# ICM and the Tournament Ladder > ICM is why a plus-chip shove on the bubble can still be a punt. Bubble factor, pay jumps, cover vs covered, and when min-cashing is the leak. Canonical: https://unmatchedpoker.com/strategy/icm-and-tournament-ladder Publisher: Unmatched Poker — https://unmatchedpoker.com/ Last reviewed: August 31, 2026 Note: Educational Hold'em magazine. Not a poker room, not financial advice, not a signup funnel. 18+/21+. --- Strategy · MTT # The bubble changes the math If you already price ICM into a 14-big-blind call, this is not a primer that talks to you like you just found the acronym. If you do not, you will leave knowing why the same Ace-Jack that is a chip-ev snap on the bubble is a fold that feels sick — and why that fold can still be the winning play. ## The short version Chip EV asks whether a line adds chips. ICM ($EV) asks whether it adds money from the prize pool. On the bubble those numbers can point opposite ways: folding Ace-Jack into a cover can be the profitable play even when Nash-in-a-cash-game would jam. Last reviewed August 31, 2026 Hold'em · Tournaments · ICM Chip EV (cEV) asks a simple question: does this action add chips, in expectation, to your stack? Independent Chip Model EV ($EV, or prize-pool equity) asks a different one: does this action add money, in expectation, to what you will take home when the tournament is paid out? Those two numbers agree when everyone is paid the same, which is never, or when you are so deep and so far from a pay jump that the ladder is a rumor. They disagree on the bubble, at pay jumps, at final tables, and in satellites. Disagree is a soft word. On the bubble they can point in opposite directions. ICM is a model. It estimates your share of the remaining prize pool from stack sizes and payouts, assuming that every chip is equally likely to win the tournament. That assumption is wrong in the same way all useful models are wrong: blinds go up, skill exists, and a 40-big-blind stack is not a 4-big-blind stack with extra zeros. It is still the right first-order correction to "chips are chips." If you ignore it, you will call off with the tenth-best stack on the stone bubble because "I never fold Ace-Jack," cash 12th in a 13-pay, and wonder why your graph looks like a ski jump. This is [MTT](/formats/mtts) math, not a cash-game argument in a hoodie. Cash games do not have a ladder. They have a drop. Different tax, different article. ## What ICM is, without the TED Talk Take three players, one prize pool, stacks that do not match. ICM says the chip leader's equity is less than their chip share, and the short stack's equity is more than their chip share. Why: the short stack is already looking at a third-place payout they cannot fall below without busting, and the leader cannot turn every extra chip into first place because first place is capped. Extra chips have declining value. That decline is the whole plot. Two-player ICM is just chip chop: if you have 70% of the chips heads-up for a 60/40 prize split after the bubble of two, you are not "owed" 70% of the remaining pool. You are owed the second-place money plus 70% of the difference between first and second. That is why heads-up ICM and chip EV converge — there is no third-place trapdoor. Add a third player and a bubble, and the trapdoor opens. Three-way ICM, then a flip that is −$EV Stacks: A 5,000, B 3,000, C 2,000. Total 10,000. Payouts: 50 / 30 / 20 (percent of the remaining pool). Call the pool 100 units so the numbers are percents. Player C, the short stack, is first with probability 2,000/10,000 = 20%. Second-place probability for C: when A wins (50%), C beats B with 2,000/(2,000+3,000) = 40%, contribution 0.50 × 0.40 = 0.20. When B wins (30%), C beats A with 2,000/(2,000+5,000) ≈ 28.6%, contribution 0.30 × 0.286 ≈ 0.086. P(second) ≈ 28.6%. P(third) ≈ 51.4%. C's ICM equity: 0.20×50 + 0.286×30 + 0.514×20 ≈ 10 + 8.6 + 10.3 = 28.9 units. Chip share is 20. The model is already paying C extra for not being able to finish worse than third without busting. Now C calls all-in against A in a pure 50/50. Ignore blinds for a second; we are isolating the ladder. Half the time C busts and is locked into 20 units (third). Half the time C doubles to 4,000, A falls to 3,000, B stays 3,000. Recompute C: P(first) = 40%. P(second) ≈ 34.3%. P(third) ≈ 25.7%. ICM ≈ 0.40×50 + 0.343×30 + 0.257×20 ≈ 35.4 units. Expected ICM after the flip: 0.5×35.4 + 0.5×20 = 27.7 units. Starting ICM was 28.9. The coin flip is about −1.2 units of prize pool, and 0 cEV. That is the sentence. A fair race can be a money loser because busting collapses you onto the bottom payout while doubling does not add a symmetric amount of first-place equity. Give C a 55% chip edge. Expected ICM: 0.55×35.4 + 0.45×20 ≈ 28.5. Still slightly below 28.9. You