You call an all-in, see that you are ahead, and lose on the river. The hand history records a loss. Your review has a different job: work out whether calling was better than folding with the information available at the time. Expected value and variance belong in that review, but they describe different properties of the decision. Confusing them makes it easy to praise a bad call that won or abandon a sound line after a few difficult results.
Expected value starts with a defined decision
Expected value, or EV, is the probability-weighted average of possible outcomes. For a poker decision, you first need to identify the alternatives, the money or chips at stake, and the range of hands your opponent can hold. A result has meaning only relative to that setup.
Consider a simplified calculation with no further betting. One action produces a net gain of 10 units half the time and a net loss of six units half the time. Its EV is two units: half of 10, minus half of six. Those units could represent chips in a teaching example. They should not quietly switch to tournament prize money halfway through the calculation.
The two-unit average is a property of the model. Neither individual outcome equals it. That is why a hand can be correctly evaluated without its actual result resembling the calculated EV.
The estimate also depends on your assumptions. Change the opponent’s range, the chance of a fold, or the cost of a later decision, and the calculation changes. Reviewing those inputs is more useful than treating a solver number as independent of the situation entered.
Tournament Poker Edge’s guide to finding leaks using a poker database makes a useful point about comparing alternatives: a result measured from the start of a hand can look negative even when it improves on folding after posting a blind. The comparison needs a consistent starting point.
Variance measures the spread around that average
Now compare another hypothetical action that gains 18 units half the time and loses 14 units half the time. Its EV is also two units, but its results lie much farther from that average.
Variance measures the average squared distance from the mean. The first example has variance of 64 squared units. The second has variance of 256. Their standard deviations, the square roots of those figures, are eight and 16 units. Equal EV clearly does not imply the same distribution of outcomes.
The distinction between an average and a distribution also appears in slot terminology, although the quantities are defined differently. Return to player describes theoretical payouts relative to stakes over repeated play. Volatility concerns how those payouts are distributed, including the balance between smaller results and less frequent larger ones. Spin Casino’s explanation of real money online slots in Canada uses both terms, making it a useful comparison for why one headline percentage cannot describe every aspect of a game’s outcomes.
The comparison has a precise limit. A slot’s RTP describes a specified game model. A poker player’s estimated return depends on the decisions being evaluated, the opponents and the assumptions used. In both cases, an average leaves unanswered questions about the spread of results.
Nor is a commercial volatility label a substitute for a numerical variance calculation. When comparing two poker lines, specify what the outcomes measure and where the probability estimates come from.
Reducing the spread need not change the mean
There are poker situations in which the distinction can be isolated. After an all-in has been called in a cash game that allows multiple runouts, players can divide the pot across more than one board. The betting decision is already complete. What changes is how the remaining cards determine each share of the pot.
Stewart N. Ethier’s paper on variance reduction in Texas hold’em and video poker examines this setting mathematically. It quantifies reductions in variance while preserving the mean return under the models studied. The hold’em treatment accounts for successive runouts being dealt without replacement.
That is a specific result about conditional outcomes after an all-in. It does not establish that a different tournament strategy preserves EV, or that every form of tournament variance can be reduced in the same way. Choosing a different betting line can change both its mean and its spread.
In tournaments, define what value means
Tournament review adds another question: are you measuring expected chips or expected prize money? Chip EV treats additional chips linearly. Prize-money evaluation must account for the payout structure and how stacks compare. The Independent Chip Model estimates prize equity from those stacks and payouts, using assumptions rather than a prediction of every future hand.
A decision that gains chips on average can therefore differ from the decision that maximizes estimated prize equity. Calling that difference simply “avoiding variance” skips the actual calculation.
For your next hand review, write down the objective before opening the result. Record the alternatives, opponent ranges, stack sizes, and relevant payouts. Then ask two separate questions: which action has the higher estimated value, and how are its possible outcomes distributed? The river card answers neither question by itself.