Gambling Mathematics Academy
Variance and Sample Size: How Many Bets Until You Know?
Why thousands of bets - not dozens - are required before short-run results say anything about skill.
Direct Answer
Distinguishing skill from luck in betting requires a large sample. At typical edges (1-3%), thousands of wagers are needed before observed win rate is a reliable estimate of true win rate. Below that, results are mostly noise.
Key Takeaways
- 01Standard deviation of a bet portfolio shrinks with the square root of n.
- 02Small edges require very large samples to detect.
- 03A 53% bettor at -110 can lose money over hundreds of bets.
- 04Closing line value is a faster proxy than win-loss results.
Why noise dominates short samples
With each independent wager, results disperse around the expected value. The standard deviation of cumulative results grows with the square root of the number of bets, while expected value grows linearly. In short samples, the noise term dwarfs the signal.
Worked example
A bettor with a true win rate of 53% at -110 odds has an edge of about 1.2%. Over 500 wagers, their 95% confidence interval still contains both losing and winning outcomes. Over 5,000 wagers, the interval narrows to a confident positive.
What this means in practice
Do not draw conclusions from a 50-bet sample. Do not abandon a process after a 100-bet drawdown. Do not declare yourself sharp after a 30-bet hot streak. Skill detection requires samples in the thousands.
Frequently asked questions
How big a bankroll do I need to survive variance?+
Standard guidance places bankroll at 1-3% per wager for serious bettors, which corresponds to roughly 100-200 units of working capital.
This article is educational only. It is not wagering, financial, or legal advice. See our editorial policy.