Tournament poker, stack pressure, and ICM
10.1 Cash-game linearity versus tournament nonlinearity
In a cash game, one additional chip of a fixed denomination is approximately worth one additional unit of cash, ignoring rake and transaction constraints. In a tournament, chips cannot be cashed out on demand. They are claims on future placement probabilities. Doubling your stack does not generally double your expected prize money, while losing your last chip sends your prize equity for future positions to zero.
10.2 The independent chip model
ICM is a model that maps current tournament stacks and the payout schedule to expected prize money. Its core assumption is that a player's probability of finishing first is proportional to their share of the chips, then second-place probabilities are computed recursively after removing a hypothetical winner, and so on. It ignores skill differences, position, blinds arriving next, and future strategic edges, but it captures the key concavity of tournament chip value.
10.3 A four-player example
Four players each have 25 chips. Prizes are 500, 300, 200, and 0. Symmetry gives each player 250 in ICM prize equity. Now Player A doubles through Player D: stacks become 50, 25, 25, 0. ICM gives A about 383.3, B about 308.3, C about 308.3, and D zero.
| Player | Before | After double-up/bust | ICM equity after |
|---|---|---|---|
| A | 25 | 50 | 383.3 |
| B | 25 | 25 | 308.3 |
| C | 25 | 25 | 308.3 |
| D | 25 | 0 | 0 |
A gained 25 chips but only about 133.3 in prize equity. D lost 25 chips and lost the full 250. This asymmetry is the tournament risk premium in miniature: losing chips can hurt more in money terms than an equal chip gain helps.
Tournament chips do not map linearly to prize equity
10.4 Bubble pressure
Near the money bubble, a medium stack may rationally decline a small positive-chip-EV gamble if losing greatly increases the chance of receiving zero while winning only modestly increases prize equity. Large stacks can sometimes exploit this by applying pressure, while the shortest stacks may have less to preserve. The exact effect depends on stack distribution and payouts, not a slogan such as "tighten up."
10.5 Satellites: the extreme case
In a satellite tournament many finishing places win the same prize. Imagine ten players remain and nine identical tickets are awarded. Once your stack is extremely likely to survive into the top nine, accumulating extra chips has almost no prize upside, while busting costs the entire ticket. This is why a hand that is a clear positive-chip-EV all-in can be a poor money-EV call in a satellite.
10.6 Stack pressure and m
The MIT tournament lecture uses Harrington's M ratio as one measure of tournament pressure. M equals your stack divided by the total forced cost of one full table orbit: small blind + big blind + all antes you personally absorb over that orbit. Modern discussion often uses big blinds directly, but M highlights how antes accelerate effective decay.
Tournament pressure
M = stack / (small blind + big blind + total antes paid per orbit)10.7 Psychology: EValuate the decision, not the last result
Tournament variance is severe. A good all-in can lose; a bad call can win. The course emphasizes process over streaks: judge whether the action was positive EV under the information available, not whether the next card rewarded you. This is simply statistical discipline applied to a noisy payoff process.