ICM is arithmetic, not mystique
The Independent Chip Model does exactly one thing: it turns a stack of chips into a share of the tournament value that is still to be distributed. It never asks how well you play, who is sitting on your left, or what the blinds are about to do to you. It asks a single question over and over — if the remaining finishing positions were handed out by chip weight, how often would each player land in each one?
That is the entire model. Most players only ever meet ICM through a calculator, which is a shame, because the machinery is three rules and one sum. Learn to run it on paper for a two- or three-player endgame and you will stop needing the calculator for the spots that come up constantly.
The three rules the model runs on
- Chips buy lottery tickets for first place. A player’s chance of finishing first is exactly their share of the chips in play. Holding 5,000 of 10,000 chips means a 50% chance at the top spot, regardless of who is holding them.
- When a player is removed, the survivors get re-weighted. To find second place, imagine the first-place finisher’s chips are simply gone, then ask who holds the biggest share of what remains. That re-weighting step is the whole trick, and it is why the model is called independent — each finishing position is resolved as a separate, chip-weighted draw.
- Nothing else exists. No blinds, no position, no skill, no future hands. Those omissions are the model’s weakness and also the reason it stays useful: it isolates the pure structural pressure of a tournament.
Put together, that gives one formula worth memorising:
Equity = Σ (chance of finishing in position p) × (share of the remaining tournament value attached to position p)
And one line that does most of the heavy lifting, the second-place recursion:
P(you finish 2nd) = Σ over every other player i of [ P(i finishes 1st) × your chips ÷ (total chips − i‘s chips) ]
Read that slowly. For each possible winner, you ask what fraction of the survivors’ chips you own, then weight that fraction by how likely that winner was in the first place. Add the results together and you have your second-place chance. Third place follows the same pattern one level deeper, and in a three-handed endgame that is as deep as you need to go.
Express the sharing schedule as fractions of what is left
Here is the habit that makes ICM tractable. Near the end of a tournament, the only thing that matters is the share of the value still to be distributed. The numbers from the start of the event are irrelevant — what matters is the remainder and how it splits across the positions that are still live.
| Position still live | Share of remaining value | As a percentage |
|---|---|---|
| 1st | 3/5 | 60 |
| 2nd | 2/5 | 40 |
| Total | 1 | 100 |
Whenever you sit down to do this by hand, convert the schedule into fractions of the remaining value first. Fractions keep the arithmetic honest and let you check your work — the shares must always total exactly one, and so must your equity figures.
Worked example: three players, two remaining positions
Three players are left. Two positions still receive a share of what remains: first takes 60%, second takes 40%. The stacks are as follows.
| Player | Chips | Share of chips in play |
|---|---|---|
| A | 5,000 | 50% |
| B | 3,000 | 30% |
| C | 2,000 | 20% |
| Total | 10,000 | 100% |
Step 1 — first-place chances. By rule one, these are the chip shares: A 0.50, B 0.30, C 0.20.
Step 2 — second-place chances. Work through the recursion for each player. Player A is the longest calculation, so show it in full:
P(A finishes 2nd) = P(B 1st) × 5,000 ÷ 7,000 + P(C 1st) × 5,000 ÷ 8,000
= 0.30 × 0.7143 + 0.20 × 0.6250
= 0.2143 + 0.1250
= 0.3393
The same pattern gives P(B finishes 2nd) = 0.3750 and P(C finishes 2nd) = 0.2857. All three second-place probabilities sum to 1.000, which is the check that you have not slipped.
Step 3 — convert to equity. Multiply each probability by the share attached to that position, then add. For player A: 0.50 × 60 + 0.3393 × 40 = 30.00 + 13.57 = 43.57 equity units.
| Player | P(1st) | P(2nd) | Equity units (out of 100) | Share of chips |
|---|---|---|---|---|
| A | 0.5000 | 0.3393 | 43.57 | 50% |
| B | 0.3000 | 0.3750 | 33.00 | 30% |
| C | 0.2000 | 0.2857 | 23.43 | 20% |
Read the last two columns against each other. A owns half the chips and 43.57 equity units. C owns a fifth of the chips and 23.43 equity units. The big stack is worth less than its chip share; the short stack is worth more. That gap is not a quirk of this example — it is the permanent signature of ICM, and it is where every practical adjustment comes from.
The number that changes your decisions: bubble factor
Now the part that actually affects your play. Player B, sitting on 3,000, faces a shove from A. Player C has folded and is not involved, and for clarity we will treat both players’ full stacks as on the line — no chips from anyone are already committed in this hand.
