ICM in Poker Explained: Chip Model Math (2026)

What Is ICM in Poker? The Independent Chip Model in 2026

Nine players left in a $109 online MTT, eight get paid. You're sitting on the bubble with A♠K♦, a genuinely strong hand, and the big stack to your left jams all-in. You cover him by a hair. Snap-call, right?

In chips, yes. In real money, calling here can burn more equity than the pot is worth. That gap between "chips say call" and "dollars say fold" is the whole reason the Independent Chip Model exists, and in 2026 it is still the single most important piece of math in tournament poker. This guide covers what ICM is, how it is actually calculated (with a full worked example whose numbers add up), and when it should change your play. When you want to feel it rather than read it, you can practice ICM spots in our ICM trainer.

TL;DR: ICM in 30 Seconds

Chips are not money. ICM is the math that turns your chip stack into its real cash value, by adding up your chance of each finishing place times what that place pays. That value is not proportional to your chip count: the chip leader is worth less than their chip share, short stacks are worth more than theirs, and near pay jumps this flips marginal all-ins from correct to losing.

Key numbers to remember

The one habit that separates tournament winners from cash players who wandered in: near the money and at the final table, stop counting chips and start counting dollars.

If you'd rather see the concept before you read the math, this short explainer covers the same ground.

Why Chips Aren't Money in Tournaments

In a cash game, a chip is a dollar. Win 40 big blinds, stand up, cash 40 big blinds. Tournaments break that rule the moment you register, because the prize pool is fixed and only the top finishers touch it.

Chip EV vs real-money equity

Chip EV is the value you get if you pretend every chip is worth the same fixed amount. It's the right model for cash and the wrong model for the back half of a tournament. Real-money equity, which is what ICM computes, accounts for the payout ladder. The two agree early, when stacks are deep and flat, and they diverge hard once real money is on the line.

The core reason is worth sitting with: the money you can win by doubling up is capped by the payout structure, but the money you lose by busting is your entire tournament life. Reward is limited, risk is total. That asymmetry is exactly what ICM measures.

This is also why every chip you add to your stack is worth slightly less in dollars than the one before it. Your first 10,000 chips buy you survival and a shot at min-cashing. Your fiftieth 10,000 chips mostly buy a bigger number that you can only convert to cash by actually finishing high. Economists call this diminishing marginal value, and in poker it is brutal near the end.

The table below takes a real three-player spot (worked out in full further down) and shows the dollar value of 1,000 chips at each stack size. Watch the value per chip fall as the stack grows.

A short stack's chips are precious. A monster stack's chips are cheap. That is not a metaphor, it is arithmetic, and it drives every ICM decision that follows.

How ICM Is Calculated: A Worked Example

ICM is one clean idea buried under some heavy bookkeeping. The idea:

$\text{ICM equity} = \sum_i P(\text{finish } i) \times \text{payout}_i$

For every possible finishing position, multiply your chance of landing there by what it pays, then add it all up. The only hard part is computing the probabilities.

From stacks to finish probabilities

The starting point is simple:

$P(\text{1st}) = \frac{\text{your stack}}{\text{total chips in play}}$

Hold 50,000 of the 100,000 chips in play and you take first 50% of the time. That single assumption, chips equal win probability, is the engine of the whole model.

Second place is where people give up and reach for a calculator. The logic: to finish second, someone else has to win first, and then you have to be the "winner" of everyone remaining. Walk through each possible first-place finisher, weight by how often they win, then compute your share of the remaining chips.

Following one clean branch

Three players: A has 50,000, B has 30,000, C has 20,000. What is A's chance of finishing second?

Someone other than A wins first. Two cases:

• B wins first (30% of the time). Now only A and C remain, with 50,000 and 20,000. A wins that pair 50,000 / 70,000 = 71.4% of the time. Contribution: 0.30 × 0.714 = 0.214.
• C wins first (20% of the time). Now A and B remain, with 50,000 and 30,000. A takes it 50,000 / 80,000 = 62.5% of the time. Contribution: 0.20 × 0.625 = 0.125.

Add them: A finishes second 0.214 + 0.125 = 34.0% of the time. Third is whatever remains, so A busts first 1 − 0.50 − 0.34 = 16.0% of the time. That is the entire model. Do it for every player and every place and you have priced the table. With five, six, or nine players the branches explode into thousands of paths, which is exactly why a solver earns its keep.

Chip EV vs ICM equity

Now put it in dollars. Three players remain, payouts are $500 for first, $300 for second, $200 for third (a $1,000 pool). Apply the finish probabilities above:

Both columns total exactly $1,000, which is the sanity check every honest ICM example must pass: dollars in equal dollars out. Nothing is invented.

