in season · 2026 season

The This Week's Forward Test

The ten places where our engine most disagrees with DraftKings this week, frozen before kickoff and scored in public after the games. These are disagreements, not advice. We publish them because a model that hides from its own record is not a model you should trust.

Read this first. In our three season test against real closing lines, disagreements like these did not beat the market: the strongest subset returned -1.75%, +5.78% and +1.43% across 2023 to 2025. That is why nothing here is a recommendation. This page exists to measure, in public, whether that result holds in 2026.

A known blind spot

The engine projects each player from his last eight games of usage. A player whose role or team changed in the offseason will show up here with a large disagreement, because the engine is pricing the job he had and DraftKings is pricing the job he has now. Early in a season that is a limitation of the model, not a signal. The settled record will show which way it cuts.