33 ICM: Bubbles, Pay Jumps & Final Tables

After this chapter you will be able to adjust your calling and shoving ranges for the dollar value of chips, rather than their raw count, near bubbles, pay jumps, and final tables.

In a cash game a chip is a dollar. You can stand up at any time, rack your stack, and turn every chip into hard currency at a one-to-one rate. Tournaments do not work this way, and the gap between how chips feel and what they are worth is the single biggest source of strategic edge in tournament poker. This chapter is about that gap.

The reason the gap exists is the payout structure. A tournament does not pay you for chips; it pays you for finishing position, and only the top sliver of finishers gets paid at all. A 1,000-runner event might pay 150 places, with the winner taking perhaps 18% of the prize pool and the player who min-cashes (finishes in the lowest paid spot) in 150th taking 0.15%. Chips are the engine that moves you up that ladder, but they are not themselves the prize. The Independent Chip Model (ICM) is the standard tool for translating a stack of chips into an expected dollar value given the remaining payouts.

Understanding ICM is what separates players who “play their stack” from players who play the money. It is also, not coincidentally, one of the areas where strong live and online regulars win the most against everyone else.

33.1 Why a chip is not a dollar

The core mathematical fact is this: chips you win are worth less than chips you lose. This is a consequence of the diminishing marginal value of chips near a payout ladder.

Imagine a satellite that pays 10 identical seats to the top 10 finishers, with 11 players left, all on roughly equal stacks. If you double your stack, you do not win two seats; there is only one seat per person, and you were already very likely to get one. Your equity barely moves up. But if you lose all your chips, you get nothing. The downside is catastrophic and the upside is nearly zero.

Regular (non-satellite) tournaments are a softer version of the same effect. The jump from 1st to 2nd is large, but the jump from “out” to “min-cash” can be just as meaningful in percentage terms, and there are many such jumps stacked along the ladder. Each one bends the chips-to-dollars curve so that it is concave: it rises steeply at the bottom (your first chips, the ones keeping you alive, are precious) and flattens out at the top (your millionth chip when you already have a monster barely adds to your cash equity).

Key idea

Tournament equity is a concave function of chips. Going from 0 to 20 big blinds changes your survival dramatically; going from 200 to 220 big blinds barely changes your expected payout. Because the curve is concave, a coin flip for your stack is a losing proposition in dollars even when it is exactly break-even in chips. This single fact drives every ICM decision in this chapter.

33.2 What the Independent Chip Model actually does

ICM takes two inputs:

The stack sizes of every remaining player.
The payout structure (the list of prizes).

It outputs each player’s dollar equity: their expected payout if the tournament were settled, right now, according to the model’s assumptions.

The model’s central assumption is simple and a little naive: the probability that a player finishes first is equal to their share of the chips in play. If you have 30% of the chips, ICM says you win 30% of the time. To compute the chance you finish second, ICM imagines you didn’t finish first, then removes each other player as the “winner” in proportion to their chip share, recomputes your first-place odds among those remaining, and so on down the ladder. Summing your probability of finishing in each paid position, weighted by that position’s prize, gives your dollar equity.

You do not compute this by hand at the table. ICM calculators do it instantly, and solvers such as HoldemResources Calculator (HRC), ICMIZER, MonkerSolver, and others bake it directly into their preflop and postflop solutions. What you carry to the table is the intuition the arithmetic produces.

A worked equity example

Three players remain in a sit-and-go that pays $500 / $300 / $200 (a $1,000 pool). Stacks:

Player	Chips	Chip %
You	6,000	60%
B	3,000	30%
C	1,000	10%

Naive “chip-chip” thinking says your 60% of the chips is worth 60% of the pool, or $600. But the top prize is only $500, and you literally cannot win $600 because you cannot win more than first place pays. ICM corrects this. Running the model gives roughly:

Player	Chip %	ICM $ equity
You	60%	~$412
B	30%	~$338
C	10%	~$249

Notice what happened. The big stack’s equity ($412) is far below its chip share ($600). The short stack’s equity ($249) is far above its chip share ($100). Player C, with a single big-blind-sized stack, is already guaranteed $200 (third pays $200) and has upside on top. The chip leader has “lost” value and the short stack has “gained” value, purely from the payout geometry. This is why min-cashing with a tiny stack on the bubble is worth fighting for, and why having a huge stack does not give you license to gamble.

