dynamicsystemsarchitecture.org

NFL Strategy Backtest — Methodology and Findings

The interactive version renders via JavaScript. This page holds the same real methodology and results as static text, readable without executing anything.

PurposeBacktest eight classic, single-variable against-the-spread strategies against the real vig-adjusted breakeven line.
StatusVerified — logic checked independently in Node against a separate Python implementation
Built fromnflverse (nflverse-data)
Depends on
Superseded by
Evidence7,276 real games, 1999–2025.

The real breakeven line

52.38% — the actual win rate required to overcome standard -110 odds. Every strategy below is measured against this, not 50%.

The eight strategies tested, and the real result for each

StrategyRecordATS %Beats breakeven?
Always Home3,473–3,60849.0%No
Always Away3,608–3,47351.0%No
Always Favorite3,615–3,43551.3%No
Always Underdog3,435–3,61548.7%No
Home Underdogs Only2,223–2,38148.3%No
Road Favorites Only2,381–2,22351.7%No — closest, still short
Always Over3,545–3,62549.4%No
Always Under3,625–3,54550.6%No

The actual finding

Not one of the eight simple, publicly-known, single-variable strategies beats the real breakeven line over 27 seasons of real games. That's the expected, credible result for an efficient market, not a disappointing one — it's evidence the closing line is doing its job. This dashboard exists specifically to show that naive rules don't have a real edge, as the honest baseline the multi-variable Strategy Composer is measured against.

→ Use the real, interactive version