Sample size calculator
Whether your record proves an edge — and how many bets it would take to know.
You are up over forty bets. Are you good, or lucky? This is the calculation that answers it, and the answer is usually not yet — a modest real edge and a warm streak look identical for far longer than anyone expects.
Your record
Win rate
55.00%
ROI
+5.00%
Break-even needs
52.38%
True rate is somewhere in
39.6–70.4%
How many bets would settle it?
Edge over break-even
2.62 pts
ROI it implies
+5.00%
Bets to prove it
2,243
To show a 55.00% win rate is real rather than luck — 95% confidence, 80% power — you would need about 2,243 bets, which is 56× the 40 you have logged.
A normal-approximation interval on the win rate, and a one-sided test of a proportion against the break-even rate. Sample size scales with the inverse square of your edge, so halving the edge quadruples the bets needed to demonstrate it. None of this says you are not winning — it says how much evidence you have, which is a different and more useful question.
What it takes to prove an edge at −110
| Win rate | Edge | ROI | Bets to prove it |
|---|---|---|---|
| 54% | 1.62 pts | +3.09% | 5,875 |
| 55% | 2.62 pts | +5.00% | 2,243 |
| 56% | 3.62 pts | +6.91% | 1,173 |
| 57% | 4.62 pts | +8.82% | 719 |
| 58% | 5.62 pts | +10.73% | 485 |
| 60% | 7.62 pts | +14.55% | 263 |
| 62% | 9.62 pts | +18.36% | 164 |
95% confidence, 80% power, against the 52.38% that −110 requires.
Why the numbers are so large
Sample size scales with the inverse square of your edge. A 60% bettor is beating break-even by 7.6 points and needs a few hundred bets. A 55% bettor is beating it by 2.6 — a third of the edge — and needs roughly nine times as many. Every step toward a realistic edge multiplies the evidence required.
That is not an argument against tracking. It is an argument against steering by win rate early, and for measuring something that resolves faster. Closing line value does resolve faster, because it scores every bet against the market’s final word instead of waiting for outcomes to pile up. A hundred bets of positive CLV is real evidence; a hundred bets of profit is barely any.
Related: break-even win rate · CLV calculator · bankroll calculator · when a 65% model still lost money
Common questions
- How many bets do you need to know if you're a winning bettor?
- More than almost anyone has. Demonstrating a 55% win rate at −110 — a genuinely strong edge worth 5% ROI — takes roughly 2,200 bets at 95% confidence. At 57% it falls to about 700, and at 60% to under 300. Below about 1,000 bets, a modest edge and a hot streak look identical.
- Is 100 bets enough to judge a betting record?
- No. Over 100 bets a 55% win rate carries a confidence interval of roughly 45% to 65%, which contains break-even, a coin flip, and a serious edge all at once. The record is compatible with almost any explanation, so it cannot single out the flattering one.
- Why does a small winning record prove nothing?
- Because variance at small samples is enormous relative to any real edge. The edge a good bettor has over break-even is two or three percentage points; the noise in a 40-bet sample is fifteen. The signal is real but it is far smaller than the thing it is buried in, and no amount of profit in the column changes that.
- How is sample size calculated for betting?
- A one-sided test of a proportion against the break-even rate the price implies: n = [z_a·√(p₀(1−p₀)) + z_b·√(p₁(1−p₁))]² ÷ (p₁ − p₀)². The important feature is the squared term in the denominator — sample size scales with the inverse square of your edge, so halving the edge quadruples the bets needed.
- Does this mean tracking bets is pointless?
- The opposite. It means win rate is a slow, noisy measure and you should not steer by it early. Closing line value tells you far more, far sooner, because it measures whether you beat the market on every bet rather than waiting for results to accumulate. Track both; trust CLV first.
- What if I'm losing — does the same logic apply?
- Exactly the same, in both directions. A losing run over a few hundred bets does not establish that you have no edge any more than a winning one establishes that you do. The calculator flags both cases, and at realistic sample sizes it usually flags neither.
More free tools
Track every bet automatically
Stop doing the math by hand. Bankroll Guardian logs your bets and shows P&L, ROI, CLV and more — free to start, no card required.