How Many Bets Before a Result Means Something? The Luck Band
Ten bets, eight winners: a system? The luck band tells you how much of a result chance alone explains, with a table for every sample size and four real strategies judged by it.
Last reviewed November 3, 2026
Key takeaways
- >Luck band ≈ ±2 × the spread of single-bet returns ÷ √(number of bets): a no-edge strategy lands inside it 19 times out of 20
- >At odds of 2.00, luck alone swings the ROI by about ±37% after 30 bets and ±12% after 300
- >Longer odds need many more bets: a 3.00 strategy needs about four times as many as a 1.50 strategy
- >Premier League totals under 1.30 won 86% of 238 bets and still lost 5.5%, beyond luck: a high hit rate proves nothing
- >Under 30 settled bets we give no verdict at all
Ten bets, eight winners, +40%: a system? Almost certainly not. The question every bettor eventually asks is how many bets it takes before a result stops being luck. There is a simple way to answer it, and it is the same calculation we put next to every result on this site.
The luck band
Every bet's return is noisy: a winner at 2.00 returns +100% of the stake, a loser −100%. Average enough of those returns and the noise shrinks, but slowly, with the square root of the number of bets. The luck band is the range that chance alone produces:
Luck band ≈ ± 2 × (spread of single-bet returns) ÷ √(number of bets)
The "2 ×" makes it a 95% band: a strategy with no edge at all lands inside it about 19 times out of 20. A result inside the band is not evidence of anything. A result outside it is worth a closer look.
How wide it is
For a strategy with no edge (prices exactly fair), the band is:
Bets | At odds 1.50 | At odds 2.00 | At odds 3.00 |
|---|---|---|---|
10 | ±45% | ±63% | ±89% |
30 | ±26% | ±37% | ±52% |
50 | ±20% | ±28% | ±40% |
100 | ±14% | ±20% | ±28% |
300 | ±8% | ±12% | ±16% |
1,000 | ±4% | ±6% | ±9% |
Two things to take from this table:
- It takes hundreds of bets. At typical odds of 2.00, you need about 300 bets before chance alone stays within ±12%, and 1,000 before it gets down to ±6%. A typical betting edge, if one exists at all, is a few percent.
- Longer odds need more bets. Outsiders win rarely and pay a lot, so their results swing harder. Because the band shrinks with the square root of the number of bets, a 3.00 strategy needs about four times as many bets as a 1.50 strategy for the same certainty.
Four real examples
All from our own paper strategies and backtests, with the band computed from the actual bets:
Strategy | Bets | ROI | Luck band | Verdict |
|---|---|---|---|---|
Our AI "quant" strategy (an expected-goals model), this season | 6 settled | +40.2% | ±57% | Too few to judge |
Home or away favourites under 1.80, all leagues | 264 | −0.6% | ±8.5% | Within luck |
Premier League over/under lines priced under 1.30 | 238 | −5.5% | ±5.0% | Losing beyond luck |
Eredivisie overs 1.50–2.50 (backtest) | 122 | +18.5% | ±17.5% | Just outside the band |
- The quant strategy won 5 of its first 6 bets this season. Exciting, and statistically meaningless: at 6 bets, anything within about ±57% of zero is what luck looks like.
- The favourites strategy placed 264 bets and finished almost exactly level. That is not "slightly unlucky"; it is the margin doing its work on a strategy with no edge.
- The Premier League short-priced totals hit 205 of 238 bets, an 86% hit rate, and still lost €131.50, beyond what luck explains. At an average price of 1.10, winning 86% is not enough: you need to win about 91% just to break even (1 ÷ 1.10). A high hit rate says nothing on its own.
- The Eredivisie overs cleared the band, narrowly. Why that still is not proof is the subject of The hot-start trap and Overfitting.
The rules we apply
- Under 30 settled bets: no verdict at all. The band is too wide to say anything.
- Inside the band: "within luck". Not an edge, and also not proof there is none.
- Outside the band, but only in one half of the period: "one good stretch", not yet a pattern.
- Outside the band and positive in both halves: "worth a closer look", and then the next question is how many rules were tried to find it.
Why this matters for your own betting
Most bettors judge a method after 20 or 30 bets. At those numbers, a method with no edge looks like a winner more than a third of the time, and a method with a small real edge looks like a loser almost as often. If you want to know whether something works, count bets, not weeks, and write the rule down before you start, so you cannot quietly change it halfway.
How the band is calculated: ±2 standard errors of the per-bet return (won: odds − 1, lost: −1), the same function Edge Terminal's Lab shows beside every result. Table: fair prices, no margin. Examples: Edge Terminal paper strategies and backtester, as of 3 November 2026. Past results only; nothing here is a tip.