33.3 The risk premium: how ICM tightens your ranges

Here is the practical engine of the whole chapter. Because doubling up gains you less equity than busting loses you, you need a bigger raw-equity edge to justify a stack-threatening confrontation than you would in a chip-EV (cash-game) world. That extra required edge is the risk premium.

In a cash game, if you are getting the right pot odds and you have 50.1% equity all-in, you call; a chip is a dollar, so any positive chip-EV spot is a positive dollar-EV spot. Under ICM, that same 50.1% all-in can be a clear fold, because the dollars you stand to lose outweigh the dollars you stand to win. You might need 55%, 58%, even 65% raw equity before a call breaks even in dollars.

The size of the risk premium depends on the situation:

Bigger near pay jumps. On the money bubble and at each pay jump, the curve is steepest, so the premium is highest.
Bigger against a covering stack. When you can be eliminated by a call or shove, ICM bites hardest. When you cover your opponent (they can’t bust you) your premium against them is small, sometimes near zero.
Bigger when other stacks are short. If there are shorter stacks who might bust for you, you have extra incentive to wait. Why risk your tournament life when someone else might bust first?
Smaller deep in chips with flat pay jumps. Early in the money, when remaining pay jumps are small relative to stacks, the premium shrinks and play approaches chip-EV.

Key idea

Risk premium is the extra equity, above the chip-EV break-even point, that you need before risking your stack. A nasty bubble can push the premium so high that hands you would snap-call with for stacks in a cash game (strong but not premium holdings) become mandatory folds. The premium is largest when you are deep on the ladder, can be eliminated, and other short stacks are present.

Bubble factor: a number for the premium

Practitioners put a number on the risk premium with the bubble factor: the ratio of the dollars you lose by busting a confrontation to the dollars you gain by winning the same chips. In a cash game the bubble factor is exactly 1.0, since a chip lost and a chip won are worth the same, so you play pure chip-EV. On a money bubble, against a stack that covers you, it might be 1.3 to 1.6; deep in a tense satellite it can climb past 2.0. A bubble factor of 1.5 means each chip you lose is worth 1.5× in dollars what each chip you win is worth. It does not mean you must win 1.5 chips for every chip risked (in a symmetric all-in you cannot win more chips than you put at risk). Equivalently, your break-even equity in a stack-off is BF / (BF + 1) = 1.5 / 2.5 = 60%, not 50%. (BF 1.0 → 50%; BF 1.3 → ~56.5%; BF 1.6 → ~61.5%; BF 2.0 → ~67%.) That is exactly the 55–65% required-equity band cited above, now pinned to a single number you can read off a solver.

Crucially, the bubble factor is pairwise: a solver computes a separate one for every pair of players at the table. Your bubble factor against the stack on your left who covers you can be 1.5 at the same moment your bubble factor against the short stack on your right whom you cover is barely above 1.0.

This is also what moves opening and open-shoving ranges, and it moves them in opposite directions depending on who can bust whom. A stack that can be covered and called should open-shove far tighter than chip-EV allows. Where a chip-EV jam from late position might be ~35–45% of hands, the same stack on a hard bubble can be squeezed down to the top ~10–15%. The covering stack goes the other way: because no one can profitably call, it can open-shove wider than chip-EV, frequently 40%+ of hands into the most trapped middle stacks.

One caveat that matters for the next example: that 40%+ figure is for stealing, jamming at trapped stacks who fold almost everything and rarely close the action. Jamming into a big blind who closes the action and can call you is a different problem. There the covering stack cannot rely on fold equity alone; it has to show a profit at showdown the times it gets called, so even a coverer keeps its jamming range into the blind meaningfully tighter, closer to the top ~25–30% than to 40%+.

A bubble calling example

A $100 online tournament is on the money bubble: 28 players left, 27 get paid, the min-cash is $180 and your $100 buy-in is otherwise gone. Effective stacks (the smaller of the two stacks in any pot) vary. You are in the big blind with 15bb. The chip leader in the cutoff (covering you many times over) open-shoves all-in, which folds out the rest of the field to you.

