We publish our misses.
Every Tuesday, once the NFL's official stats land, we score last week's served projections against what actually happened and post it here — no cherry-picking, no re-runs. Lower is better. MAE is the average miss in PPR points per player; rank correlation says whether we ordered players correctly (1.0 = perfect).
This season, week by week
| Week | Pos | Players | MAE | RMSE | Rank corr |
|---|---|---|---|---|---|
| Loading… | |||||
Backtest
The served model (the Herd blend): its members trained on 2022–24 only, scored on every week of the 2025 season. The blend weights for each half of 2025 were set on the other half, so no week is scored by weights that saw it. Skill positions count players it projected 3+ points, the same rule as the weekly numbers. This is the number we started from.
| Position | Player-weeks | MAE | RMSE | Rank corr |
|---|---|---|---|---|
| Loading… | ||||
Skill positions: players projected at 3+ PPR points. K and D/ST: every starter. Rank correlation is Spearman, averaged across weeks. Source: GET /v1/accuracy on the public API.