Tournaments & ICM
Bubble and final table:
understanding ICM
Why winning chips is not enough in tournaments. An interactive example, the effects of pay jumps and a method for studying your decisions with ICM.
Beginner to intermediate
01ICM estimates a share of prizes from stacks. It does not predict your next result.
02The same matchup changes depending on who covers whom and which stacks remain.
03To improve, compare similar situations while changing only one parameter.
01 / The principle
A good decision in chips can cost tournament value
You are approaching a pay jump. A player puts you all-in, you think you have enough equity to call, but another player has a tiny stack. Your pot odds describe the price of calling in chips. On their own, they do not measure the value of risking elimination before that short stack.
Chip EV is a decision’s expected gain in chips. Here, $EV represents its effect on expected tournament winnings, expressed in the prize unit. Maximising the former does not always maximise the latter. ICM, the Independent Chip Model, helps you study this difference.
The model starts with every remaining player’s stack and the payout schedule. It assigns finishing-position probabilities and then a value to each stack. This value is a theoretical average: no amount is paid to you at this point, and it does not measure profit after entry fees.
02 / Try it yourself
Three players, 100 chips, 100 prize units
Take three players A, B and C with 50, 30 and 20 chips. The places pay 50, 30 and 20 units respectively. Everyone is already in the money: we are observing late-tournament pay jumps, rather than a bubble where the next player eliminated receives nothing.
In classical ICM, A has a 50 % chance of finishing first. For second place, the model adds two paths: B wins and A beats C, or C wins and A beats B. This gives 30 % × 50/70 + 20 % × 50/80, or approximately 33.93 %. A’s probability of finishing third completes the total to 100 %.
A’s value therefore becomes 50 × 50 % + 30 × 33.93 % + 20 × 16.07 %, or approximately 38.39 units. A’s 50 % share of the chips does not confer 50 % of the prizes, because each of the other players already has at least third place. Adjust the parameters below to observe this relationship.
03 / The cost of risk
Ten chips won do not always offset ten chips lost
Keep the example’s payouts and look from B’s perspective. B’s stack of 30 is worth 32.75 units. Compare two hypothetical transfers of 10 chips between A and B, without blinds or antes. These states isolate the mechanism; they do not describe a complete hand.
| Situation | Stacks A / B / C | B’s ICM value |
|---|---|---|
| Before the transfer | 50 / 30 / 20 | 32.75 |
| B wins 10 from A | 40 / 40 / 20 | 35.67: approximately +2.92 |
| B loses 10 to A | 60 / 20 / 20 | 29.50: −3.25 |
04 / Read the table
Whom can you eliminate, and who can eliminate you?
On the bubble, the next pay jump can take you from zero to a paid place. At a final table, payouts rise as players are eliminated. In both cases, examine all relevant stacks, not just your opponent’s. A very short stack can substantially change the value of your survival.
A player covers you when they have at least as many chips as you: losing your stack to them can eliminate you. Conversely, losing to a player you cover leaves you chips if your stacks are unequal. That does not make the matchup free: your remaining stack and the pay jumps still matter.
- 1
Identify the pay jump
Record the number of players remaining and the next prizes. Do not automatically apply a hand played far from the money.
- 2
Compare stack coverage
Work out your remaining stack after losing against each opponent. The largest stack and the smallest expose you to different scenarios.
- 3
Distinguish shoving from calling
When shoving, you can win without a showdown if everyone folds. When calling an all-in, that possibility disappears. A pushing range therefore cannot be used as a calling range.
05 / The assumptions
What this ICM value leaves out
Classical ICM values stacks without assigning a skill advantage to a player. At identical stacks, it does not distinguish an expert from a beginner. It also ignores position and approaching blinds in this standalone valuation. A hand analysis does use positions and bets: it is the final estimate of stack value that remains simplified.
Future Game Simulation, or FGS, extends some calculations by simulating future hands and the movement of blinds. It is particularly worth studying when stacks are very short. Its results still depend on the chosen parameters and strategies; a more detailed model does not turn an uncertain assumption into certainty.
In PKO tournaments, an elimination can also pay an immediate bounty and change the value of future bounties. Our example does not account for these. A satellite with identical tickets also requires its actual payout structure. Choose the correct format before any analysis: conclusions from a regular tournament cannot simply be copied across.
06 / Put it into practice
One study session, one question, three comparisons
After your session, choose a preflop decision near the bubble or a pay jump. The aim is not to collect hand charts: it is to understand which change shifts your decision. Here is a protocol to repeat in a suitable calculator.
- 1
Reconstruct the context
Enter stacks, positions, blinds, antes, actions, remaining players and payouts. Check imported data before running a calculation.
- 2
Write your hypothesis
State your proposed action and why. Describe the assumed opposing range. A conclusion based on a very tight range can change if your opponent plays wider.
- 3
Compare chip EV and ICM
Keep all other parameters and range assumptions identical. Record the actions and displayed units; numbers in chips and shares of prizes are not directly comparable.
- 4
Change one variable
Then test another opponent stack or pay jump. Return to the initial case before each variation so you know what explains the difference.
- 5
Keep one lesson
Write down the decisive factor and the conditions in which it applies. Revisit the hand a few days later without looking at your first answer.
07 / Move to a tool
Calculating your hands and training are distinct needs
ICMIZER lets you study preflop decisions using models including ICM, chip EV and FGS. The calculator becomes useful when your question requires precise stacks, payouts and ranges. Start with this page’s example to discover the principle; reconstruct the full context to work on your own hand.
According to the official help centre, Basic gives access to the ICMIZER calculator. Pro adds features including the Replayer and MTT Coach, formerly SNG Coach. The Coach provides push/fold decision drills. The pricing page distinguishes desktop and mobile quotas: check these before selecting a plan.
If you mainly want answers to a few specific questions about your own hands, evaluate the calculator first. If you plan regular quiz sessions and use the replayer, compare Pro against that concrete need. The right subscription is one your routine actually uses.
Your questions
Before your next session.
Does ICM matter only on the bubble?
No. It applies to stack valuation when several finishing places receive different payouts. The bubble and major pay jumps make its effects particularly visible; they are not when those effects first appear.
Should I always play tighter at the final table?
No. Stack coverage, pay jumps and previous actions change decisions. Some calls become less attractive, while pressure on other stacks can create opportunities to raise.
Does this calculator tell me which hands to shove?
No. It only estimates the ICM value of stacks. Calculating a decision also requires blinds, possible actions, opposing ranges and the probabilities of different outcomes.
Can I use this example for a PKO tournament?
Not to value the whole decision. Bounties and their evolution add payouts that this example ignores. You need a format-appropriate model and correct bounty data.
Do I need to buy ICMIZER Pro to understand ICM?
No. The guide and example are enough to approach the mechanism. The calculator then supports personal analyses; Pro is mainly justified if you use its extra features, including MTT Coach and the Replayer.
Keep learning