You look down at A♣J♦. In a vacuum, by chip-EV cash-game logic, AJo (AJ offsuit) is miles ahead of a wide button/cutoff shoving range. You’d call instantly. But run the ICM numbers and the picture changes:

Because the chip leader is jamming into your big blind (you close the action and can call), his range is the tighter “into a caller” range from the previous section, not the 40%+ steal-jam he would fire at a trapped folding stack. Call it the top ~25–30% of hands (broadly: suited aces, broadway-ish hands, most pairs, suited connectors; reads vary). Against that, AJo has only about 52–54% raw equity: the AK and AQ in his range dominate it and the pocket pairs flip-or-beat it, so it is a marginal edge at best, and a slight underdog against the tighter end of his range. (If he were somehow jamming the full 40%+ steal range, AJo’s equity would climb toward 57–60% and the call would come right back.)
You are closing the action in the big blind, so you are getting a real price. With the antes and the blinds already dead in the middle, you are risking about 14bb to win a pot of about 17.5bb, which means your chip-EV break-even is only about 44–46%, not 50%.
The ICM risk premium against this covering chip leader, on a hard bubble with several shorter stacks waiting to bust, adds roughly 8–12 percentage points of required equity on top of that pot-odds break-even. That lifts your real break-even to about 52–58%.
AJo’s ~52–54% sits at the bottom of, or just below, that 52–58% required band. At the genuine midpoints, roughly 53% equity against ~55% required, equity falls short of the requirement, so this is a fold. You surrender at most a sliver of chip-EV edge, and against the tighter end of his range none at all, to protect a near-certain min-cash and a healthy 15bb stack.

If three other players have 3–6bb and are about to be blinded out, the premium climbs toward the top of that 8–12 point range and the fold stops being close at all: let them bust.

Common mistake

Calling off too light on the bubble. The most expensive recurring tournament error is treating a strong-looking hand (AJ, KQ, 99, even AQ) as an automatic stack-off against a covering shove on the bubble. These hands are big chip-EV calls but losing, or at best break-even, dollar-EV calls. The covering stack knows this, which is exactly why they can widen their shoving range and still profit. Tighten up dramatically when you can be eliminated, you are not the shortest stack, and a pay jump is one elimination away. Conversely, do not let this make you a nit when you cover the aggressor: then the premium is small and you defend close to normally.

The flip side: ICM is a weapon, not just a constraint

ICM does not only tell you when to fold; it tells you when to attack. The same forces that make your opponents fold strong hands give the player applying the pressure enormous leverage. Call this ICM pressure.

If you are a big or middling stack who covers the table on the bubble, you can shove and re-shove relentlessly into players who cannot call without a near-premium, because they are paying the risk premium, not you. The textbook spot is a big stack to the immediate left of a medium stack who is to the left of a desperate short stack: the medium stack is trapped, since busting before the short stack would be a disaster, so they fold almost everything and you take their blinds and antes uncontested.

Key idea

On the bubble, chips in the hands of a covering stack are worth more as a threat than as a call. The correct bubble big-stack strategy is high-frequency, low-showdown aggression: attack the medium stacks who have the most to lose, and avoid the desperate short stacks who have little left to lose and will call wide.

33.4 Satellites: ICM taken to the extreme

A satellite awards seats (or tickets) into a larger event instead of a cash ladder, and usually every winning seat is identical in value. That flat top makes the chips-to-dollars curve about as concave as it ever gets: satellite ICM is regular ICM with the dial turned to maximum.

The consequences are dramatic and a little counterintuitive:

Once you have enough chips to be “safe,” stop playing hands. If 10 seats are paid and you are comfortably in the top 10 with two or three players clearly shorter, your job is to avoid all variance. You should fold hands as strong as AK, even pocket aces preflop, to a shove that risks your seat when folding keeps you safely qualified. This is the most famous satellite move and it is correct: aces win the pot ~80% of the time, but if winning the pot gains you almost nothing (you were already getting a seat) and losing it costs you the seat entirely, the 20% disaster dominates.
The bubble is brutal and binary. With 11 players left for 10 seats on roughly even stacks, nobody wants to play a pot. Two big stacks colliding is a catastrophe for the loser and a gift to the other nine. The short stack, paradoxically, has leverage: they can shove any two cards, and the medium stacks usually cannot call.
Chip accumulation has a hard ceiling. Being the massive chip leader in a satellite is nearly worthless beyond the point of safety, because all the extra chips buy you is the same single seat everyone else in the top 10 gets. There is no first-place premium to chase.

