When more chips are not enough.

After-play study. This is a small, explicitly simplified model, not an actual tournament hand or an ICM strategy recommendation.

Winning more chips and winning more prize money are different objectives.

A tournament chip cannot be cashed out at a fixed price. The same stack can have a different modeled prize value depending on the other stacks and the payout structure. That is why a chip-profitable gamble does not automatically improve your expected tournament payout.

The Independent Chip Model, or ICM, translates stacks and prizes into a model of tournament equity. It is useful precisely because it asks a different question from a basic pot-odds calculation. It is not a forecast that knows each player's skill, future cards or future decisions.

An example small enough to audit.

Imagine three equal stacks of 10 units each and a $100 prize pool paying $50, $30 and $20. Under standard ICM, each equally sized stack has the same expected payout: 100 / 3 = $33.33, rounded. The model assigns each player a one-third chance of each finishing position.

Now consider a hypothetical choice between keeping those stacks unchanged and taking a gamble against one other player. Winning transfers their ten units to you; losing transfers your ten to them and puts you third. There are no blinds, antes, dead money, ties or intermediate outcomes in this benchmark. It is deliberately not a literal hand with a bet waiting in the pot.

The fictional transfer gamble, with amounts rounded to cents

Outcome	Your chips	Your modeled payout
Keep the state	10	$33.33
Win the gamble	20	$43.33
Lose the gamble	0	$20.00

The break-even probability is no longer 50%.

Let q be your probability of winning this no-tie gamble. The expected prize value is q × (130 / 3) + (1 − q) × 20. Set that equal to the original 100 / 3 and solve: q = 4 / 7 = 57.14%.

For chip value, preserving ten versus winning twenty or losing everything has a 50% break-even threshold. In this constructed example, the prize model needs 7.14 percentage points more win probability. That difference is an example of a risk premium. It is not a fixed adjustment to attach to every tournament call.

At a hypothetical 55% chance to win, the gamble has an expected 11 chips rather than ten, a gain of +1 chip. But its expected payout is 0.55 × (130 / 3) + 0.45 × 20 = $32.83, which is $0.50 below preserving the state. The disagreement is about the objective, not a contradiction in arithmetic.

Does this mean a tournament player should always avoid all-ins?

No. The example removes dead money, folds, skill differences and future decisions to isolate one effect. A real decision needs its actual fold, win, loss and tie states. A profitable shove can gain value from opponents folding. A call cannot borrow that fold benefit from the player who already shoved.

The player who did nothing can gain value.

The uninvolved player's stack remains ten units. After either decisive outcome, they face a 20-unit stack heads-up and are guaranteed at least second prize. Their ICM value is (1 / 3 × $50) + (2 / 3 × $30) = $36.67.

They gained approximately $3.33 in modeled prize equity without gaining a chip. This is a useful explanation for a viewer who sees someone benefit while two other players clash. It is also why an account of a tournament hand can need context about a player who folded before the action became dramatic.

Changing the prizes changes the question.

If the entire $100 were awarded to first place in this same model, ICM value would be proportional to chips. The equal-risk gamble would again break even at 50%. A flat-prize satellite creates a different incentive structure. Bounties introduce additional rewards that cannot be represented merely by changing an ordinary final-place payout label.

These comparisons show why “near the money” is not a complete input. Prize structure, who covers whom, other short stacks and the decision's actual outcomes all matter. Do not infer a correct action from an event-stage label alone.

What this model leaves out.

Standard ICM uses stack sizes to assign finishing probabilities. It does not directly model differences in player skill, seating position, the next blind increase or how future hands will be played. A precise number can therefore be precisely conditional on a simplified model.

Likewise, the absence of an immediate money bubble does not establish that chip-value and prize-value strategies coincide. The size of a relevant effect must be evaluated for the situation; it should not be assumed to switch on at one particular field size.

A screenshot showing “ICM” does not settle a hand if the stacks or payouts are missing. Nor does a dollar-valued result certify the ranges used to calculate it. Keep the game state, model and range assumptions separately reviewable.

The minimum useful tournament-context note.

Record the number of players remaining, available payout information, relevant stack distribution, who covers whom, blind and ante structure, bounty format if any, and the action up to the decision. Identify estimates and missing information instead of inventing precise stacks elsewhere in the field.

Then write two questions: “What does the chip-value model say?” and “What changes when the prize objective is included?” Agreement is informative. Disagreement is a reason to inspect assumptions, not to select whichever answer better fits the eventual result.