PUBLIC PROOF · HISTORICAL RESULTS
See the evidence behind the board.
DFSReady is built to reduce the daily decision load. These replays show what the projection and decision layers actually did across historical slates—before we ask you to trust the product.
EVIDENCE INDEX
Start with the result that matches your question.
| Replay | What it answers | Coverage | Open |
|---|---|---|---|
| 2025–26 projection holdout | Does the projection rank and estimate actual fantasy production? | 24,208 played rows · 164 slates | View evidence → |
| 2024–25 salary-aware diagnostic | Does a legal lineup surface behave differently from raw projections? | DraftKings salary archive · 112–114 slates | View diagnostic → |
The links show the complete decision context and the limitations attached to each replay. There is no hidden “best season” filter.
BASELINE HOLDOUT · 2025–26
The projection clears a simple average-player baseline.
| Surface | MAE | RMSE | Correlation | Top 3 actual / slate |
|---|---|---|---|---|
| Training-mean baseline | 12.2188 | 15.1284 | — | — |
| DBB2 only | 8.6175 | 11.0039 | 0.6916 | 53.3417 |
| DBB2 + style/archetype fit | 8.6168 | 11.0029 | 0.6917 | 53.3498 |
The adjustment is directionally positive but practically tiny. It remains a bounded context/tie-breaking surface, not a headline edge.
FEED COMPARISON · EXACT INTERSECTION
New DailyEngine feed versus the legacy replay.
Both models were evaluated on the same 24,208 joined player-games, with 24,152 rows where both sources marked the player as appearing. The target was recomputed from the authoritative historical box score.
| Model | MAE | RMSE | Correlation | Top 3 actual / slate |
|---|---|---|---|---|
| Legacy DFSReady | 8.6134 | 11.0028 | 0.6906 | 53.5008 |
| DailyEngine | 8.0747 | 10.5602 | 0.7417 | 53.9921 |
This comparison supports the feed change. It does not establish salary value, ownership leverage, lineup ROI, or future-season performance.