need roughly 60% equity in this toy tree before the call is +$EV. "We're ahead" is not a reason. Ahead by how much, against this ladder, is a reason. That toy is three-handed with a flat-ish payout. Real bubbles are worse for the medium stack. Ten left, nine paid, you sitting 11th of 10 in chips with 18 big blinds, and a shorter stack shoving: your call can be +cEV and a fireable offense in $EV. We walk one of those below. We also wrote the hand as a history: [the bubble fold that feels sick](/hands/bubble-fold-that-feels-sick). ## Bubble factor, in English Bubble factor is a ratio: how many chips you lose, in prize-equity terms, when you bust a chip, versus how many you gain when you win one. A bubble factor of 1 means chips are chips. A bubble factor of 2 means losing a chip hurts twice as much as winning one helps, so you need 2-to-1 the chip equity a cash-game player would need. Medium stacks on a hard bubble post the biggest numbers. The chip leader's bubble factor is lower — they can apply pressure. The shortest stack's bubble factor is also lower — they are already looking at zero or min-cash as a coin, and they have to go through people. You, with 20 big blinds and six shorter stacks still in, are the person the model is yelling at. You do not need to compute a bubble factor at the table. You need to feel the shape: - Hard bubble, many people paid, you are mid-stack: fold hands you would never fold in a cash game. This is not becoming a nit. This is not paying a tax to people who will bust themselves. - You are the covering stack versus a short shove: you can call wider than the mid-stack can, because busting them locks the field toward the money and you still have a stack. - You are covered, and calling off puts you out: you need a bigger edge. Ace-Jack offsuit versus a 14-big-blind under-the-gun shove is the poster child. Cover versus covered is not a slogan from a training site. If you call and lose, do you still have a stack that means something, or do you take a photo with the floor person? If you call and win, do you cover the table or do you merely survive? Those four futures are the decision. "I had a pair" is not. ## Pay jumps, not just the bubble The stone bubble is the loud version. Pay jumps are the quiet version that steal more money from people who "played to win" as an identity. Final table of nine, standard ladder: 9th is a min-cash with extra zeros, 1st is a life event, and 6th-to-5th is a jump that is larger than your hourly at this buy-in for a month. When three stacks are under 8 big blinds and you have 22, you do not need to be the one who volunteers for a race. Let them race. Ladder. Then play for first when the field has thinned and your bubble factor has come back toward 1. Playing for first is not the opposite of ICM. Playing for first is what you do when the model says extra chips are worth taking. Deep in a flat structure, or when you are the chip leader with a table of desperate shorts, chip accumulation is $EV. ICM suicide is the other thing: calling off 30 big blinds with Ace-Queen on the stone bubble because a podcast said "min-cashes don't pay the bills." Min-cashes don't. Busting 12th of 11 paid, with a stack that was a lock to convert, also doesn't. You can hold both thoughts. The leak is treating "play for first" as a personality and "ladder" as cowardice. They are both just regions of the same payout function. If you cannot switch, you are not a crusher. You are a one-gear recreational with a better vocabulary. ## Satellites are a different sport In a satellite, N tickets are awarded, each the same value, and everyone else gets nothing — or a rounding-error min-cash that is not a ticket. ICM here is violent. Once there are N stacks that cannot reasonably bust, extra chips are almost worthless. Shoving 20 big blinds with Ace-five because "I need chips" is how you donate a seat. Folding Ace-Queen into a mid-stack who is also folding their way into a seat is correct so often it should bore you. Satellite grind is not glamorous. It is survival with a hard cap on the prize. If you hate it, do not play satellites. If you play them, stop applying Sunday-major continuation-bet theory on the bubble of a 20-seat feeder. The model is: do not bust. The chip leader who is bullying is not a genius. They are the only person at the table whose extra chips still have a job — denying you a seat. Fold. Repeat. Take the ticket. Play the main. That is the whole satellite book, and people still ignore it because folding for two hours does not feel like poker. ## Street by street: the +$EV fold that is −cEV Sunday major, nine-handed, 10 players left, 9 paid. Min-cash is 1.7 buy-ins. First is about 80. Blinds 6,000/12,000 with a 12,000 big-blind ante (modern "ante in the BB" style — 2,000 extra in the pot