Every factor. Every weight. On purpose.

This page renders directly from the model's config file. Nothing is hidden and nothing here is marketing copy. When we change a weight, this page changes with it. If you disagree with a number, you know exactly which number.

How a projection is built

  1. Base projection: volume times efficiency times team environment. Roughly three quarters of the final score. Targets come from team pass attempts times target share, with vacated targets redistributed when players move.
  2. Environment modifier: o-line quality, offense quality, game script, coaching pass lean. Bounded to 12%.
  3. Matchup modifier: schedule difficulty specific to the player's archetype plus weather exposure. Bounded to 4%. Fantasy playoff weeks count double.
  4. Situation modifier: human-entered flags for injuries, holdouts, committees, ascending roles. Plus the age curve.
  5. Games played: 17 times a durability factor built from three seasons of missed games, regressed halfway to the league mean.

Scoring rules

pass_yd
0.04
pass_td
4.0
interception
-2.0
rush_yd
0.1
rec_yd
0.1
td
6.0
xp
1.0
fg_short
3.0
fg_mid
4.0
fg_long
5.0
reception
1.0 / 0.5 / 0.0

Regression factors

How much of last season we throw away, per stat. High number means the stat does not repeat.

defense
60%
td rate
70%
durability
50%
team epa
35%
point differential
50%

The modifier stack

environment

oline
0.05
offense quality
0.04
positive script
0.03
coach pass lean
0.04
new coordinator noise
0.02

matchup

schedule
0.03
weather
0.02
playoff week weight
2.0

Situation flags (manual, human judgment)

injury status
x0.9
holdout
x0.85
suspension
x0.8
new team
x0.97
committee
x0.88
ascending
x1.06
descending
x0.94
contract year
x1.02

Age curves

RB declines from age 27, 3.0% per year
WR peak 24 to 29
TE peak 25 to 30
QB declines from age 36, 3.0% per year

Rookies

No NFL volume exists, so rookies start from the historical average rookie season by position and draft slot (2015 through 2024), get a 7% translation discount, and are blended 60/40 with market ADP. No rookie ranks more than 24 spots ahead of his ADP unless we manually confirm the opportunity. Landing spot does the rest of the differentiating.

Ranking and tiers

The overall board ranks by points above a replacement player at the position, so a 300-point QB does not bury a 250-point RB. Kicker and DST edges are damped because they do not repeat. Tier breaks appear where the gap between consecutive players is at least 0.6 standard deviations of the local gaps. Value and fade flags fire when the model and ADP disagree by more than 12 spots, scaled up as ADP gets later.

Built on nflverse public data. Defensive metrics regressed 60% because defense does not repeat year over year. TD rates regressed 70% because touchdowns are noise wearing a jersey.