ICM
In short: ICM (Independent Chip Model) converts tournament chips into money: a stack's value is the sum of every prize times the probability of finishing in that place, with probabilities computed by the Malmuth–Harville recursion from stack sizes. Chips have diminishing value — a leader's stack is worth less than its chip share, a short stack more. Who is under the most pressure depends on the payouts and on who covers whom: in short-handed spots with steep payouts the medium stacks facing bigger stacks are the most squeezed, not the shortest stack.
What ICM Is
- Independent Chip Model (ICM) — a mathematical model for evaluating the monetary value of tournament chips.
- Based on the probability of finishing in each place of the tournament.
- Considers the stack sizes of all players and the payout structure.
- Chips have diminishing returns: each extra chip is worth less money than the previous one.
What Chip Value Depends On
- Tournament chips ≠ monetary value.
- The value of a stack depends on:
- your own stack size;
- opponents' stack sizes;
- the payout structure;
- the number of remaining players.
The ICM Formula
The monetary value of a player's stack is the sum of all prizes, each multiplied by the probability of finishing in that place:
ICM_value = Σ (Prize_i × Probability_i)
where:
- Prize_i — prize for place i;
- Probability_i — probability of finishing in place i.
Probability of Each Place (Malmuth–Harville)
The probability of finishing in a specific place is calculated recursively (Malmuth–Harville model), starting from first place.
- Probability of finishing first:
P(1st) = Player_chips / Total_chips
- Probability of finishing second — some other player j takes first, then the player wins the race among the rest:
P(2nd) = Σ over j ≠ player of P(j is 1st) × Player_chips / (Total_chips − Chips_j)
- Every further place — remove the player who took the place and repeat the same step for the next place with the remaining chips.
When to Use ICM
- Bubble play — just before the money.
- Final table.
- Deals — prize pool negotiations.
- Other strategic decisions in tournaments once payouts are close enough to matter.
ICM Pressure: Who Is Squeezed
ICM pressure is the effect of the payout structure on strategy: - Big stacks can pressure medium stacks: the covering stack risks less when it loses an all-in. - Medium stacks covered by bigger stacks are usually the most vulnerable — they can bust and lose their chance to ladder past shorter stacks. - Which stack is "tight" depends on the bubble type: on a big-field money bubble with a flat min-cash the short stacks are under heavy pressure; on a final table bubble or in a short-handed spot with steep payouts the short stack is the least squeezed (see Example 2).
Inputs for the Calculation
- Player stacks (chip counts) - current stacks of all active players; - updated in real time.
- Payout structure - prize size for each place; - percentage distribution of the prize pool; - number of paid places.
- Number of players - current number of active players; - tournament positions.
Example Input
// Player stacks
const chipStacks = [15000, 12000, 8000, 5000, 3000];
// Payout structure in dollars
const payoutStructure = [1000, 600, 400, 250, 150];
// Or, for a different tournament, as fractions of the prize pool (sum = 1.0)
const payoutPercentages = [0.4, 0.24, 0.16, 0.10, 0.10];
Calculation Algorithm
- Initialization - get all players' stacks; - get the payout structure; - determine the number of prize places.
- Recursive calculation - for each player calculate the probability of taking the next place; - recursively compute ICM for the remaining players; - sum the weighted probabilities.
- Optimization - use memoization for repeated calculations; - cache intermediate results.
Example 1: Three Players, Simple Calculation
Situation: - 3 players: 6000, 3000, 1000 chips. - Prizes: $100, $60, $40.
ICM values (Malmuth–Harville, rounded to $0.1): - Player 1 (6000): ~$82.5 - Player 2 (3000): ~$67.7 - Player 3 (1000): ~$49.9
A plain chip share would give $120 / $60 / $20: the leader's stack is worth much less than its chips, and the 1000-chip stack more than twice its chip share — it already has $40 locked in for third place.
Example 2: Four-Handed Bubble, Three Paid
Situation: - 4 players, only 3 are paid. - Stacks: 5000, 4000, 3000, 1000. - Prizes: $300, $200, $100.
Equity needed to call an all-in for the effective stack (no dead money; chipEV needs 50%): - Short stack (1000): about 55–58% against anyone — less than any medium stack needs against a bigger stack. - Medium stack 4000 vs the leader (5000): about 72.5%; 3000 vs 5000: about 70%; 3000 vs 4000: about 69%. - Leader (5000) vs 4000: about 64%; vs 3000: about 58.5%; vs 1000: about 52%.
Strategic conclusions: - The medium stacks (4000 and 3000) are under the most pressure against stacks that cover them: they need 69–72.5% equity to call off. - The short stack (1000) has the least to lose here — the medium stacks can still bust before it — so it can play closer to chipEV. - Medium stacks avoid confrontation with bigger stacks. - The big stack can open and bluff more often, above all against the medium stacks.
This is a short-handed bubble with steep payouts. On a big-field MTT money bubble with a flat min-cash the pressure shifts towards the short stacks.
Application in the Bot
A bot can use ICM for: - automatic calculation of stack value; - strategic recommendations in tournaments; - deal analysis between players; - EV evaluation of different decisions.
API Interface (sketch)
// Main function for the bot
function calculateICMForBot(gameState) {
const { stacks, payouts, playerCount } = gameState;
return {
icmValues: calculateICM(stacks, payouts),
recommendations: generateRecommendations(stacks, payouts),
bubbleFactor: calculateBubblePressure(stacks, payouts)
};
}
Drill it in the trainer
Theory sticks when your hands do the reps: range quizzes, an AI coach and reviews of your own hands — free start on Telegram.
Train in the bot