Why the chip leader is not worth their chip share

Player A holds half the chips but only 38% of the money. They gave up $116 of chip-share value simply because they cannot finish better than first, and first is capped at $500. Player C, the short stack, holds 20% of the chips but 29% of the money, because even the last-place payout of $200 is guaranteed value that props up a small stack. This single table is the reason big stacks should apply pressure and short stacks should tighten up, which we get to below.

Risk Premium and Bubble Factor

The worked example prices a frozen table. Real decisions are about risking that equity. Two numbers describe the risk: risk premium and bubble factor.

The same all-in, two different answers

Four players left in a single-table event, three get paid: $1,000 / $600 / $400 / nothing for fourth. You and the villain both have 40,000 chips; two short stacks sit on 10,000 each. The villain open-jams, you cover exactly, and you have to call for your whole stack. Say your hand is a 55% favorite to win.

In chips, that is a trivial call. You win 40,000 chips 55% of the time and lose 40,000 chips 45% of the time:

0.55 × (+40,000) + 0.45 × (−40,000) = +4,000 chips. Positive. Chip EV says call, and chip-EV break-even is just 50%.

Now price it in dollars with ICM:

Folding is worth $688.89. Calling as a 55% favorite is worth only 0.55 × $915.56 = $503.56. Folding beats calling by about $185, even though you are a favorite and even though the call prints chips. To make calling break even under ICM you would need roughly $688.89 / $915.56 = 75% equity, not 50%. That 25-point gap between chip break-even and ICM break-even is the risk premium, and it is enormous here because busting on the bubble costs you everything while doubling barely moves your locked-in equity. Want to internalize the feeling? Run this exact spot in the trainer and vary your equity until the call turns green.

Bubble factor by tournament stage

Bubble factor packages that same idea into one multiplier: how much a lost chip costs you versus how much a won chip gains you. In our spot above, folding risks $688.89 of equity to gain only $226.67, a bubble factor near 3.0. The table below gives working ranges by stage. These are heuristic bands consistent with published ICM study from GTO Wizard and others, not hard laws, so treat them as a compass rather than a GPS.

The closer you are to a big pay jump, the higher the bar, and the more equity you need before risking your stack. That is the whole of ICM strategy compressed into one chart.

How ICM Changes Your Strategy

ICM does not change what a good hand is. It changes how much equity you need before you are willing to bet your tournament life on it. The adjustment depends entirely on your stack.

Adjusting to your stack size

Short stacks live and die on fold equity. Your job is to be the one shoving, not calling, because when you jam you can win the pot uncontested, and when you call you can only realize your raw equity. Push wide, call tight. A medium stack has the worst of it near a bubble: enough to lose by busting, not enough to bully. Medium stacks should avoid marginal spots against anyone who covers them and pick on the true short stacks instead. If your bankroll is not built to ride the swings these spots create, sort that out first with a look at poker bankroll math and our bankroll management guide for poker.

The big stack plays the opposite game. The chip leader from our worked table gave up $116 of chip value, and pressure is how they earn it back. When you cover the table, every all-in you make puts the other players' tournament lives at risk, not yours. Open more, three-bet jam more, and attack the medium stacks who cannot call without risking a pay jump. You are effectively taxing everyone else's fear.

Stack geometry: watch the other stacks

The most missed ICM skill is looking past your own stack. A call that is fine when the other stacks are even can be a disaster when a shorter stack is about to bust. If two players are on fumes, folding almost anything is correct because someone else may bust for you and lock in a pay jump for free. Who covers whom matters as much as your own cards. Modeling variance across a full tournament makes this geometry click faster than any single hand can.

When ICM Matters Most

ICM is not always loud. Sometimes it whispers, and knowing when to listen saves your stack. The two spots where it dominates are the money bubble and the final table.

On the money bubble, the jump from nothing to a min-cash is the largest percentage jump in the whole tournament, so bubble factors spike and calls tighten to almost nothing. At the final table, every pay jump is large in absolute dollars, so the same logic runs throughout. If you only apply ICM in two places, apply it at these two.

Final-table deals and chops

When a final table gets short, players often stop and split the prize pool. ICM is the fair baseline for that split, because it prices each stack in dollars rather than raw chips. A naive chip-chop just divides the pool by chip percentage, and that can shortchange short stacks below their guaranteed money.

