Strategy finder

We test thousands of strategies on older seasons, then check each one on later seasons it never saw. A strategy that only got lucky usually falls apart in the check.

Markets

Filters per strategy

Discover on seasons through

Min. bets (discovery)

Show

Discovery seasons pick and rank the strategies (by the luck check: the chance a no-edge system does that well). Check seasons come after and were not used to pick anything. "Held up" means it also made money in the check. Even then, test so many ideas and some will hold up by chance, so treat these as leads, not sure things. Uses closing lines; 1 unit per bet. Data: nflverse, through .

How the finder avoids fooling you

Discovery and check

Every strategy is scored on older "discovery" seasons, then tested on later "check" seasons that played no part in picking it. "Held up" means it also made money in the check. Why this matters.

Why most fade

When thousands of rules are tested, some look great by chance alone. Typically about half of the top discovery results lose money in the check seasons, which is the honest picture of how rare real edges are.

Using the results

Treat a result that held up as a lead to study in the backtester: does it make sense, does it hold across most seasons, and how big were its losing stretches? Backtesting guide.