Week 6 · 2026 · Oct 15–19 · 14 games, none played yet
Sixteen games, the model’s number against the market’s — each labeled, never merged. The bar under each matchup is the model’s win probability: calibrated out of sample (a team shown at 74% has won 77% of the time across fourteen seasons), not the market’s, and not a recommendation — the market’s own probabilities remain slightly sharper.
Lines and schedule from nflverse, retrieved 2026-10-14 — not when this page was built.
Widest disagreement this week
Furthest the model sits from the line. That makes it the most interesting game to look at — not the best bet. Accuracy does not rise with disagreement.
Kickoff 2026-10-18 13:00 ET
How far each model input pushes the prediction away from the market line, in points (positive favors CLE). Exact linear attribution of the market-anchored member — these four sum to most of the gap.
Kickoff 2026-10-15 20:15 ET
How far each model input pushes the prediction away from the market line, in points (positive favors DEN). Exact linear attribution of the market-anchored member — these four sum to most of the gap.
Kickoff 2026-10-18 16:05 ET
How far each model input pushes the prediction away from the market line, in points (positive favors LAR). Exact linear attribution of the market-anchored member — these four sum to most of the gap.
Kickoff 2026-10-18 13:00 ET
How far each model input pushes the prediction away from the market line, in points (positive favors CLE). Exact linear attribution of the market-anchored member — these four sum to most of the gap.
Kickoff 2026-10-18 16:25 ET
How far each model input pushes the prediction away from the market line, in points (positive favors LV). Exact linear attribution of the market-anchored member — these four sum to most of the gap.
Kickoff 2026-10-18 13:00 ET
How far each model input pushes the prediction away from the market line, in points (positive favors PHI). Exact linear attribution of the market-anchored member — these four sum to most of the gap.
Kickoff 2026-10-18 13:00 ET
How far each model input pushes the prediction away from the market line, in points (positive favors ATL). Exact linear attribution of the market-anchored member — these four sum to most of the gap.
Kickoff 2026-10-18 20:20 ET
How far each model input pushes the prediction away from the market line, in points (positive favors GB). Exact linear attribution of the market-anchored member — these four sum to most of the gap.
Kickoff 2026-10-18 09:30 ET
How far each model input pushes the prediction away from the market line, in points (positive favors JAX). Exact linear attribution of the market-anchored member — these four sum to most of the gap.
Kickoff 2026-10-18 16:25 ET
How far each model input pushes the prediction away from the market line, in points (positive favors KC). Exact linear attribution of the market-anchored member — these four sum to most of the gap.
Kickoff 2026-10-18 13:00 ET
How far each model input pushes the prediction away from the market line, in points (positive favors NYG). Exact linear attribution of the market-anchored member — these four sum to most of the gap.
Kickoff 2026-10-18 13:00 ET
How far each model input pushes the prediction away from the market line, in points (positive favors NE). Exact linear attribution of the market-anchored member — these four sum to most of the gap.
Kickoff 2026-10-18 13:00 ET
How far each model input pushes the prediction away from the market line, in points (positive favors TB). Exact linear attribution of the market-anchored member — these four sum to most of the gap.
Kickoff 2026-10-18 13:00 ET
How far each model input pushes the prediction away from the market line, in points (positive favors IND). Exact linear attribution of the market-anchored member — these four sum to most of the gap.
Kickoff 2026-10-19 20:15 ET
How far each model input pushes the prediction away from the market line, in points (positive favors SF). Exact linear attribution of the market-anchored member — these four sum to most of the gap.
They live on Betting, evaluated against every market with the house price de-vigged, the expected value and the stake at your bankroll.
This site says the model does not beat the closing line, and that page lists bets. It is not a contradiction: those are the games where the model departs most from the market, and out of sample accuracy does not rise with disagreement — 49.3%, 50.9% and 48.8% by bucket, none of them reaching the 52.4% breakeven at −110. A lean is where the model stands, never a promise, and the full measurement is on that page.
As they stood after the last game played. Under Defense, a high number means a permissive defense: expected offense against it is offense + opponent defense, added.
| Team | Elo | Offense | Defense | Net |
|---|---|---|---|---|
| SEA | 1,679 | 0.003 | -0.033 | 0.035 |
| BUF | 1,658 | 0.084 | 0.015 | 0.069 |
| SF | 1,656 | 0.057 | 0.008 | 0.048 |
| MIN | 1,632 | -0.034 | -0.038 | 0.004 |
| JAX | 1,626 | 0.035 | -0.009 | 0.044 |
| NE | 1,613 | 0.027 | -0.011 | 0.038 |
| DEN | 1,606 | 0.003 | -0.019 | 0.023 |
| LAR | 1,587 | 0.055 | -0.010 | 0.064 |
| BAL | 1,576 | 0.052 | 0.000 | 0.051 |
| KC | 1,569 | 0.034 | -0.002 | 0.036 |
| DET | 1,558 | 0.063 | 0.018 | 0.045 |
| CHI | 1,536 | 0.021 | 0.007 | 0.014 |
| HOU | 1,532 | -0.008 | -0.032 | 0.024 |
| PHI | 1,525 | -0.002 | -0.010 | 0.009 |
| PIT | 1,523 | -0.004 | 0.006 | -0.009 |
| CIN | 1,522 | 0.027 | 0.023 | 0.004 |
| GB | 1,522 | 0.035 | 0.004 | 0.031 |
| ATL | 1,464 | -0.001 | 0.004 | -0.005 |
| CAR | 1,450 | 0.007 | 0.024 | -0.017 |
| IND | 1,449 | 0.004 | 0.016 | -0.012 |
| NYG | 1,445 | -0.005 | 0.021 | -0.026 |
| TB | 1,440 | -0.002 | 0.002 | -0.004 |
| DAL | 1,438 | 0.040 | 0.046 | -0.006 |
| CLE | 1,425 | -0.061 | -0.014 | -0.047 |
| LAC | 1,418 | -0.034 | -0.017 | -0.018 |
| WAS | 1,414 | 0.012 | 0.031 | -0.020 |
| LV | 1,406 | -0.043 | 0.009 | -0.051 |
| NO | 1,380 | -0.025 | 0.007 | -0.032 |
| NYJ | 1,365 | -0.032 | 0.018 | -0.051 |
| MIA | 1,364 | -0.015 | 0.024 | -0.039 |
| ARI | 1,358 | 0.001 | 0.032 | -0.030 |
| TEN | 1,266 | -0.051 | 0.029 | -0.080 |