The idea in plain terms
A play has negative expectation when, averaged over a very large number of repeats, the player ends up behind by a predictable amount per play. The average is a property of the game, worked out from its payouts and probabilities. It is not a forecast for any single play, and it does not mean the player loses every time or every session.
The calculation is the expected value of one play: multiply each possible result by its probability and add them up. When that total is below zero, the play has negative expectation. When it is above zero, the expectation is positive. When it is exactly zero, the game is fair.
A small worked illustration
The numbers below are invented to keep the arithmetic simple. They are not figures from any real game.
Imagine a fair coin flip with a stake of 1 unit. The player wins 0.9 units if it lands heads and loses 1 unit if it lands tails. Each outcome has probability one half.
- Expected result per play: 0.5 × 0.9 + 0.5 × (−1) = 0.45 − 0.5 = −0.05 units.
- That is an average loss of 5% of the stake on each play.
- Over 1,000 plays, the expected total is 1,000 × (−0.05) = −50 units.
No single play ever loses 0.05 units. Each play either wins 0.9 or loses 1. The −0.05 only appears as an average across many plays.
Short-run results
In the short run, luck dominates. The spread of results can be measured with the standard deviation. For this game the standard deviation of one play is 0.95 units. Over n plays, it grows roughly with the square root of n, while the expected loss grows in direct proportion to n. Using the normal distribution as an approximation, the picture looks like this:
| Plays | Expected result | Typical spread | Chance of being ahead (approx.) |
|---|---|---|---|
| 100 | −5 units | about 9.5 | roughly 3 in 10 |
| 1,000 | −50 units | about 30 | roughly 1 in 20 |
| 10,000 | −500 units | about 95 | far below 1 in a million |
After 100 plays, finishing ahead is a perfectly ordinary outcome, which is one reason a negative-expectation game can feel winnable. After 10,000 plays, the average loss is more than five times the typical spread, and finishing ahead becomes extremely unlikely. These percentages are approximations, and the exact values depend on the details, but the direction is dependable: the longer play continues, the more the built-in average dominates.
Why this is not “evening out”
A common misreading is that wins and losses balance, so a lucky run must be followed by an unlucky one. That is not how it works. The coin has no memory, and past results do not change future odds. What the law of large numbers actually says is that the average result per play settles toward the expected value. It does not say that the total gets pulled back to zero or that a gap gets repaid. In this example the average per play drifts toward −0.05, while the total loss keeps growing.
This is also why a hot streak is weak evidence about the game. A small sample size can show almost anything. Someone who is ahead after 50 plays has learned very little about whether the game’s expectation is positive or negative.
Where negative expectation shows up
The idea appears whenever the payout is lower than the odds justify:
- Casino games. The house edge is the average share of each stake the casino keeps. Roulette-style and slot-style games are built so that this share is positive for the casino.
- Betting. Prices usually include a bookmaker margin, so bets placed at those prices carry negative expectation on average for the bettor.
- Lotteries. A ticket typically returns less on average than it costs, which is why lottery odds are so long relative to prizes.
Some activities do not have a fixed built-in average. Poker played against other people, for instance, depends on skill and on who is at the table, though fees taken by the venue can still push the average below zero for many players. Even where skill matters, results are uncertain, and nothing here suggests that anyone can reliably profit.
What negative expectation does not tell you
- It does not predict the next play or the next session.
- It does not mean every player loses the same amount. Individual results vary widely.
- It does not disappear if stakes are adjusted. Changing stake size changes how fast the average loss accumulates and how wild the swings are, but the sign stays negative. The average loss is a share of the total amount staked, so staking more means the expected loss is larger in money terms.
- It does not stop being relevant because of a lucky run. Variation around the average is normal, and a win is not evidence that the odds changed.
Putting it into practice as a reader
Seeing the average clearly can make decisions about entertainment more honest. If a game has negative expectation, the money staked over time is best thought of as the price of playing, not as an investment. The expected cost per unit staked can be estimated; the actual outcome of any given night cannot.
Gambling can be harmful, and no pattern in past results makes a loss more or less likely next time. Anyone who finds it hard to stop, or who is tempted to play to win money back, can contact a local support service for problem gambling.