Three players at a final table with $10,000 / $6,000 / $4,000 left to play for, holding 60,000 / 30,000 / 10,000 chips:

The straight chip-chop pays the short stack only $2,000, which is below the $4,000 that third place already guarantees. No rational player accepts that. ICM correctly pays player C almost $5,000, because their locked-up minimum is worth real money. If you are ever offered a deal, price it against ICM before you agree. If you play staked, run the split through the staking and markup calculator so your backer's share is right.

Satellites, where ICM is extreme

Satellites are ICM turned up to maximum. Every seat pays exactly the same, so the moment you have enough chips to lock a seat, one more chip is worth almost nothing and busting costs you everything. Correct satellite play looks insane to a cash player: folding aces preflop can be right when you are already in, which inverts the usual preflop range charts by position. That is not a myth, it is ICM at its logical extreme.

When ICM barely moves

Deep in a tournament with 200 players left and everyone holding similar stacks, ICM and chip EV are nearly identical. Pay jumps are tiny and far away, so a chip is close to a chip. Play a near chip-EV game here and save the tight ICM folds for when the ladder gets steep. Applying bubble-factor tightness 300 players from the money just bleeds chips you will need later.

Common ICM Mistakes

Almost nobody covers these well, and they cost real money. Here are the three that show up most often, on the felt and in staked players' hand histories.

Calling off too wide near a pay jump

The number-one leak. Players see a 55% or 60% favorite and call on instinct, exactly like our worked spot where 55% was a clear fold. Under a bubble factor of 2 or 3, being ahead is not enough. You need to be a big favorite, often 70% or more, before a stack-off is correct. When in doubt near the money, fold and let someone else bust.

Ignoring who covers whom

The second leak is tunnel vision on your own hand. Whether the player jamming into you covers you, or you cover them, can completely flip the decision, because only the player at risk of busting pays the ICM tax. Always ask who dies if this goes wrong before you commit chips.

Treating ICM as gospel

The opposite error is trusting ICM blindly. It assumes everyone plays equally well and ignores your future edge, position, and blind pressure. A strong player with a clear skill edge should deviate to keep chips in play for spots where that edge compounds. The equity you leave on the table by folding a thin edge can be worth it if you expect to out-play the field later.

FGS and DCM: models that go further

ICM is a snapshot. It ignores that blinds keep rising and that play continues. Future Game Simulation (FGS) and the Dependent Chip Model (DCM) try to fix this by looking a few hands ahead. They are more accurate near the money and more expensive to compute. For 99% of decisions, plain ICM plus judgment is enough, and it is what every solver you will actually use is built on. One more caveat: in PKO or bounty formats the math shifts, because part of every stack is a cash bounty you collect the instant you knock someone out, so a covered player is worth more than pure ICM says.

Practice, Tools, and the History of ICM

Reading about ICM builds intuition slowly. Drilling real spots builds it fast, because the numbers stop being abstract the moment one costs you a buy-in.

Train real ICM spots

The fastest way to internalize any of this is to run spots until the right fold feels obvious. Train final-table ICM decisions with real stacks and payouts, then check your calling ranges against hand equity so you know exactly how often you are ahead. Pair that with an honest look at your risk of ruin and how many buy-ins you actually need, because ICM discipline only pays off if your bankroll survives long enough to compound it. And if you cash a final table, remember the taxman: our poker tax calculator handles the part nobody wants to think about.

ICM software compared (2026)

You do not have to compute this by hand. Several tools do it for you, each with a different sweet spot.

If you are learning, start free and drill spots. If you are a grinding professional, a paid solver pays for itself in one avoided bubble punt.

Where ICM came from

The math predates poker. Statistician David Harville published the finish-order probability formula in 1973 for horse racing, ranking runners by ability rather than chips. Mason Malmuth adapted the idea to poker tournaments in 1987, and the poker community named it the Independent Chip Model. The word "independent" is a warning label: the model assumes each chip finishes independently of skill, position, or who is sitting where. It is not perfect, and it was never meant to be. It is the best simple answer to a genuinely hard question, which is why it has shaped tournament strategy for nearly forty years. For the formal derivation, the Wikipedia entry on the Independent Chip Model has the Harville formulas; PokerNews and GTO Wizard's ICM Basics cover the strategy side, and Harville's original 1973 paper in the Journal of the American Statistical Association is where the whole thing started.

The chip in front of you is not a dollar. Once you can price it the way ICM does, the folds that used to feel weak start to feel like free money, and the calls that felt brave start to look like the leaks they always were. If chip-dumping or soft-play at a final table ever comes up, that is a separate integrity problem covered in our chip dumping guide. ICM is about honest math on an honest table.