Most betting records prove nothing, and say otherwise
A 2% edge over 500 even-money bets swings between roughly -7% and +11% across a season. So a bettor with a real edge routinely shows a losing year, and a bettor with none routinely shows a great one. Every tracker prints the number and stops.
A real record, read honestly
220 settled bets, +1,300 profit, 5.91% return:
+5.91% flat-bet ROI, but the interval spans zero (-7.0% to +18.9%) — indistinguishable from break-even
The interval runs -7.0% to 18.9%. Nothing about that record distinguishes it from break-even, and no amount of presentation changes it. Below 30 settled bets nothing is characterised at all.
Slice it enough and something always looks good
Tag your bets and each tag becomes a hypothesis. Test enough hypotheses and one clears the bar by luck:
| Tags tested | Bar required | Chance one is luck |
|---|---|---|
| 1 | z 1.96 | 5% |
| 5 | z 2.58 | 23% |
| 10 | z 2.81 | 40% |
| 20 | z 3.02 | 64% |
At ten tags there is a 40% chance that something looks significant when nothing is. So the bar rises with the number of slices checked, and the report says how many cleared the ordinary bar and how many survived the correction. A bettor who invents labels until one looks profitable is running a search, not a test.
Why closing line value instead
Profit takes years to say anything. The closing line is the market's best public estimate, and beating it consistently converges far faster — which is why it is the signal the engine calibrates on rather than merely displays. It also refuses to overclaim: under 30 graded bets it declines to give a verdict at all.