Betting results are dominated by variance over any sample a person is likely to accumulate in a season. This is not a caveat to mention and move past; it is the central fact about interpreting any published record, including the one on this site.
The scale of the problem
Suppose a method genuinely wins 54 percent of bets at -110, which would be a very strong edge. Over 100 bets, an ordinary run of bad luck still produces a losing record often enough that a losing 100-bet sample is unremarkable. Distinguishing a 54 percent method from a 50 percent method with reasonable confidence takes samples in the thousands, not the dozens.
| Sample size | What it can distinguish |
|---|---|
| 30 bets | Essentially nothing. Any win rate from 30% to 70% is unsurprising. |
| 100 bets | Very little. A real edge and no edge produce overlapping results. |
| 1,000 bets | A large edge starts to separate from break-even. |
| 5,000+ bets | A modest edge becomes measurable with some confidence. |
Why unit totals mislead worse than win rates
A record can show a positive unit total on a poor win rate if a small number of longshots landed. That is a description of what happened, not evidence of a method. Concentration matters: if removing the three biggest winners flips the total negative, the total is a statement about those three results.
Note
This is why closing line value is the more useful early signal. It produces a reading on every single bet rather than only on the ones that settle favourably, so it accumulates evidence far faster than profit does.