OutlierBaseline Projection Lab

Project the next season—not the last one again.

Compare current and historical players across ten seasons of NFL production. Recent form, position-specific aging, and expected games shape the next season's projections.

Current model Age Curve v1 Explainable · format-aware · season-level

Model inputs

Choose the league and player

Teams Quarterbacks Reception scoring TE premium Position Player

Learning from ten seasons

Position age curve

RB production by age

Preparing the selected player's comparison...

Player-season Position average Selected player

Age comparison

Scroll inside the chart to explore every age. Hover over a point for its season details.

Position distribution

Expected scoring and standard deviation

Current draftable-player projections under this scoring format. The spread shows how much season scoring varies inside each position.

Draft capital map

Expected season points by position and round

Compare this season's projections with actual season scoring grouped by historical draft position.

Season view Compare waiting

Graph position

● QB ● RB ● WR ● TE · Choose one position to show average ± one scoring SD.

Scroll inside the graph or the table below to compare all rounds and positions.

Lines show scoring averages, not a promise of availability or the best roster strategy. Gaps mean missing data. Historical variation is not a calibrated forecast interval.

Where does waiting cost the most?

How to use the timing guide

Compare a position in this round with the same position one or two rounds later. A target window means one of the largest qualifying scoring drops from waiting—not the highest raw fantasy score. Positions within five season points of the largest drop share the timing priority.

Historical comparisons use only years represented in both rounds and give each matching year equal weight. A signal requires at least five shared seasons, ten player-seasons and five distinct players on each side, a mean drop of at least ten season points, and a decline in at least two-thirds of the shared years. These are transparent screening rules, not a statistical confidence test. Single-year views show differences but cannot establish a multi-year signal.

Current-season timing uses the model's projected points; one player on each side is sufficient for a comparison, so it is labeled projection-only. Missing rounds stay unavailable. No declining curve is imposed: later-round averages can be higher.

This assumes you still need the positions being compared. It does not know your roster, draft slot, keeper rules, or which players will actually remain. An ADP round is an average, not a promise of availability. A quiet window is not proof a position is bad or safe to ignore; bench upside and individual bargains still matter. Historical ADP format limitations apply to every historical timing label. These labels do not change player rankings or the separate Market +/- draft tags.

Historical sources: MyFantasyLeague ADP API · nflverse regular-season statistics.

How Age Curve v1 works

A transparent baseline we can improve

Recent form

The latest three seasons receive 5:3:2 weights, with partial seasons contributing less.

Position aging

Year-over-year changes from the same player and position estimate improvement or decline at the next age.

Expected games

Recent availability is blended with a healthy-season baseline instead of assuming everyone plays all 17 games.

Uncertainty

Player history and the same-position age group create the displayed one-standard-deviation range.

Historical player records include recently retired players. The last recorded season describes the available statistics.

Historical statistics and player identities: nflverse.

This first version intentionally does not guess coaching changes, depth-chart usage, offensive efficiency, or strength of schedule. Those become separate, visible model inputs rather than hidden adjustments.