Weekly player stats
Historical weekly opportunity and production from the approved nflverse release.
- Coverage
- NFL regular seasons 2018-2025
- Freshness
- Through 2025 Week 18
PROJECTION ENGINE · HISTORICAL V0.1
How our fantasy football projections work.
One audited path from prior-week opportunity to a point estimate, uncertainty range, and ranked list.
Historical test only — not current 2026 lineup advice.6.3%lower error than baseline
13,711held-out player-weeks
6/6test seasons won
80.7%80% interval coverage
LIVE MODEL MAP · V0.1.0
Every active input, model, and output.
Follow a signal from source data to a ranked player list. Every active node links to its evidence.
- InputsWeekly player stats
Historical weekly opportunity and production from the approved nflverse release.
- Player opportunityRecent WR opportunity
Targets, catches, yards, air yards, target share, and production from prior weeks only.
- ModelsWeekly WR projection model
A regularized model estimates weekly PPR points with empirical uncertainty bands.
- ProjectionsHistorical PPR projections
Frozen weekly point estimates and intervals used for out-of-time evaluation.
- Rank policyPPR rank and tier policy
Expected points determine order; empirical interval separation creates tiers.
- OutputsHistorical WR rankings
An archived ranking snapshot for evaluation, not current lineup advice.
Recent WR opportunity
Targets, catches, yards, air yards, target share, and production from prior weeks only.
- Coverage
- WRs with at least two prior appearances
- Freshness
- Locked before each forecast week
Weekly WR projection model
A regularized model estimates weekly PPR points with empirical uncertainty bands.
6.3% betterMAE vs trailing baseline
- Coverage
- Historical WR PPR forecasts, 2020-2025
- Freshness
- Refit before each held-out season
Historical PPR projections
Frozen weekly point estimates and intervals used for out-of-time evaluation.
- Coverage
- 13,711 held-out player-weeks
- Freshness
- Archived 2025 Week 18
PPR rank and tier policy
Expected points determine order; empirical interval separation creates tiers.
- Coverage
- Historical wide receivers in PPR
- Freshness
- Same as each projection snapshot
Historical WR rankings
An archived ranking snapshot for evaluation, not current lineup advice.
- Coverage
- 167 eligible wide receivers
- Freshness
- Archived 2025 Week 18
OUT-OF-TIME EVALUATION
It beat the simple baseline every year.
Each test season was hidden while the model was fit on earlier seasons. The baseline used the same eligible player-weeks.
| Season | Rows | Model MAE | Baseline MAE | Improvement | 80% coverage | Weekly rank corr. |
|---|---|---|---|---|---|---|
| 2020 | 2,154 | 5.07 | 5.42 | +6.5% | 77.2% | 0.58 |
| 2021 | 2,345 | 4.72 | 5.03 | +6.0% | 80.9% | 0.60 |
| 2022 | 2,233 | 4.70 | 5.06 | +7.1% | 81.0% | 0.60 |
| 2023 | 2,317 | 4.53 | 4.83 | +6.1% | 82.3% | 0.62 |
| 2024 | 2,271 | 4.59 | 4.91 | +6.4% | 80.2% | 0.62 |
| 2025 | 2,391 | 4.31 | 4.58 | +5.9% | 82.6% | 0.61 |
Baseline: Trailing four-appearance mean PPR points on the identical player-week rows.
FROZEN TEST OUTPUT
What one historical ranking looked like.
2025 season before Week 18 (historical reconstruction). This table proves the projection-to-ranking path; it is not a live rankings product.
| Rank | Tier | Wide receiver | Team | Expected PPR | Historical range |
|---|---|---|---|---|---|
| 1 | 1 | Puka Nacua | LA | 17.8 | 11.0–26.5 |
| 2 | 1 | Jaxon Smith-Njigba | SEA | 17.2 | 10.4–25.9 |
| 3 | 1 | Chris Olave | NO | 16.9 | 10.1–25.6 |
| 4 | 1 | Wan'Dale Robinson | NYG | 16.8 | 10.0–25.5 |
| 5 | 1 | Ja'Marr Chase | CIN | 16.5 | 9.8–25.3 |
| 6 | 1 | Amon-Ra St. Brown | DET | 16.4 | 9.6–25.1 |
| 7 | 1 | Zay Flowers | BAL | 14.9 | 8.2–23.7 |
| 8 | 1 | Michael Wilson | ARI | 14.3 | 7.6–23.1 |
| 9 | 1 | Luther Burden III | CHI | 13.6 | 6.8–22.3 |
| 10 | 1 | A.J. Brown | PHI | 13.6 | 6.8–22.3 |
| 11 | 1 | Courtland Sutton | DEN | 13.5 | 6.7–22.2 |
| 12 | 1 | Mike Evans | TB | 13.2 | 6.4–21.9 |
| 13 | 1 | Parker Washington | JAX | 12.8 | 6.0–21.5 |
| 14 | 1 | Stefon Diggs | NE | 12.7 | 5.9–21.4 |
| 15 | 1 | Ricky Pearsall | SF | 12.3 | 5.5–21.0 |
| 16 | 1 | CeeDee Lamb | DAL | 12.3 | 5.5–21.0 |
| 17 | 1 | Chris Godwin Jr. | TB | 12.0 | 5.2–20.7 |
| 18 | 1 | Jakobi Meyers | JAX | 11.5 | 4.7–20.2 |
| 19 | 1 | Terry McLaurin | WAS | 11.3 | 4.5–20.0 |
| 20 | 1 | Jameson Williams | DET | 11.3 | 4.5–20.0 |
| 21 | 1 | Mack Hollins | NE | 11.3 | 4.5–20.0 |
| 22 | 1 | Justin Jefferson | MIN | 11.2 | 4.4–19.9 |
| 23 | 1 | Christian Watson | GB | 11.1 | 4.3–19.8 |
| 24 | 1 | George Pickens | DAL | 10.7 | 3.9–19.4 |
| 25 | 1 | Nico Collins | HOU | 10.6 | 3.9–19.4 |
| 26 | 1 | Tee Higgins | CIN | 10.5 | 3.7–19.2 |
| 27 | 1 | Quentin Johnston | LAC | 10.2 | 3.4–18.9 |
| 28 | 1 | Deebo Samuel Sr. | WAS | 10.2 | 3.4–18.9 |
| 29 | 1 | Chimere Dike | TEN | 9.9 | 3.1–18.6 |
| 30 | 1 | DK Metcalf | PIT | 9.8 | 3.0–18.5 |
| 31 | 1 | Adonai Mitchell | NYJ | 9.8 | 3.0–18.5 |
| 32 | 1 | Khalil Shakir | BUF | 9.4 | 2.7–18.2 |
| 33 | 1 | Rashee Rice | KC | 9.3 | 2.5–18.0 |
| 34 | 1 | Tetairoa McMillan | CAR | 9.2 | 2.4–17.9 |
| 35 | 1 | Brian Thomas Jr. | JAX | 9.2 | 2.4–17.9 |
| 36 | 1 | DeVonta Smith | PHI | 9.2 | 2.4–17.9 |
CURRENT BOUNDARY
What this model does not know yet.
- This is a historical baseline, not a current-season projection.
- The eligibility pool is reconstructed from prior appearances and can lag trades, signings, and injuries.
- The model does not use live injuries, depth charts, routes, coaching, betting markets, or news.
- Intervals are empirical training-residual bands and are evaluated for coverage, not player-specific variance.
Live injuries, depth charts, coaching, markets, and news become active only after each signal has rights, timing, and validation evidence. See the research standards →