Variance (in Betting)

Variance is the gap between what a bet should return on average and what actually lands in your account after it settles. Every priced-up wager has an expected value (EV) — the long-run average outcome if you repeated it thousands of times — but no single bet, or even a run of a few dozen, plays out at that average. Variance is the statistical name for that scatter: how wildly results bounce around EV before the sample is large enough for the average to assert itself.

It matters because betting is judged on outcomes, but outcomes are a noisy signal of quality. A bettor with a genuine 5% edge can still lose money over a month. A bettor with no edge at all — or a negative one — can string together a winning week purely on variance. The size of that noise depends on the odds involved: backing short-priced favourites at 1.20 produces frequent, small swings, while backing 6.00 outsiders produces rare but violent ones. Higher odds always mean higher variance for a given edge, because the win/loss outcomes are more lopsided.

The practical consequence is that variance, not lack of skill, is usually what wipes out undercapitalised bettors. A strategy can be correctly priced and still produce a losing sequence long enough to break a bankroll that wasn’t sized for it. This is why serious bettors think in terms of standard deviation and sample size, not just “am I up or down this week.”

Example

Say a bettor has built a model for ATP tennis matches and finds a player priced at 2.50 by the market who they believe wins 45% of the time. The market’s implied probability at 2.50 is 40%, so the extra 5 percentage points is the edge. Staking €20 flat per bet:

• A win returns profit of €20 × (2.50 − 1) = €30
• A loss costs the full €20 stake
• Expected value per bet = (0.45 × €30) − (0.55 × €20) = €13.50 − €11.00 = €2.50

Over 20 bets, expected profit is 20 × €2.50 = €50, built on an expected 9 wins out of 20 (20 × 0.45).

Now look at two entirely plausible 20-bet runs with that same 45% true win rate:

Run A — 9 wins, 11 losses (matches expectation almost exactly): profit = (9 × €30) − (11 × €20) = €270 − €220 = €50. The bettor looks sharp and the numbers agree with the model.

Run B — 5 wins, 15 losses: profit = (5 × €30) − (15 × €20) = €150 − €300 = −€150. Same 45% edge, same €2.50-per-bet EV, but the bettor is down €150 and questioning the whole model.

Is Run B unusual enough to mean the model is wrong? Not really. With p = 0.45 and n = 20, the standard deviation of the win count is √(20 × 0.45 × 0.55) ≈ 2.2 bets. Five wins is under two standard deviations below the expected nine — an uncomfortable but statistically ordinary result, not proof the edge doesn’t exist. This is exactly the kind of stretch that convinces good bettors to quit good strategies, and bad bettors to keep funding bad ones.

Key Points

• Variance is the price of having an edge at all: you cannot access positive EV without accepting that any given run of bets can land well below (or above) it. Wanting the profit without the swings isn’t an option in this game.
• Small samples are close to meaningless: 20, 50, even 100 bets is often too few to separate a real edge from noise, especially at odds above 2.50. Judge a strategy on hundreds of settled bets, not a hot or cold fortnight.
• Higher odds mean higher variance for the same edge: a value bet at 5.00 will produce far bigger bankroll swings than an equally-edged bet at 1.80. Factor this into how much of your bankroll any single wager should risk.
• Staking size controls how much variance you feel: flat staking, fractional Kelly, or fixed percentage staking all change the amplitude of your swings without changing your underlying edge. Undersized bankrolls turn ordinary variance into ruin.
• A losing streak is not automatically a broken model: before scrapping a strategy, check whether the drawdown is inside the range normal variance would produce, using something like the standard deviation of expected wins, rather than reacting to the scoreboard alone.
• Bookmakers manage variance too: it’s why they balance books, limit sharp accounts, and price in margin — they’re trying to shrink their own exposure to the same swings a bettor experiences, just from the other side of the book.