Three outcomes matter to B:
- B folds. B keeps the stack and the equity that comes with it: 33.00.
- B calls and wins. The stacks become B 6,000, A 2,000, C 2,000. First-place chances are 0.60 / 0.20 / 0.20. B’s second-place chance is 0.20 × 6,000 ÷ 8,000 + 0.20 × 6,000 ÷ 8,000 = 0.30, so B’s equity becomes 0.60 × 60 + 0.30 × 40 = 48.00.
- B calls and loses. B is gone, and so is the equity: 0.
So the call risks 33.00 equity units to gain 15.00. Divide the risk by the reward and you get B’s bubble factor:
33.00 ÷ 15.00 = 2.2
The bubble factor converts directly into the equity B needs to break even on the call:
Required equity = bubble factor ÷ (1 + bubble factor) = 2.2 ÷ 3.2 = 0.6875
B needs roughly 68.75% against A’s shoving range to call correctly. Cross-check it the direct way: if B wins with probability w, the call is break-even when w × 48.00 = 33.00, so w = 0.6875. The two methods agree, which is a good sign that the arithmetic is clean.
Compare that to the same decision in a ring game, where chips convert to value at a flat rate and B would call with anything better than a coin flip. The tournament is asking B for eighteen extra points of equity before the call becomes correct. That is the entire practical content of ICM in one number.
Bubble factor to required equity, at a glance
| Bubble factor | Equity needed to call | What it feels like |
|---|---|---|
| 1.0 | 50.0% | Ring-game logic — chips are just chips |
| 1.5 | 60.0% | Mild pressure, early steps in the schedule |
| 2.0 | 66.7% | Standard bubble, medium stack against a big one |
| 2.2 | 68.8% | The example above |
| 3.0 | 75.0% | Short stack facing the one player who covers them |
Two things to notice. First, the required-equity curve flattens as the bubble factor climbs — going from 2.0 to 3.0 buys you less extra tightness than going from 1.0 to 2.0. Second, this table only tells you what a break-even call looks like; it says nothing about whether you should be the one shoving.
A five-step checklist for any endgame spot
- Strip the schedule to what remains. Write the live positions as fractions of the value still to be distributed. Ignore every number from the start of the event.
- Write every stack as a share of total chips. Those shares are the first-place probabilities, no adjustment needed.
- Run the re-weighting step only for the players who matter to you. In a three-handed pot you need two of the three second-place figures at most.
- Price the decision. Equity lost if you are wrong, divided by equity gained if you are right, gives your bubble factor.
- Convert. Required equity equals bubble factor over bubble factor plus one. Compare that threshold to the actual hand range you are facing.
A useful sixth step: always compute the ring-game threshold alongside it. If the chip-only math says 50% and ICM says 69%, the nineteen-point gap is the tournament talking, and it should change the hand you need without changing anything about how you read the opponent.
Where the model quietly fails
ICM is a floor, not a ceiling. It assumes every chip is equally likely to be the chip that decides the tournament, which is false in a single hand but reasonable across a whole endgame. It assumes no one has an edge, so a stronger player’s stack is systematically undervalued by the model. It ignores position, which in a three-handed game is nearly everything. And it is blind to the blinds — the next increase can move every stack’s relative weight before you have played a hand, which is why future-game simulation exists as a supplement rather than a replacement.
The practical framing: ICM tells you how much tighter than chip-neutral you must be, not how to play the hand. A high bubble factor makes a marginal call a clear fold and simultaneously makes a marginal shove from the covering stack a clear shove. Both adjustments come from the same calculation.
Why small-field club and online-club tournaments shift the numbers
The math above stays the same, but three structural features of smaller club-style events move the inputs.
- Flatter schedules mean lower bubble factors. When the step up between the last two positions is modest, the equity you risk is closer to the equity you gain, and the bubble factor drifts toward 1. You end up playing closer to chip-neutral than a large-field final table would suggest.
- Faster levels compress the re-weighting. More hands per blind increase means stacks change relative weight more often, so a snapshot calculation ages quickly. Recompute after every significant pot rather than every orbit.
- Small fields mean short-handed play is the norm. Three- and four-handed endgames are exactly where the hand arithmetic is easiest, so a player willing to do it on paper has a real edge over one who only consults software between sessions.
None of this requires a tool. In a three-handed game with a two-position schedule, the whole calculation is four divisions and three multiplications. Practising it there is how the adjustment becomes instinctive in a full-ring final table, where the numbers are bigger but the structure is identical.
Laws and rules covering poker vary widely by location and by operator. This article is informational only and is not legal advice or gambling advice.
