2026 NFL Week 4 Digest
All 16 games of Week 4 in the 2026 NFL season are now graded. Before kickoff our model picked 13 of the 16 winners, four more than the market favorite. THE PLAY, our one headline call for each game, won 7 and lost 9, a gain of $3.92 on $80 staked. During play the live engine's win probability scored a Brier of 0.162 over 13,501 moments, ahead of the incumbent model at 0.168 and behind the market at 0.149, where lower is better.
Pregame calls
Before kickoff the model picked a winner in 16 games, and those picks won 13 and lost 3. The market favorite, the side the market priced above 50%, won 9 and lost 7. Against the spread our side won 9, lost 5 and pushed 2, and on totals it won 4 and lost 12. Scored as probabilities, our pregame win chances had a Brier score of 0.191 against 0.222 for the market, where lower is better, so ours were the closer forecasts this week.
THE PLAY and the $5 ledger
THE PLAY is the one headline call we publish for each game, and the $5 ledger stakes every pregame pick at $5 at the price it showed. THE PLAY won 7 and lost 9 across 16 games, winning 43.8% of the time at an average price of 46.7 cents, a gain of $3.92 on $80 staked. The largest single return was a gain of $14.23 on NE wins in NE at BUF, priced at 26 cents.
Every pick in the ledger together, 4,123 of them, won 41.2% of the time at an average price of 40.4 cents and finished with a gain of $777.20 on $20,615 staked. By market, player props finished with a gain of $1,659.30, spreads a loss of $149.07, totals a loss of $739.22 and moneylines a gain of $6.19.
The live engine
During play the live engine reprices every game about every ten seconds. Across 13,501 in-game moments in 16 games its win probability scored a Brier of 0.162. The incumbent, a simpler baseline model, scored 0.168 and the market scored 0.149. Lower is better.
| Quarter | Moments | Engine | Incumbent | Market |
|---|---|---|---|---|
| Q1 | 2,380 | 0.237 | 0.237 | 0.220 |
| Q2 | 4,760 | 0.188 | 0.195 | 0.174 |
| Q3 | 2,631 | 0.152 | 0.162 | 0.140 |
| Q4 | 3,730 | 0.088 | 0.095 | 0.080 |
| Game | 13,501 | 0.162 | 0.168 | 0.149 |
In-game edges and hot plays
An in-game edge is a stretch where the engine disagreed with the market price on the same side, counted once at the first price it showed. Across 16 games the engine logged 1,075 edges, and the 957 that could be graded won 46.2% of the time at an average price of 50.8 cents. Held to the final they returned a loss of 6.0 cents per $1 contract after fees, and sold 15 minutes after entry they returned a loss of 4.8 cents.
Hot plays are the edges that come right after a score, a turnover or a big play. Across 16 games the engine logged 584 hot plays, and the 553 that could be graded won 45.2% of the time at an average price of 51.1 cents. Held to the final they returned a loss of 7.3 cents per $1 contract after fees, and sold 15 minutes after entry they returned a loss of 4.1 cents.
Live player props
Live prop projections were graded in 16 games, 480 props in all. At kickoff the average projection missed the final stat by 17.4, measured in each stat's own units, and by the end of the third quarter the average miss was 8.5. The Brier score of the projected chance of going over the line was 0.264 at kickoff and 0.129 at the end of the third quarter. The market's own chance scored 0.249 on the 359 props with a market price.
- Receiving yards, 180 props: an average miss of 23.2 yards at kickoff and 10.7 yards at the end of the third quarter.
- Receptions, 174 props: an average miss of 1.7 catches at kickoff and 0.8 catches at the end of the third quarter.
- Rushing yards, 94 props: an average miss of 18.6 yards at kickoff and 7.7 yards at the end of the third quarter.
- Passing yards, 32 props: an average miss of 66.8 yards at kickoff and 39.6 yards at the end of the third quarter.
The best and worst calls
The best call is the pregame winner pick where we disagreed most with the market and were right, and the worst is the one where we disagreed most and were wrong. The best came in ARI at NYG: we gave NYG a 56% chance to win when the market gave them 43%, and NYG won 36 to 24. The worst came in DAL at HOU: we gave HOU a 64% chance to win when the market gave them 58%, and DAL won 34 to 30.
Every game
| Game | Final | Our winner | Spread | Total | THE PLAY | $5 result |
|---|---|---|---|---|---|---|
| PIT at CLE | CLE 27, PIT 24 | CLE 50%, won | CLE +2.5, won | Over 38.5, won | NO: PIT by 3.0+, won | +$5.64 |
| IND at WAS | IND 30, WAS 13 | IND 52%, won | WAS +4.5, lost | Over 48.5, lost | WAS wins, lost | −$5.00 |
| ARI at NYG | NYG 36, ARI 24 | NYG 56%, won | NYG +2.5, won | Over 43.5, won | Over 44.0, won | +$4.55 |
| DAL at HOU | DAL 34, HOU 30 | HOU 64%, lost | HOU −3, lost | Under 48.5, lost | NO: DAL by 2.0+, lost | −$5.00 |
| GB at TB | GB 17, TB 14 | GB 63%, won | GB −3, push | Over 38.5, lost | Over 39.0, lost | −$5.00 |
| JAX at CIN | JAX 22, CIN 17 | JAX 54%, won | JAX +2.5, won | Under 51.5, won | JAX ML, won | +$6.00 |
| LA at PHI | LA 24, PHI 20 | LA 54%, won | PHI +3.5, lost | Over 42.5, won | PHI wins, lost | −$5.00 |
| NE at BUF | NE 29, BUF 26 | BUF 64%, lost | NE +7, won | Under 49.5, lost | NE wins, won | +$14.23 |
| NYJ at CHI | CHI 23, NYJ 12 | CHI 66%, won | CHI −3.5, won | Over 43.5, lost | CHI -3.5, won | +$4.55 |
| TEN at BAL | BAL 24, TEN 18 | BAL 85%, won | BAL −11.5, lost | Over 42.5, lost | Over 42.5, lost | −$5.00 |
| MIA at MIN | MIN 15, MIA 10 | MIN 81%, won | MIN −10, lost | Over 38.5, lost | Over 38.5, lost | −$5.00 |
| DEN at SF | SF 24, DEN 14 | SF 67%, won | SF −3, won | Over 48.5, lost | SF -3.0, won | +$4.67 |
| KC at LV | KC 30, LV 27 | KC 57%, won | LV +4.5, won | Under 47.5, lost | Under 47.5, lost | −$5.00 |
| LAC at SEA | SEA 30, LAC 23 | SEA 83%, won | SEA −7, push | Under 42.5, lost | NO: over 42.5, lost | −$5.00 |
| DET at CAR | CAR 32, DET 26 | DET 57%, lost | CAR +3.5, won | Under 51.5, lost | CAR wins, won | +$9.29 |
| ATL at NO | ATL 45, NO 24 | ATL 57%, won | ATL +1.5, won | Under 47.5, lost | NO: over 47.5, lost | −$5.00 |
How to read this
A win rate means little without the price paid, so every rate here sits beside the average price, and a price of 53 cents means the market put the chance at about 53%. A Brier score measures how far a probability sat from what happened, from 0 for a perfect forecast to 1 for a certain one that was wrong, so lower is better. In-game returns are per $1 contract after fees.
Every game, every edge and every prop behind these numbers is on the NFL results page, graded as each game finishes.