The backtest that looks incredible usually knew something it shouldn't have. A sports betting backtest only counts if it uses what was known, and could have been bet, when the decision is made. You'll learn the data contract that keeps final injury reports, revised stats and unavailable odds out of the past. Then you set up a walk-forward test with an untouched holdout and a metric chosen in advance, so you can't tweak your way to a pretty curve. The lesson also explains why a 12 percent return on 100 plays says very little. A good test is meant to disappoint you early, before real money is on the line.
Learning objectives
- Explain look-ahead leakage and why a backtest may use only what was known at decision time.
- Set up a walk-forward test with an untouched holdout and a metric chosen in advance.
- Recognize why a 12% sample ROI on 100 plays is not evidence of a repeatable return.