from the posting of the BB). You are in the big blind with 22 big blinds (264,000). Under the gun has 14 big blinds and has been folding into the money for two orbits, then looks down and moves in. Two mid-stacks fold. The button, 31 big blinds, folds. Small blind, 9 big blinds, folds. It is on you. [Hero's hole cards] Preflop — stone bubble, you cover, you also die if you lose UTG open-shoves 14bb. Folds to BB. Hero 22bb. AJo is not a hand you are proud of, and it is also not trash. In a cash game, or in a turbo with no ladder, this is a call. You are ahead of a 14bb UTG shove that includes pairs, suited aces, and a bunch of King-Queen. Call it 55–60% equity versus a linear shove, maybe 52–55% versus a tighter UTG panic range of 77+/AQ+. Chip EV is green. Look at the other seats before you look at your cards again. Six players have 8–16 big blinds. Two of them are in the blinds next orbit. If you fold, you are 21bb after posting, still covering most of the danger, and the field is one deal closer to bursting because those short stacks cannot wait. If you call and lose, you are out in 10th, prize zero. If you call and win, you have 36bb and you have removed one of the people who was about to bust for you. The call, priced in chips You put 14bb more into a pot that already has 14bb (their shove) + 2.5bb blinds + 1bb ante ≈ 17.5bb, plus your 1bb already in. Rough pot odds: 14 to win ~32.5, about 43% equity required in a cEV vacuum. AJo clears that versus almost any shove range UTG will have. The chip-ev calculator is not the question. The question is the 1.7 buy-ins you forfeit when the 43% happens. The $EV Sketch, not a full 10-way ICM dump: your current equity is already close to a min-cash plus a little, because eight of the other nine stacks are in more danger than you. Folding keeps that. Calling creates two branches. Win: you jump toward a pay-jump stack, maybe 3–4 extra buy-ins of equity if you then play well. Lose: you go from "almost locked 1.7 plus a shot at 4–8" to zero. The lose branch is larger than people feel, because zero is not "I took a shot." Zero is erasing a coupon the field was about to hand you. Plug in a conservative 55% win, +3 buy-ins of extra equity when you win, −1.8 when you lose (the 1.7 min-cash you were going to convert, plus a sliver of extra). Expected extra: 0.55×3 + 0.45×(−1.8) = 1.65 − 0.81 = +0.84. That version is still green — if winning really adds 3 buy-ins of ICM. On a flat min-cash bubble with a huge field still above you in chips, winning a 14bb pot does not add 3 buy-ins. It adds a handful of big blinds and one fewer opponent. Reprice the win at +0.8 buy-ins of $EV and the same math goes negative: 0.55×0.8 + 0.45×(−1.8) = 0.44 − 0.81 = −0.37 buy-ins. That is a fold. It will feel sick. It is supposed to. If you had called [Qc 7d 2s — you already put the chips in] There is no flop decision. The money went in pre. They show Ace-Queen. You lose. That's a cooler, not a leak, if the call was +$EV. If you called because "Ace-Jack is too strong to fold" while six shorter stacks were drowning, you spewed prize pool. Same cards. Different crime. The full walk-through of this shape of fold lives in [the bubble fold that feels sick](/hands/bubble-fold-that-feels-sick). ## Leak: ICM suicide as a brand "I never fold a pair on the bubble" is not courage. It is a refusal to price the ladder. Pocket sevens versus a chip-leader shove that covers you, with four shorter stacks still in, is a fold more often than Twitter wants it to be. You are not folding sevens forever. You are folding sevens until two of those short stacks die, the bubble bursts, and extra chips start paying you again. The other leak is the mirror: min-cashing as a goal. If you are the chip leader with 40 big blinds and you open-fold Ace-Queen to a 12bb shove because "we're on the bubble," you are not using ICM. You are hiding. Leaders apply pressure. Medium stacks take the pain. Know which one you are before you write a speech. ## Deviate: when chip EV is the right god Leave ICM tightness when the ladder is flat, the bubble is miles away, or you have a massive chip lead and the pay jumps below you are rounding error compared to first. Early levels of a big field: play close to chip EV, including the 15bb open-shoves in [push-fold shape](/strategy/short-stack-push-fold). PKO bounties pull the same direction — a bounty is cash now, and it can justify a call that pure ICM would hate. Final table, three left, payouts still steep: mix. Heads-up: chip EV, with the chop-math mentioned above. Satellites: the other extreme. If your "deviation" is the same in all five of those worlds, it is not a deviation. It is a refusal to read the payout screen. ## Cover, covered, and who gets to bully If you cover the shover and several other stacks, calling