A bet with value can still lose, and a run of them can lose in a row. That is not a sign that the method is broken: it is variance, the gap between what should happen on average and what actually happens in any given stretch of bets. Understanding it is the difference between sticking with a profitable approach and abandoning it at the worst moment.
Take odds of 2.10 on an outcome that really happens 50% of the time. Each 1 staked is worth +0.05 on average, a 5% edge. But no single bet returns 5%: it either wins 1.10 or loses 1. The edge only shows up as a long-run average, and the bets in between are dominated by luck. Football makes this worse than most sports because a single goal, often a deflection or a penalty, decides many results.
The lower your win rate, the longer your losing streaks. A bettor winning half of their bets will typically hit a run of six losses within 100 bets. At a 30% win rate, which is common when backing odds around 3.30, the typical worst run in 100 bets is ten losses in a row, and one of eight or more is very likely (86%). Value betting on longer odds means accepting this up front.
Typical worst losing streak by win rate (median of the longest run, with the chance of a run of 8 or more in 100 bets)
| Win rate | Worst streak in 100 bets | Worst streak in 500 bets | Chance of 8+ losses in a row in 100 bets |
|---|---|---|---|
| 50% | 6 | 8 | 17% |
| 40% | 7 | 11 | 49% |
| 30% | 10 | 15 | 86% |
| 20% | 15 | 22 | 99% |
Using the same example (odds of 2.10, true probability 50%), the table shows the range your return per unit staked will fall in about 95% of the time. After 100 bets the result could land anywhere between a 16% loss and a 26% profit, with exactly the same edge. Only around 2,000 bets narrow the range enough to separate a genuine edge from luck. This is why ten or twenty bets say almost nothing about whether a method works.
Likely range of return on investment (about 95% of outcomes) for a 5% edge at odds of 2.10
| Number of bets | Expected return | Random noise (one standard deviation) | Likely range |
|---|---|---|---|
| 100 | +5% | ±10.5% | −16% / +26% |
| 500 | +5% | ±4.7% | −4% / +14% |
| 1000 | +5% | ±3.3% | −2% / +12% |
| 2000 | +5% | ±2.3% | 0% / +10% |
Variance cannot be removed, only absorbed. Size your stakes so that a streak like the ones above takes a small bite out of the bankroll, which is what flat staking and a fraction of the Kelly criterion are for; the bankroll management tutorial covers how. Do not raise stakes to win back losses, do not drop a method after a bad week, and judge it over hundreds of bets, never over a handful.
The Kelly criterion calculator in the tools section turns your estimated probability and the odds on offer into a recommended stake, so a bad run does not push you into betting more than the bankroll can take.
Work out how much to stake with our free Kelly criterion calculator.