Proving an edge means showing your profits are too big to be luck. A rough rule: you need n = 4 x Var / edge² bets for about 95% confidence. At -110 odds with a 4% edge, that works out to about 2,260 bets. Most bettors quit long before that.
Up 15 units after 200 bets and feeling like a genius? Slow down. You might be one. The numbers just can't tell yet.
Why your record lies to you early
Every bet has two parts. There's your edge, the small steady thing you're trying to measure. And there's noise, the big random swing of win or lose. On a single -110 bet you either make 0.91 units or lose 1 unit. Your edge might be 0.04 units. The noise is roughly 25 times bigger than the signal.
Noise shrinks with the square root of the number of bets. Edge grows with the number of bets. Edge wins in the end. The end is just a lot further off than most people think.
The formula
For a 95% confidence check (about two standard errors), you want:
edge x n ≥ 2 x SD x √n
Solve for n and you get:
n = 4 x Var / edge²
Where:
- edge = your expected return per unit staked (ROI). A 4% edge means you make 0.04 units per unit bet on average.
- Var = the variance of the profit on one 1-unit bet.
For a bet paying b units on a win, with win probability p:
- p = (1 + edge) / (1 + b)
- Var = p x (1 - p) x (1 + b)²
That's the whole thing.
The worked numbers
We double checked these with a short script.
| Odds | Edge (ROI) | Win rate needed | Variance per bet | Bets for ~95% confidence |
|---|---|---|---|---|
| -110 | 4% | 54.48% | 0.904 | about 2,260 |
| +300 | 4% | 26.00% | 3.078 | about 7,700 |
| -110 | 2% | 53.43% | 0.907 | about 9,070 |
Here's the first row, step by step:
- At -110, a win pays b = 100/110 = 0.909 units.
- p = 1.04 / 1.909 = 0.5448.
- Var = 0.5448 x 0.4552 x 1.909² ≈ 0.904.
- n = 4 x 0.904 / 0.04² = 3.616 / 0.0016 ≈ 2,260.
Two lessons in that table
Halving your edge quadruples the sample. Going from a 4% edge to a 2% edge at the same odds takes you from about 2,260 bets to about 9,070. Edge is squared in the denominator, so small edges take forever to confirm. And most real edges in sports betting are small.
Longshots need way more bets. A 4% edge on +300 plays needs about 7,700 bets, over three times as many as the same edge at -110. If you bet plus money, expect bigger swings, longer losing streaks, and a record that stays fuzzy a lot longer. That's not a reason to skip underdogs. It's a reason to size them carefully and be patient.
What this means for you
- Don't trust a hot 100 bets. At -110 with zero edge, going 58-42 over 100 bets is well within normal luck.
- Track everything. You can't measure an edge you didn't record. Odds, stake, closing line, result.
- Use closing line value (CLV) as an early signal. Results take thousands of bets to mean much. Beating the closing number consistently tends to show up much sooner, and it is the metric sharp bettors watch while the results catch up.
- Size bets for the noise, not the dream. Fractional Kelly sizing helps you survive the long stretch where you can't tell skill from luck yet.
Caveats
This is a rough estimate. It assumes your bets are independent, your edge is constant, and your stakes are flat. Real betting is messier. Edges drift, markets get sharper, and correlated bets (same game parlays, multiple legs on one team) make the noise bigger. Treat these numbers as a minimum.
Also, 95% confidence still means a 1 in 20 chance of fooling yourself. If you test lots of strategies, one of them will look great by pure luck.
Try it yourself
EdgeTilt has free calculators so you can run this on your own bets: a Kelly bet sizer, vig removal, an odds converter, an arbitrage check, and a CLV tracker. Plug in your real numbers and see where you stand. Try the free tools at edgetilt.com.
FAQ
How many bets do I need to prove a 4% edge at -110? About 2,260 bets for roughly 95% confidence, using n = 4 x Var / edge².
Why do plus-money bets need a bigger sample? Their variance per bet is higher. At +300 the variance is about 3.4 times the variance at -110, so you need about 3.4 times as many bets for the same edge.
Can I use fewer bets if I track CLV? CLV is not proof of profit, but it is a faster and widely used signal of whether you are getting good numbers. Many bettors use it to judge a strategy before results settle.
Is 100 bets enough to know anything? Not about your edge. It is enough to learn whether your process and tracking work.
Does this apply to poker or other games? The same idea applies anywhere results are noisy. Poker players run into it as variance in bb/100.
This is educational content, not betting advice. 21+. Gamble responsibly. If you or someone you know has a gambling problem, call 1-800-GAMBLER.