How the WR Model Works

What you're looking at

Each week our model projects a receiving-yards and receptions line for every team's WR1 and WR2 (plus other receivers who play at least 40% of snaps). "Our Line" is the model's median projection rounded to the nearest half point; "Range" is its 25th–75th percentile band.

The model

A LightGBM quantile ensemble trained on 2021-present player-game data: target and air-yards share, snap rates, depth-chart role, injuries, team pass-rate-over-expected and pace, opponent pass-defense EPA and pressure, Vegas totals/spreads, and Next Gen Stats separation metrics. The five quantile predictions form a full probability curve, so any sportsbook line can be priced as a calibrated P(over).

Edges and flags

"Book" is the sportsbook line we evaluated (Hard Rock when it posts one). "P(Over)" is the model's calibrated probability the player clears that line. "Edge" compares that to the de-vigged sharp-market consensus (Pinnacle preferred). The ⚑ flag marks plays where the edge is at least 4 points AND the sharp market leans the same way against Hard Rock's price.

Grading

After games finish, every prediction is graded against the evaluated line, and flagged edges get closing-line-value (CLV) tracking — how much the market moved toward or away from our number by kickoff, in probability terms. The performance strip shows the last four graded weeks.

Receiving Yards Receptions WR1/WR2 only
Player Team Opp Our Line Range (q25–q75) Book Book Line P(Over) Edge
Jaxon Smith-Njigba WR1 SEA NE 5.0 3.4–7 draftkings 6.5 42.1% -5.3 pts
A.J. Brown WR1 PHI WAS 5.0 3.2–6.6 draftkings 4.5 51.4% -1.4 pts
Puka Nacua WR1 LA SF 5.5 4–6.9 draftkings 7.5 42.1% -1.2 pts
Keenan Allen WR1 LAC ARI 2.5 2–3.5
Mack Hollins WR1 NE SEA 2.5 1.8–4.3
Courtland Sutton WR1 DEN KC 4.0 2.5–5.4
DK Metcalf WR1 PIT ATL 3.5 2.4–5.3
Terry McLaurin WR1 WAS PHI 3.5 2.3–5
Brandon Aiyuk WR1 SF LA 2.5 2–3.8
Justin Jefferson WR1 MIN GB 4.0 2.6–6.4
Jerry Jeudy WR1 CLE JAX 3.0 1.5–4.5
Nico Collins WR1 HOU BUF 3.5 2.3–4.9
Jaylen Waddle WR1 MIA LV 3.0 2.1–4.8
Ja'Marr Chase WR1 CIN TB 5.0 3.7–7.6
Amon-Ra St. Brown WR1 DET NO 5.5 3.5–6.9
Drake London WR1 ATL PIT 4.0 2.8–6.3
Chris Olave WR1 NO DET 5.0 3.3–6.3
George Pickens WR1 DAL NYG 3.5 2.5–5.7
Khalil Shakir WR1 BUF HOU 3.5 2.3–5.5
Garrett Wilson WR1 NYJ TEN 3.5 2.6–5.9
Wan'Dale Robinson WR1 NYG DAL 4.5 3.6–6.3
Christian Watson WR1 GB MIN 3.0 2.3–5.1
Michael Wilson WR1 ARI LAC 5.0 3.3–6.2
Tre Tucker WR1 LV MIA 3.5 1.5–4.6
Parker Washington WR1 JAX CLE 3.5 2.4–5.1
Josh Downs WR1 IND BAL 2.0 1.3–3.4
Zay Flowers WR1 BAL IND 4.0 2.7–6.3
Rashee Rice WR1 KC DEN 5.0 3.5–6.3
Rome Odunze WR1 CHI CAR 2.5 1.4–4
Tetairoa McMillan WR1 CAR CHI 3.5 2.4–4.6
Elic Ayomanor WR1 TEN NYJ 2.0 0.9–3.4
Romeo Doubs WR2 GB MIN 2.0 1.7–3.7 draftkings 2.5 49.7% -6.2 pts
Deebo Samuel Sr. WR2 WAS PHI 3.0 2–4.3 draftkings 2.5 52.2% -4.4 pts
Mike Evans WR2 TB CIN 3.5 2.3–5.1 draftkings 3.5 49.7% -4.4 pts
Cooper Kupp WR2 SEA NE 2.5 1.8–3.5 draftkings 2.5 50% -1.3 pts
DeAndre Hopkins WR2 BAL IND 1.0 0.4–2
Davante Adams WR2 LA SF 3.0 2–4
Amari Cooper WR2 BUF HOU 1.0 0.4–2.8
Stefon Diggs WR2 NE SEA 3.0 1.9–4
Tyreek Hill WR2 MIA LV 3.5 2.5–5.8
Marquez Valdes-Scantling WR2 PIT ATL 1.0 0.6–2.8
Jakobi Meyers WR2 JAX CLE 3.0 2–3.7
Michael Pittman WR2 IND BAL 2.0 1.5–3.6
Jauan Jennings WR2 SF LA 2.5 1.3–3.4
Darnell Mooney WR2 ATL PIT 1.5 0.8–3.5
CeeDee Lamb WR2 DAL NYG 4.0 2.5–5.3
Tee Higgins WR2 CIN TB 3.0 1.7–5
DeVonta Smith WR2 PHI WAS 4.0 2.4–5.7
Jameson Williams WR2 DET NO 3.5 2.3–5.6
Quentin Johnston WR2 LAC ARI 2.5 1.6–3.5
Xavier Hutchinson WR2 HOU BUF 2.0 1.1–3.3
Marvin Mims Jr. WR2 DEN KC 2.0 0.8–3.1
Jordan Addison WR2 MIN GB 1.5 0.9–3.7
Malik Nabers WR2 NYG DAL 4.5 3.3–6.9
Devaughn Vele WR2 NO DET 1.5 1.2–3.2
Jalen Coker WR2 CAR CHI 3.5 2.2–4.7
Marvin Harrison Jr. WR2 ARI LAC 2.5 1.2–3.5
Adonai Mitchell WR2 NYJ TEN 2.5 1.2–4.5
Xavier Worthy WR2 KC DEN 1.5 1–3.4
Chimere Dike WR2 TEN NYJ 2.0 1.2–2.9
Luther Burden III WR2 CHI CAR 3.5 1.7–4.1
Methodology. A LightGBM quantile model (q10–q90) trained on nflverse player-game data since 2021 projects each receiver's yardage and receptions distribution from role (target/air-yards share, snaps, depth chart, injuries), team tendencies (pass rate over expected, pace), opponent pass defense (EPA, pressure), and game environment (Vegas totals, spread, venue). Quantiles become a full probability curve, calibrated against three seasons of real closing lines, so P(over) is priced for the actual book line. Edges compare our probability to the de-vigged sharp market; validation is strictly walk-forward (the model never sees the week it predicts). Predictions freeze when published and grade automatically after games — history and CLV are tracked on every flagged edge.