A satellite seat example

A satellite awards 10 identical seats worth $1,000 each into a championship event. Eleven players remain (the seat bubble) and you sit comfortably in the upper half with about 18bb. Two opponents are down to 4bb and 6bb and will almost certainly be all-in within an orbit or two. A stack that covers you open-shoves; the action folds to you in the big blind and you find A♠A♥.

Run the seat-EV, not the chip-EV:

Fold. You keep a healthy stack with two much shorter players about to bust ahead of you. Your chance of qualifying is roughly 90–95%, so your seat-EV is about 0.92 × $1,000 ≈ $920, and it climbs every time a short stack busts.
Call and win (~80%). You now have a huge stack, but a satellite seat is capped at $1,000, the exact same seat folding was already going to win you. Your seat-EV inches up to perhaps $990. You gained almost nothing.
Call and lose (~20%). The shover covers you, so you are out: seat-EV $0.
EV(call) ≈ 0.80 × $990 + 0.20 × $0 ≈ $792.

Folding aces is worth about $920 versus $792, roughly $130 of real money, even though aces are an 80% favorite. That is ICM at its most extreme, which is why “fold aces to lock a seat” is meant literally.

Common mistake

Playing a satellite like a regular tournament. Trying to “win” a satellite by accumulating a giant stack and busting people is a losing approach once you are near the seat-bubble. The goal is to finish in the seats, and every seat is equal. In a satellite, folding aces to lock a seat is the textbook-correct, money-maximizing play. The error is the player who calls there and high-fives their cooler hand while it costs them their qualification.

33.5 Final-table ICM dynamics

A final table is where ICM is at its most lucrative and most punishing, because the pay jumps are enormous. A typical final-table payout might run something like this for nine players (percentages of pool):

Place	Approx. % of pool
1st	30%
2nd	20%
3rd	14%
4th	10%
5th	8%
6th	6%
7th	5%
8th	4%
9th	3%

The jump from 9th to 8th is one percentage point; the jump from 2nd to 1st is ten. But notice the early jumps are still steep relative to a short stack’s equity: a player clinging to 3bb in 9th has almost nothing locked up, and laddering just one or two spots lifts their guaranteed cash by a third to two-thirds (3% → 4% → 5% of the pool), while climbing into the middle of the table doubles it. That short stack should be willing to gamble more than you’d think to climb.

Key final-table principles, all flowing from the concave curve:

Medium stacks are handcuffed; big stacks and tiny stacks are free. If you have a comfortable middle stack, you have a great deal to lose by busting before several shorter players and relatively little to gain by doubling. You should fold a lot and avoid the big stacks. The chip leader, by contrast, can apply relentless pressure, and the desperate short stack has so little equity that they can ship it in lightly to try to ladder or double.
Pick your spots against the player who covers you. Confrontations where you can be eliminated carry the full risk premium. Confrontations where you cover your opponent are nearly chip-EV, and that is where a big stack should be looking to get it in.
Ladders are real money. When two other players are critically short, sometimes the correct play is to simply fold and let them bust, banking the pay jump for free. The dollars you save by laddering one spot can dwarf the dollars you’d win in a marginal pot.
Deal-making is applied ICM. When a final table pauses to discuss a deal, the chip-chop (“everyone takes their chip percentage of the remaining pool”) overpays the chip leader and underpays the short stacks, exactly as our 3-handed example showed. An ICM deal is the fair one. If you are short, push for an ICM (or “ICM-plus-a-bit-left-to-play-for”) deal; if you are the chip leader, a straight chip-chop is in your favor. Know which one you’re being offered.

A worked final-table example

Five players remain in a tournament paying $10,000 / $6,000 / $4,000 / $2,800 / $2,000. Stacks:

Seat	Player	Chips
BTN	You	480,000 (≈24bb)
SB	Short	120,000 (≈6bb)
BB	Medium	600,000 (≈30bb)
UTG	Big	900,000 (≈45bb)
CO	Medium-2	300,000 (≈15bb)

Action folds to you on the button with 9♠9♣, blinds 10k/20k with antes. The 6bb short stack is in the small blind; the 45bb big stack has already folded; the 30bb medium is in the big blind and covers you. With the big stack out of the hand, the big blind is now the only live player who can bust you.

