The rules of hold'em are the same in both formats. The economics are not, and the difference is more consequential than most players allow for.
The structural differences
| Cash game | Tournament | |
|---|---|---|
| Chip value | Fixed. One chip is one unit of money, always | None directly. Value comes only from finishing position |
| Stack depth | Chosen; you can rebuy to the maximum whenever you like | Falls continuously as blinds rise |
| Leaving | Any time, with the chips in front of you | Only by busting or by the tournament ending |
| Blinds | Fixed | Increase on a clock |
| Variance | Moderate | High, and concentrated in rare deep runs |
| Skill expression | Deep-stacked, multi-street play | Increasingly preflop and short-stack technique as blinds climb |
The rising blind level is the engine of everything else. Because the blinds go up and your stack does not, your effective depth measured in big blinds shrinks even when you are winning pots. A 100 big blind stack is a game of position and postflop skill; a 12 big blind stack is a game of shove-or-fold, where most of the decision has been made before the flop is dealt. A tournament walks you through both, and rewards very different competences at each end.
Why chips won are worth less than chips lost
This is the whole of ICM in a sentence, and it follows from how prizes are paid.
Consider the simplest possible case: a satellite where the top five finishers each win an identical seat, and sixth place wins nothing. Once you have enough chips to be safe, extra chips are worth exactly zero to you — no prize is larger than a seat. But the chips you could lose are worth everything, because losing them costs you the seat. In a satellite, the chip leader should be folding hands that would be trivial calls in a cash game, and it is not close.
Ordinary tournament payouts are a softened version of the same shape. The prize for first is many times the prize for the bubble, but it is never proportional to the chips. Doubling your stack does not double your expected prize money. So every chip you accumulate is worth slightly less than the one before, and every chip you risk is worth slightly more than the one you would gain.
A worked three-handed example
The Independent Chip Model converts stacks into a share of the prize pool by estimating each player's chance of finishing in each position. Enter three stacks and three prizes:
With the default figures — stacks of 5,000, 3,000 and 2,000 and a 50/30/20 payout — the model gives:
| Seat | Chips | Chip share | ICM share of the pool |
|---|---|---|---|
| A | 5,000 | 50.0% | 38.4% |
| B | 3,000 | 30.0% | 32.8% |
| C | 2,000 | 20.0% | 28.9% |
The chip leader holds half the chips and is entitled to 38.4% of the money. The short stack holds a fifth of the chips and is entitled to 28.9%. That transfer is not a quirk of these numbers; it happens in every tournament with a flat-ish payout, and it gets more extreme the closer the payouts are to each other.
The calculator uses the standard method of estimating finishing positions from chip counts. It is a model, and it makes assumptions a real table violates — it ignores position, blind level and how well anyone plays. Treat the output as the correct order of magnitude for the pressure you are under, not as a decimal-precise instruction.
What ICM actually changes at the table
- Marginal calls become folds for the big stack. A coin flip that is break-even in chips is a loss in money, because you gain less by winning than you lose by losing.
- Short stacks gain leverage on the bubble. Everyone at risk of busting behind you is playing tighter than the cards justify, which is the entire mechanism behind bubble aggression.
- Pay jumps matter more than chips. Near a big step up in prize money, survival is worth an amount that has nothing to do with your equity in the current pot.
- Satellites invert everything. Where the prize is a flat seat, the chip leader has almost nothing to gain and should be close to unwilling to play a pot.
None of this applies in a cash game, where a chip is a chip and the only question is whether the pot odds are met. The moment a payout structure exists, they stop being the whole answer.
Variance and what it demands
Tournament results are lumpy in a way cash results are not. A large-field tournament pays a small fraction of the field, so a genuinely strong player still loses on the majority of entries and makes their return on rare deep runs. That distribution is why tournament players need a bankroll measured in hundreds of buy-ins rather than tens, and why an unprofitable-looking year proves considerably less than it appears to. See bankroll management.