Why 20 Bets Cannot Validate a Strategy
A strong win rate in a short run can still be consistent with a wide range of underlying results. See an interval example and why odds matter.
A short run leaves wide uncertainty
Suppose a record contains 20 bets and 12 wins. The observed win rate is 60%. But 60% is a share in a small sample, not a reliable estimate of future results. Assuming independent observations and a simplified binomial model, the approximate 95% Wilson interval for 12 successes out of 20 is about 38.7%–78.1%. That is a wide range: the same run is compatible with very different long-run win rates.
An interval describes uncertainty about a parameter under specific assumptions. It does not say the true win rate must lie inside this exact interval, and it does not predict the next bet. Real wagers may be dependent, cover different sports and carry different prices. A simple binomial calculation may be unsuitable for such data.
Win rate does not replace ROI
Twelve wins from twenty do not tell you how much was staked, what the odds were, how much each winner returned or whether some bets were voided. Win rate ignores the price of a win and the payout size. On its own, it does not prove profitability or an edge. Keep win rate, ROI and variance distinct.
What to record for a fair evaluation
Track more than win or loss: date, original odds, market, stake, settlement status, void return, commission and why the bet was included. Do not delete losing periods or change criteria after seeing results. If events are related or one league dominates the sample, account for that when interpreting it.
Plan the sample before analysis instead of stopping when a run looks persuasive. Checking every few bets increases the risk of selecting a lucky slice. Separate development of an idea from an independent evaluation, then predefine the metric and decision rule.
For one independent sample, report the number of bets, wins, win rate and interval. If prices differ, add net financial return and ROI; if there are missing observations or voids, explain how they were handled. Sportexa's article on variance explains why results fluctuate.
Takeaway
Twenty bets can reveal a recordkeeping error or test whether a journal works. They are not enough to declare a strategy proven. More observations can reduce uncertainty, but do not remove it automatically: sound methods, consistent rules and a complete record still matter.
Show the method with the estimate
Report an interval beside the observed share, along with the method, confidence level and sample size. Do not compare two win rates by a few percentage points when samples are small and intervals are wide. The Wilson interval is often preferable to a simple symmetric estimate for a small proportion sample, but it still assumes independent binary observations with a common success probability. If bets are clustered by match, market or time period, or the strategy changed, the basic formula may understate uncertainty. Use a method that accounts for the structure or limit the conclusion to a description of the collected data. Explain how voids and missing observations were handled.