Two separate ideas
Casino games are games of chance with a built-in house edge. No strategy, system, pattern or timing changes the odds in the player’s favor over time, and most players lose money over time. Nobody should gamble money they cannot afford to lose. Variance does not alter that. It describes how results scatter around the average, and the way that scatter feels from the player’s seat can be very misleading. If gambling is causing problems, a local support service can help.
Two numbers describe any bet. The first is the expected value, the average result per bet in the long run. For casino games it is negative for the player. The second is the variance, which measures how far individual results typically land from that average. Its square root, the standard deviation, is usually easier to picture because it is in the same units as the money bet.
Variance in different games
Every game has variance, though it differs widely.
- Slots can have enormous variance, because a small chance of a very large payment sits alongside many losing spins.
- Roulette varies by bet. An even-money bet produces results that are mostly small wins or losses. A single-number bet produces long losing runs and rare big payments.
- Blackjack has moderate variance per hand, but sessions still swing widely, which is why blackjack variance is a topic of its own.
- Dice and card games vary according to the bets chosen and the payouts attached to them.
A bet that pays a large multiple on a small chance is high variance. A bet that pays about even money on a chance near half is low variance. Neither is cheaper on average because of its variance, since the cost depends on the house edge and the amount wagered.
A worked example, hypothetically
Take a single-zero wheel with 37 pockets and a $10 bet, and compare two bets.
Even-money bet on a color. There are 18 winning pockets out of 37. A win pays $10, a loss costs $10. The expected result is (18/37 x 10) - (19/37 x 10) = -$10/37, which is about -$0.27 per bet. The standard deviation of a single bet is just under $10.
Straight-up bet on a number. A win pays $350 (35 to 1) with probability 1/37, and a loss costs $10 with probability 36/37. The expected result is (350 - 360)/37, which is also about -$0.27 per bet. The standard deviation, though, is about $58 per bet.
Both bets cost the same on average. After 100 bets, the expected loss is about $27 for either. But the typical spread of results after 100 bets is about $100 for the even-money bet (the single-bet figure times the square root of 100, which is 10) and about $584 for the single-number bet. The even-money player’s result will often land within roughly $100 either side of the average loss. The single-number player’s result is far more scattered: with no hit at all, the loss is the full $1,000 wagered, while one or two hits can leave the player ahead or only modestly behind, because each hit pays $350. The average is the same; the experience is not.
These are illustrative numbers for one standard bet type, and real sessions will differ.
How swings feel
Short-term variance shapes feelings about a game in predictable ways.
- A winning session feels like skill or a good machine. A losing one feels like bad luck or a rigged game. Both reactions treat noise as information.
- Early wins can feel like confirmation. Early results carry the same randomness as later ones, but they color beliefs about the whole experience, which is part of why beginners luck is so persuasive.
- A loss tends to weigh more than an equivalent win. This is the pattern behind loss aversion, and it is often paired with a strong urge to win back what was lost.
- Memory favors the dramatic. The big win is retold. The many long, quiet losing sessions fade, so the overall record looks better than it was.
How variance misleads about the real odds
The law of large numbers says that the average result settles toward the expected value only across a very large number of bets. A player’s own record is a tiny sample. Several errors follow from reading it as if it were large.
- Concluding a game is generous. A few winning sessions on a high-variance game are well within what chance produces, and they say almost nothing about the cost over thousands of bets.
- Concluding a game is rigged. A long losing run on a game with a negative expectation is also ordinary. Without a very large number of recorded results, the two cannot be told apart.
- Expecting a rebound. After a run of losses, it can feel as though a win is owed. This is the gamblers fallacy: independent results do not balance out on a schedule.
- Reading luck as skill. In games where choices matter little, a good run is rarely informative about decisions.
The house edge shows up in the long run because the operator sees an enormous number of bets. An individual player sees a small number, with variance dominating.
Matching expectations to the spread
Understanding variance is useful for setting realistic expectations, not for gaining an edge. In practice that means:
- Expect that sessions can end well above or well below the average, and that the average cost is still there underneath.
- Recognize that bigger swings need a larger buffer to avoid hitting the end of a planned budget early, which is part of bankroll management.
- Remember that raising the stake after a loss increases the swings, not the chances.
- Fix a spending limit in advance, treat it as an entertainment cost, and stop when it is gone.
A result in the short term, good or bad, is mostly the noise of variance. It is not a reliable guide to the odds of the game or to what will happen next.