A profitable record looks like proof. Often it is not, because small samples swing widely on luck alone.
The key measures
- Profit in units: the sum of wins and losses, measured in units.
- ROIi (or yield): profit divided by the total staked. Stake 200 units in total and make 8.8, ROI is 4.4%.
- Average odds: needed to judge the hit ratei.
- Number of bets: the sample size.
The example
200 bets, average odds 1.80, 58% hit rate. ROI is 4.4%. It looks good. But how much could be luck?
The error of ROI (one standard deviation) over n bets is about the standard deviation of one bet divided by √n. Here one bet has a standard deviation of about 0.89 units, so over 200 bets the error is 6.3 points of ROI.
An ROI of +4.4% over 200 bets cannot be told apart from zero. Such a record is encouraging, not conclusive.
What to do in practice
- Ask for hundreds or thousands of bets, with the odds.
- Look at CLVi, which needs less data to show something.
- Check that tips were logged before the match and that losses are not missing.
- Compare with the market: what counts is whether you beat the closing price.
This is how we publish our own record: the performance page, with results that update automatically and small samples labelled as such.