Variance and Sample Size: How Many Bets Until You Know Your Edge?
Short-run results are mostly noise. See with a worked example how variance behaves, why a real edge can look like a loss after 100 bets, and how many bets it takes before the numbers mean something.
You placed 100 bets, you are down 8 units, and you wonder if the strategy is broken. Or you are up 12 units and you wonder if you finally cracked it. Both feelings are usually noise. In betting, the result of a small sample says very little about the quality of your decisions, and knowing how little is one of the most useful skills you can build.
Why results lie in the short run
Every bet is a weighted coin flip. Even a bettor with a real edge loses a big share of the time, and the swings between wins and losses are far larger than the edge itself. That swing is called variance, and its size is measured by the standard deviation of your results.
Here is the key asymmetry. Your expected profit grows in a straight line with the number of bets: double the bets, double the expected profit. But the noise grows only with the square root of the number of bets: double the bets, noise grows about 41 percent. So the edge slowly wins over the noise, but it takes a lot of bets.
A worked example
Take a bettor with a genuine 5 percent edge, flat 1 unit stakes, all at decimal odds of 2.00. That means a 52.5 percent chance to win each bet. The standard deviation of a single bet is almost exactly 1 unit.
| Bets | Expected profit | Standard deviation | Typical range (about 95%) |
|---|---|---|---|
| 100 | +5 | 10 | -15 to +25 |
| 400 | +20 | 20 | -20 to +60 |
| 1,000 | +50 | 32 | -13 to +113 |
| 2,500 | +125 | 50 | +25 to +225 |
Read the first row slowly. After 100 bets, a bettor who truly has a 5 percent edge has roughly a 3 in 10 chance of being in the red. After 400 bets that drops to about 1 in 6. Only around 1,600 bets does the typical range stop including zero. A smaller edge needs far more: halve the edge and you need four times as many bets.
So how many bets until you know?
There is no magic number, but the logic is simple. The sample you need grows with the square of the noise divided by the edge. Higher odds mean more variance, so longshot bettors need bigger samples than people who bet near even money. A thin edge means a bigger sample too.
This is why judging a strategy on a hot month is a trap. A few dozen bets can make a break-even approach look brilliant, or a good one look broken. The honest answer is usually "hundreds at the very least, thousands to be sure."
Use a faster signal alongside results
If profit takes thousands of bets to speak, you need something that speaks sooner. Closing line value is that signal. It compares the price you took with the final market price, so it does not depend on whether the bet happened to win. Consistently beating the closing line across a few hundred bets is much stronger evidence than the same few hundred results in profit. It pairs naturally with value betting, where the whole point is taking prices better than fair.
Variance also tells you how to size your stakes. Large stakes on a small sample turn normal noise into a painful drawdown, which is why bankroll management exists. Plan for a losing streak you have not seen yet, because it is coming.
Seeing it in your own numbers
Doing this by hand means a spreadsheet, a closing odd for every bet and a lot of patience. Betamigos logs each bet, calculates CLV and P&L across all your books, and shows your running totals so you can tell a rough patch from a real problem. Open your dashboard and let the sample build, because the sample is the whole game.
Betamigos is an odds-analytics tool, not a sportsbook, and none of this is a promise of profit. Bet responsibly. 18+.