How Many Bets Before You Know You Are Winning?
A 60% record over 50 bets is noise. So is a 40% one. The number of bets it takes to tell a real edge from luck is far larger than almost anyone assumes — and it is the reason so much of what gets sold as handicapping skill is a coin that landed heads.
10 min read
The short answer
- Separating a 55% win rate from break-even at −110 takes about 5,689 bets at conventional statistical standards.
- At ten bets a week, that is nearly eleven years.
- A 55-45 record over 100 bets has a 95% confidence interval of roughly 45% to 64% — it includes losing badly.
- Even 550-450 over 1,000 bets does not conclusively clear the 52.38% break-even bar.
- The smaller the edge, the harder it gets: halving the edge roughly quadruples the bets required.
- This is why closing line value matters — it gives a readable signal in dozens of bets rather than thousands.
Why results are such a slow instrument
Betting produces a stream of binary outcomes with a tiny signal buried in them. A strong professional edge might be three percentage points of win rate. The random variation between one run of a hundred bets and the next is many times that. The signal is real; it is simply much smaller than the noise it arrives in.
Which means the natural way to evaluate a bettor — look at the record — is close to useless over any realistic period. Not imprecise. Genuinely uninformative.
How many bets it actually takes
The formal version of the question: how many bets before a claimed win rate can be distinguished from break-even, with 95% confidence and 80% power — the conventional thresholds.
| Claimed win rate | Edge over break-even | Bets required | At 10 bets/week |
|---|---|---|---|
| 53% | 0.62 pts | 102,110 | 196 years |
| 54% | 1.62 pts | 14,910 | 29 years |
| 55% | 2.62 pts | 5,689 | 11 years |
| 56% | 3.62 pts | 2,975 | 5.7 years |
| 57% | 4.62 pts | 1,823 | 3.5 years |
| 60% | 7.62 pts | 665 | 1.3 years |
Read the top row again. A 53% bettor — genuinely profitable, a real edge, someone who would make money for decades — cannot prove it from results inside a human lifetime.
The reason is in the denominator of the formula: the required sample scales with the square of the edge. Halve the edge and you need roughly four times as many bets. This is why the table explodes at the top and not merely rises.
n ≈ (z_α × √(2 p̄ q̄) + z_β × √(p₀q₀ + p₁q₁))² / (p₁ − p₀)² p₀ = break-even rate implied by the price p₁ = the win rate being claimed The (p₁ − p₀)² is the whole story.
Both z terms are normal quantiles: one for the confidence level, one for the power.
What your own record is telling you
Turn it around. Instead of asking how many bets you need, ask what the bets you already have actually establish. A confidence interval on the win rate answers that.
| Record | Win rate | 95% confidence interval | Clears 52.38%? |
|---|---|---|---|
| 55-45 | 55.0% | 45.2% – 64.4% | No — includes losing badly |
| 110-90 | 55.0% | 48.1% – 61.7% | No |
| 550-450 | 55.0% | 51.9% – 58.1% | No — only just fails |
| 1100-900 | 55.0% | 52.8% – 57.2% | Yes |
Every row shows the same 55% record. The first tells you almost nothing — the true rate could plausibly be 46%, which is a disaster. It takes two thousand bets at that rate before the interval clears break-even and the record becomes evidence rather than an anecdote.
The win rate confidence calculator does this for your own record, and the record calculator keeps the underlying tally straight.
Why hot streaks are guaranteed to exist
Variance does not merely blur the picture, it manufactures convincing false ones. Over enough bets, extreme runs are not unlikely — they are certain.
A bettor at exactly break-even will produce a five-loss streak about 74% of the time within a hundred bets, and a seven-loss streak about a quarter of the time. Those same dynamics run in the other direction, which is the part worth sitting with: winning streaks of the same length are equally routine, and equally meaningless.
The streak calculator gives the exact probability of a run of any length at any win rate.
What to do about it
The situation is genuinely awkward: the thing you most want to know is the thing your data cannot tell you for years. Three responses are useful.
Measure the price, not the result. Closing line value removes the coin flip entirely and compares your price against the market’s final one. It gives a readable signal in dozens of bets rather than thousands, which is why serious bettors track it in preference to their own record.
Stake as though you might be wrong. Everything in bankroll management follows from the sample-size problem. You cannot know your edge is real for years, so stake at a size that survives being mistaken about it.
Discount every record you are shown, including your own. A 100-bet sample is a story, not evidence. When somebody advertises a win rate, the only questions that matter are how many bets, at what prices, and whether the losers were published at the time.
Common questions
- How many bets do I need to know if I am a winning bettor?
- To distinguish a 55% win rate from the 52.38% break-even at −110, roughly 5,689 bets at 95% confidence and 80% power — about eleven years at ten bets a week. A smaller 53% edge needs over 100,000 bets, because the requirement scales with the square of the edge.
- Is a 60% record over 50 bets good?
- It is not evidence of anything. A 95% confidence interval on 30-20 runs from roughly 46% to 72%, which comfortably includes losing money. Records at that sample size are consistent with almost any true win rate.
- Why does a smaller edge need so many more bets?
- Because the required sample scales with the square of the edge. Halving the edge roughly quadruples the bets needed — which is why 55% needs about 5,689 bets and 53% needs over 100,000.
- What is a confidence interval on a win rate?
- The range of true win rates consistent with what you observed. A 55-45 record over 100 bets has a 95% interval of about 45% to 64% — so the record is compatible with being a strong winner or a heavy loser.
- If results take years, how do I evaluate a bettor sooner?
- Track closing line value instead. Comparing the price taken against the market’s final price removes the randomness of the outcome and gives a usable signal within dozens of bets rather than thousands.