In a cash game, by pure chip-EV, 99 on the button 5-handed is an automatic open-raise, and a small open is standard in every framework. The open is the easy part. Where ICM actually changes a decision is one step later, in how you react if the covering big blind plays back:

The short stack in the SB has 6bb and is the player most likely to bust next. There is real value in simply not dying before they do. Every orbit they survive, they pay blinds and antes and bleed toward all-in, and every time one of them busts you ladder a pay jump for free.
So you open small, say to 2–2.25bb, which is correct under chip-EV and ICM alike. The 6bb SB is no problem: you cover him, so any confrontation with him is near chip-EV, and if he jams his desperate range you have a comfortable call. The stack that matters is the 30bb big blind who covers you.
The genuine ICM decision arrives if that big blind re-shoves all-in over your open, and here ICM flips the answer. A 24–30bb re-jam over a min-open, fired into a player who closes the action, is not the 40%-wide steal-jam from earlier; even a loose covering stack re-jams a value-leaning range here: the better pairs, the better aces, suited broadways, with only a handful of bluffs. Against a realistic range like that, 99 is no better than a coinflip in chips, a slight dog, in fact, to the overpairs and the AK/AQ it keeps running into. But so much dead money is already in the middle that your chip-EV break-even is only about 44%, so in pure chip-EV terms calling is still marginally profitable. Chip-EV says call; ICM says fold. Stacking off for your tournament life against the only player at the table who covers you, while a 6bb short stack is about to bust in front of you, means paying a risk premium of roughly 10–15 points of equity, which lifts your real break-even into the high 50s and leaves a coinflip far short. Busting in 5th ($2,000) instead of laddering past that short stack for free is exactly the disaster the concave curve warns about.

The lesson lives in the response. Raise small, keep control, and be ready to let go of a strong hand against the one stack that can end your tournament.

Drill

For your next ten online MTTs, paste the final-three-table stack situation and payout list into an ICM calculator (HRC, ICMIZER, or any free ICM equity tool) after you bust or cash. For each, find the single closest all-in decision you faced and check: what raw equity did you actually need, and what did you assume at the table? Track how often your gut over-valued a “strong” hand against a covering stack near a pay jump. Most players discover they were calling off 5–10% too light on bubbles. Re-calibrate until your instinct matches the model, then trust the instinct in real time.

33.6 Putting it together

ICM is not a separate game bolted onto poker; it is the correct accounting system for every tournament decision once real money is on the line. The whole chapter reduces to a few load-bearing truths:

Chips you can win are worth less than chips you can lose: the curve is concave, so coin flips for stacks lose money.
The risk premium tightens your calling ranges, most sharply near pay jumps, against covering stacks, and when other short stacks are present.
The same pressure you fear is a weapon when you hold it. Cover the table near a bubble and you can attack the trapped medium stacks almost for free.
Satellites are ICM at the extreme: fold aces to lock a seat without apology.
Final tables are where pay jumps are largest: ladder deliberately, respect the covering stack, and know whether a proposed deal is an honest ICM chop or a chip-chop tilted against you.

None of this is certain. ICM itself is a simplification: it ignores position, skill edges, and the future value of fold equity, and serious players adjust for those (a topic the model’s critics and its modern refinements address). The standard version above is the Malmuth-Harville model, with its chip-share finishing-order assumption; modern high-stakes solvers refine it with future-game simulation (FGS), which actually plays out future hands instead of settling the tournament instantly. But as a framework for avoiding needless busts on the bubble, it is indispensable. Learn to feel the concave curve under your decisions, and you will fold the hands that need folding and fire the shoves that need firing, while the players around you are still pretending a chip is a dollar.

The flat cash-per-chip shortcut, and its limits

A common quick valuation — cash per chip = regular prize pool ÷ total chips in play (used to price bounties and to put a rough dollar tag on a stack) — is a chip-EV approximation. It treats every chip as worth the same flat amount and so ignores exactly the ICM pressure on the regular prize pool that this whole chapter is about: near a pay jump a chip is worth far less than that average, and your first few chips far more. Use it for a fast estimate, never for a close bubble or final-table decision.