has a side benefit: you can bust someone and still sit down. If you are covered, calling is a binary life. The same Ace-Ten offsuit is a call from the 70bb chip leader versus a 12bb button shove on the bubble, and a fold from the 16bb big blind who would be out. People hate this because it feels like the leader "gets to play poker" and you do not. Correct. That is what the chip lead purchased. If you want to be the one who gets to play poker, stop donating 18bb pots on the way to the bubble. ICM pressure is a real shove range, not a myth. Chip leaders should open wider and shove wider at 15–20bb when the mid-stacks cannot call without lighting $EV. Mid-stacks should overfold those opens. Short stacks should not cooperate by folding themselves into blinds-out death — at 6bb you are not "using ICM," you are waiting to get blinded into a worse shove. Push-fold charts still apply to the shorty; the ICM adjustment is mostly in the calling ranges of the people who cover them. If that last sentence is the only thing you take, it is already better than 80% of the bubble table talk you will hear this year. ## Min-cash versus playing for first, without the speech A min-cash is a consolation prize that is still more than zero. Taking it on purpose when you had 40bb and a playable table is a leak. Folding AJo on the stone bubble when you have 22bb and the field is full of 8bb time bombs is how you convert a min-cash plus a window into more. Those are not the same hand. If your friends cannot tell the difference, get new friends or stop explaining your folds in the lobby. ROI lives in the right tail. That is true. The way you get to the right tail in a modern field is not "call off every time I am ahead in chips." It is "survive the spots where the model charges a tax, then apply chip-ev pressure when the tax drops." People skip the first half because it is emotionally expensive. They skip the second half because bullying feels mean. You can lose money both ways. The graph does not care which speech you made. Work a second number, smaller. You are 8th of 9 at a final table. 9th just busted, 8th pays 4 buy-ins, 7th pays 6. Two stacks have 4bb. You have 19bb, not in the blinds. UTG+1 (cover, 45bb) opens, cutoff (7bb) shoves, it is on you with King-Queen offsuit. Chip EV likes a call. $EV likes a fold and a short-stack bust in the next two hands. Fold. If both short stacks double and you still have KQo in a similar spot with the jump behind you, the call gets better. Recalculate. Do not marry the fold, and do not marry KQo. ## What to study in 20 minutes You do not need to become an ICM specialist this week. You need three pictures: - A 10-left, 9-paid bubble with mixed stacks. Who can shove on whom. Who has to overfold. - A final-table pay jump with two micro-stacks. Who is allowed to gamble. - A satellite bubble with equal tickets. Who is already in, whose extra chips are worthless. Take a hand you played in each picture and ask only: if I bust, what do I get, and if I double, what do I actually jump to? If you cannot answer in a sentence, you did not look at the payouts. Looking at the payouts is the study. Solvers that include ICM (again: licensed tools, not a pirated dump) are useful after you can say the sentence. Before that, the solver is just a way to launder a call you wanted to make. For the mechanics of the shove itself — 10bb, 12bb, 15bb, antes, 9-max shape — use the [push-fold article](/strategy/short-stack-push-fold) and the [push-fold primer](/tools/push-fold-primer). ICM sits on top of that shape. It does not replace it. A 10bb stack still shoves a lot of hands. The people who cover that stack stop calling as wide. That is the partnership. And if you folded AJo, watched the shorty triple the next hand, and felt like a coward: that feeling is not information. The model already priced the times they triple. You are not entitled to a parade because you made a −cEV, +$EV fold. You are entitled to still being in the tournament. Play poker when the ladder says extra chips are on sale again. ## FAQ **What is ICM in poker?** Independent Chip Model estimates your share of the remaining prize pool from stack sizes and payouts. Extra chips have declining dollar value. Chip EV and dollar EV disagree on the bubble, at pay jumps, at final tables, and in satellites. **Why does the bubble change the math?** Busting before the money turns a stack into zero dollars, while folding and letting a shorter stack die can lock a min-cash. A chip-EV shove with Ace-Jack into a cover can be a dollar-EV punt. **Is folding Ace-Jack on the bubble a leak?** Often no. In cash it is usually a jam at 12bb. On an MTT bubble against a cover, with shorter stacks still alive, the fold can be +$EV and −cEV. Feelings are not a pay table. Last reviewed August 31, 2026.