Final Table Strategy: How to Use ICM

A final table is a different game to the one you played to reach it. The cards behave the same way; the money does not. Decisions that were routinely correct an hour earlier become expensive mistakes, and hands you would never have played become mandatory.

The reason is ICM. This is how to actually use it, seat by seat, in the spots where it decides the outcome.

The one thing that changes

Everywhere else in a tournament, chips are the objective and accumulating them is the strategy. At a final table, survival acquires a dollar value of its own, because every elimination moves you up a ladder that pays real money.

The consequence, in one line: a call that breaks even in chips loses money in dollars. The chips you stand to win are worth less per chip than the chips you stand to lose, so the equity you need to call rises above the chip-EV break-even point. That gap is your risk premium.

Everything below follows from that. If the mechanics behind it are unfamiliar, the Independent Chip Model explained covers where the numbers come from.

Who is actually under pressure

The most common misreading of ICM is that short stacks must play tightest. The opposite is true, and getting this right reorders the whole table.

The big stack. Smallest risk premium at the table. Losing a pot costs chips, not a finishing position, and cannot bust them. They can attack constantly and the table cannot fight back — this is the most profitable seat in tournament poker by a wide margin.

Medium stacks. The most constrained players in the room. They have accumulated real equity and can still be eliminated, so every marginal spot is a fold. If you have a medium stack and it feels as though you cannot play a hand, that is the model working, not you being timid.

Short stacks. Least to lose. A player with three big blinds will be all-in within two orbits regardless, so their risk premium is small and they can shove very wide. Attacking them is far less profitable than it looks.

The clean statement: pressure flows from the top of the ladder downwards, and it is heaviest in the middle. The player just above the shortest stack is the one paying the most.

The bubble

The starkest ICM spot in poker: one elimination from the money, where the difference between busting and folding is the entire minimum cash.

If you are big or comfortable, raise constantly. Everyone at the table has a large risk premium and you have almost none. You are effectively being paid a fee every orbit by players who cannot afford to fight back.

If you are medium, fold almost everything. Not because folding is inherently good, but because a fold guarantees a ladder climb and a marginal call risks the whole cash. This is the phase where hands that look like clear calls — A-J, medium pairs against a shove — are genuine folds.

If you are the shortest stack, ignore most of it. You cannot fold into the money if you will be blinded out first, and you have the least to protect. Take reasonable spots.

If two players are shorter than you, stop playing entirely for a while. Every orbit both of them survive costs them blinds, and every hand you do not play costs you almost nothing. Waiting is an active strategy here.

Calling ranges tighten far more than shoving ranges

The asymmetry that decides most final table hands: ICM restricts calling much more than it restricts shoving.

When you shove, you win the blinds and antes uncontested a large share of the time, and that outcome carries no bust risk at all. When you call, you are always in a pot for your tournament life. So as ICM pressure rises, shoving ranges stay wide and calling ranges collapse.

Concretely, at a final table with a real pay jump ahead:

• Where chip EV might call an all-in with any pair, ace-nine or king-jack, ICM might restrict it to nines and up, plus A-Q and A-K.
• Where chip EV might shove 40% of hands from the button, ICM often supports shoving more, because everyone behind is folding wider than they should.

The practical rule: when in doubt at a final table, shove rather than call. The aggressive line carries fold equity; the passive one does not.

Two things follow immediately. First, min-raising loses value here — it commits chips without generating fold equity, which is precisely the wrong trade. Shove or fold with short and medium stacks, rather than making small raises. Second, any hand that is only a marginal chip-EV call is a fold, without further thought.

Pay jumps decide how tight

Not all final tables have the same ICM pressure. What matters is the shape of the remaining payouts.

Flat structures — where fifth through ninth pay similar amounts — produce mild pressure. There is less to protect, so play closer to chip EV.

Steep structures, where first pays several times what third does, produce extreme pressure lower down and much less at the top. Near the bottom of the ladder, survival is everything; heads-up for a big first prize, it is nearly a chip-EV game again.

A useful habit before every final table: look at the payout list and find the largest jump ahead of you. That jump is the size of your risk premium in practical terms. If third to second is a small increase and second to first is enormous, laddering matters less than winning, and you should play accordingly.

Playing each seat

Big stack. Raise into medium stacks constantly. Avoid unnecessary confrontation with other big stacks — they can bust you and their premium is low too. Ease off the very short stacks; they will call wide. Your profit comes from players with something to lose.

Medium stack. Steal from the players below you, fold against the players above you. Do not call off your stack in a spot that is close in chips. Your equity is protected by not busting, and your chip accumulation should come from uncontested pots.

Short stack. Shove or fold; almost never call, and almost never raise small. Look for spots where the players behind you have the most to lose — shoving into two medium stacks is far better than shoving into the chip leader, who can call much wider.

Heads-up. ICM pressure largely evaporates. There is one jump left, both players face the same risk, and it becomes close to a chip-EV game. Loosen up substantially.

Reading the table before the hand

The habit that separates players who understand ICM from players who have read about it is doing the work before the action reaches them.

Three questions, asked once per orbit rather than once per hand:

Who is closest to busting? Not just the shortest stack in chips — the shortest stack in big blinds, accounting for who is about to post. A player with eight big blinds who is about to be in the blinds has two hands left, not eight.

Who is protecting the most? Find the medium stacks. These are the players who will fold too much, and they are your primary targets when you raise.

What is the next jump worth? If the difference between the current payout and the next one is small, everyone should be playing closer to chip EV and the folding you are counting on will not materialise.

Answering those three before you look at your cards means that when a hand does arrive, the decision is already half made — which matters, because final table decisions come with a clock and an audience.

Studying it away from the table

You cannot compute ICM while a clock is running. What you can do is build the reflexes beforehand.

Take real final tables you have played, enter the stacks and payouts into an ICM calculator for final tables, and look at what each stack was actually worth. Then take the hands where you called off and check the equity you needed. A handful of these does more than any amount of reading, because seeing a call you were proud of turn out to be a $40 mistake is not something you forget.

For the hand-versus-range side of that comparison — what equity you actually held against the shoving range — an equity calculator gives you the number that ICM's requirement is measured against.

Deals

Final table deals are an ICM calculation made explicit. A chip-chop based on ICM equity is the fair baseline: it pays each player exactly what the model says their stack is worth.

Compare any offer against that number. Above it, taking the deal is a straightforwardly profitable trade of variance for certainty. Below it, you are paying for certainty — sometimes worth doing, but you should know the price you are paying.

The one adjustment ICM cannot make for you is skill. If you are clearly the strongest player remaining, your equity is higher than the model says, and a deal at exact ICM equity is a bad trade for you.

Keeping perspective

Final tables are rare, and their results are dominated by chance. Playing every ICM spot perfectly still leaves you busting in ninth place regularly, and running a few final table outcomes through a poker variance calculator makes clear how little a handful of them proves about anyone's skill.

That is the right frame for it. ICM does not make you win final tables. It makes each one worth measurably more over a career — and at a final table, a few per cent of equity is a lot of money.