Model record
Every projection is saved shortly before kickoff and compared with the result once the game's stats are in.
Includes a replay of 10 past games (360 projections) through the current model, using only what was known before each game. A replay is a fair test of the statistics, matchup and injury numbers, but it is kinder than real life in two ways: league-wide measurements include later seasons, and injury reports are the final ones. More games are replayed every hour.
Yards
Avg miss: how far off, on average. Plain avg: his simple 16-game average with no factors. Bias: + means the model runs high.
| Model | Bet | Graded | Avg miss | Plain avg miss | Closer than plain | Bias | |
|---|---|---|---|---|---|---|---|
| v0.12 | Passing yards | 16 | 72.3 | 60.3 | 44% | −17.0 | Plain average closer |
| v0.12 | Receiving yards | 101 | 22.0 | 21.5 | 43% | −2.7 | Plain average closer |
| v0.12 | Rushing yards | 45 | 19.4 | 18.9 | 40% | +0.8 | Plain average closer |
Touchdowns
If the model says 60% and about 60% score, it is well calibrated.
| Model | Bet | Graded | Projected at least one | Actually did | Avg miss | Plain avg miss |
|---|---|---|---|---|---|---|
| v0.12 | Passing touchdowns | 16 | 77% | 81% | 1.10 | 1.02 |
| v0.12 | Touchdowns scored | 117 | 29% | 27% | 0.40 | 0.41 |
Does each factor help?
The average miss with the factor removed, compared with the model as it is, in yards.
| Model | Bet | Matchup | Coach / team | Rest / bye | Game time |
|---|---|---|---|---|---|
| v0.12 | Passing yards | Hurts by 6.3 | Hurts by 1.9 | Hurts by 0.1 | Hurts by 0.8 |
| v0.12 | Receiving yards | Hurts by 0.3 | Hurts by 0.1 | No difference | Helps by 0.2 |
| v0.12 | Rushing yards | Hurts by 0.6 | Hurts by 0.6 | Helps by 0.1 | Hurts by 0.6 |
Latest results
| Player | Team | Game | Bet | Projected | Actual |
|---|---|---|---|---|---|
| Jameis Winston | NYG | 2026 wk 10 | Passing yards | 172.2 | 0 |
| C.J. Stroud | HOU | 2026 wk 9 | Passing yards | 214.6 | 384 |
| Jordan Love | GB | 2026 wk 9 | Passing yards | 234.6 | 372 |
| CeeDee Lamb | DAL | 2026 wk 9 | Receiving yards | 62.3 | 195 |
| Dak Prescott | DAL | 2026 wk 9 | Passing yards | 210.5 | 335 |
| Jared Goff | DET | 2026 wk 9 | Passing yards | 285.8 | 170 |
| Kayshon Boutte | HOU | 2026 wk 9 | Receiving yards | 21.8 | 125 |
| Malik Willis | MIA | 2026 wk 9 | Passing yards | 216.7 | 128 |
| Christian Watson | GB | 2026 wk 9 | Receiving yards | 84.6 | 171 |
| Matthew Golden | GB | 2026 wk 9 | Receiving yards | 79.8 | 164 |
| Ja'Marr Chase | CIN | 2026 wk 9 | Receiving yards | 76.8 | 155 |
| Tee Higgins | CIN | 2026 wk 9 | Receiving yards | 102.2 | 32 |
Proposed weights: Yards bets
162 of 200 graded projections needed before weights are proposed.
| Factor | Weight now | Proposed | Games with it | |
|---|---|---|---|---|
| Matchup | 100% | 162 | Waiting for data | |
| Weather | 100% | 0 | Waiting for data | |
| Coach / team | 100% | 162 | Waiting for data | |
| Rest / bye | 100% | 16 | Waiting for data | |
| Game time | 100% | 148 | Waiting for data | |
| Injuries / lineup | 100% | 162 | Waiting for data |
Proposed weights: Touchdown bets
133 of 200 graded projections needed before weights are proposed.
| Factor | Weight now | Proposed | Games with it | |
|---|---|---|---|---|
| Matchup | 100% | 117 | Waiting for data | |
| Weather | 100% | 0 | Waiting for data | |
| Coach / team | 100% | 120 | Waiting for data | |
| Rest / bye | 100% | 14 | Waiting for data | |
| Game time | 100% | 111 | Waiting for data | |
| Injuries / lineup | 100% | 117 | Waiting for data |
What the proposals aim for
The goal is expected value (EV): profit per dollar bet. EV can only be measured against sportsbook lines, and those are not connected yet. Until they are, the proposals choose the weights that would have made projections closest to what happened, which is what makes EV possible. Weights move slowly with few games and faster with many. Nothing changes until a proposal is applied.
353 more projections are saved and waiting for results.