF).
- GlossaryPoker Terms Starting With P5 poker terms beginning with P, defined with examples and linked to the articles that explain them: Paired Board, Polarized Betting Strategy, Polarized C-Bet, Pot Control.
- GlossaryPoker Terms Starting With R2 poker terms beginning with R, defined with examples and linked to the articles that explain them: Raise First In, Range Advantage.
Frequently asked questions
What is ICM in poker?
ICM, the Independent Chip Model, estimates what a tournament stack is worth in prize money. It uses every stack in play and the remaining payouts to work out each player's chance of finishing in each place.
How do you calculate ICM?
Your chance of finishing first is your share of all the chips. Your chance of each lower place depends on who has already placed: you take the next place in proportion to your share of the remaining chips. Multiply each chance by its prize and add them up. It takes too many steps to do by hand, which is why people use a calculator.
What is risk premium in poker?
The extra equity a call needs in a tournament over the chip-only break-even. In our bubble example, a call that needs 50% in chips needs 78.6% in prize money, a risk premium of 28.6 points.
When should I use ICM?
Near the money, on the bubble and at final tables, where prize jumps are close enough to change what a chip is worth. Early in a tournament, it has little to say.
Who is hurt most by ICM on the bubble?
Medium stacks, especially when shorter stacks are still in. They can bust against a bigger stack, and they have plenty to gain by waiting for someone else to go out. Big stacks benefit, because they can attack those medium stacks.
Does ICM apply in winner-takes-all tournaments?
No. With only one prize, a stack's value is exactly its share of the chips, so the risk premium is zero and chip EV and prize money agree.
Why does ICM say to fold a hand that is ahead?
Because winning adds less to your stack's value than losing removes. Being a favourite is not enough: you have to be enough of a favourite to pay for that gap.
Is ICM always right?
No, it is a model. It ignores position, future hands, skill differences and how opponents actually play. It still gives a much better estimate than chips alone, so most players use it to build instincts and adjust from there.
Related
Concepts and tools that build on this page.
- ConceptMinimum Defense Frequency (MDF)Minimum defense frequency is how often you must continue against a bet so a pure bluff cannot show an automatic profit. How to work it out, and why it is a guide rather than a rule.
- Free toolPoker Equity CalculatorExact hand-versus-hand equity: every remaining runout dealt and counted, not simulated.
- Free toolPot Odds CalculatorWork out the exact equity you need to call any bet, and the minimum defense frequency that goes with it.
- ConceptExploitative vs GTO PlayGTO is a strategy that cannot be beaten. Exploitative play gives up that protection to win more from a specific mistake. How much each is worth, what deviating costs, and when to do it.
- ConceptAlpha (Break-Even Bluff Frequency)Alpha: how often a bluff must succeed to break even (= bet / (bet + pot)).
- Free toolPLO Equity CalculatorExact four-card Omaha equity, hand against hand: every remaining runout counted under the two-plus-three rule.
Source: Every value on this page is computed with the Malmuth-Harville model when the site is built, using our own example stacks and prizes. Blinds and antes are left out to keep the arithmetic clear, so these are teaching numbers rather than solver output.