Track record every call saved before kickoff, graded after the stats land, misses included
We don’t project fantasy points. We project winning situations.
Each week we grade whether the lower-sweat player at the same position actually outscored the other. A coin flip gets 50%; our long-term goal is to beat 65% (a goal, not a result yet). Why we built this →
How we grade
- Saved before kickoff. Each data build saves every player's usage score, rank, sweat meter (with its reasons), injury status, matchup, QB context and game lines. For each player we keep the last save before his own kickoff; once his game starts, his entry is frozen and never edited.
- Graded after the stats land. When nflverse posts the week's player stats (usually Tuesday morning), we score every saved call in PPR. A player with no stat line counts as 0 (did not play).
- Lower sweat should beat higher sweat. For every pair of players at the same position, we check whether the one with the lower sweat scored more. A coin flip gets 50%; the range shown is a 95% interval (pairs overlap, so treat it as approximate).
- Rank correlation (Spearman, -1 to 1) shows how well the usage score and the sweat meter ordered players against actual points. 0 = no better than random.
- Top-12 hit rate: how many of our usage top-12 at each position finished top-12. A random pick would hit about 12 / players in the pool.
- Busts = finished below the position's replacement line that week (QB12, RB24, WR24, TE12 among graded players).
- Kickers and team defenses are saved the same way and graded separately in default (Yahoo-style) points, never mixed into the QB/RB/WR/TE numbers. Same pair test, with a 25+ point gap as the "big gap". They started being saved in Week 6 of 2026.
- Games already final when a week was saved are left out. The record starts with Week 5 of 2026; earlier weeks were never published, so we don't grade them.
- Fantasy context only. No gambling ads, no sportsbook partners.