Week 4 Model Picks

Week 3, if you had bet the posted card: 1-1 on spreads, 0-3 on parlays, -3.10 units. Chicago +4.5 covered by fifteen. Miami +10.5 did not, and all three moneyline parlays died. Season to date: -11.41 units. You didn’t bet it. Here is Week 4, on the same rules. A spread has to clear a raw edge of 4.0. Totals stay on the page and off the card. Three 1-unit moneyline parlays. The budget is eight units. This week spends eight.

Our Bets

Spreads (5 units)

  • 1 unit on Jaguars +2.5 (-105). The Model has Jacksonville by 5.4, against Cincinnati -2.5. That is a 7.9-point edge, the biggest on the board, and the only game this week where The Model and the market pick different teams and The Model is more confident. Trevor Lawrence is at 74.7 in the file, Joe Burrow at 76.4. The bet is not “The Model loves Jacksonville.” The bet is “2.5 is too few.”
  • 1 unit on Bears -3.5 (-105). The Model has Chicago by 10.9 against the Jets’ 3.5. The Chicago quarterback in the file is Case Keenum at 74.5, with matched history this week after sitting at a 45 default last week. If the depth chart changes again, this number is stale. Same caveat as Week 3: the bet is “3.5 is too few,” not a scouting report on Keenum.
  • 1 unit on Dolphins +10.5 (-112). The Model has Minnesota by 4.0, not 10.5. Miami’s row is Malik Willis at 40.7. Ten and a half is a lot of points twice in a row.
  • 1 unit on 49ers -3.0 (+100). The Model has San Francisco by 8.6 against a field goal. Brock Purdy is at 89.3, Bo Nix at 47.1. Even money on a number The Model calls three points too short.
  • 1 unit on Seahawks -7.0 (-115). The Model has Seattle by 11.8 against 7.0. Justin Herbert’s row is at 29.0 in the file. If that is an injury split and not the real Herbert, this number moves — read the starter notes below before betting it.

Parlays (1 unit each)

  • Jaguars ML (+120) / Colts ML (-192) / Chiefs ML (-218). Pays +388.
  • Saints ML (-148) / Lions ML (-180) / Bills ML (-298). Pays +248.
  • 49ers ML (-148) / Bears ML (-175) / Seahawks ML (-380). Pays +233. That is three card bets in a trench coat. It hits when the card hits.

No Thursday spread. Pittsburgh at Cleveland is CLE -2.5 and The Model calls it Cleveland by 0.3. A 2.8-point edge stays off the card. Green Bay at Tampa Bay is the hard one: The Model has Tampa Bay by 0.5, the raw edge is 3.99, and the table rounds it to 4.0. Week 3 we left New Orleans off at 3.98. Same rule. It stays off.

The Model

Same ensemble: Ridge, LightGBM, XGBoost, and a small neural net, refit on games through Week 3, using team stats, EPA, and starter QBR. Through three weeks it is 26-22 straight up, 21-27 against the spread, and behind the market on margin error, 11.5 to 10.6. It is not a money printer.

The Model picks for Week 4, 2026: per-model margin of victory, pick, DraftKings spread, edge, win probability, and moneyline for all 16 games

The table is the full 16-game card: per-model margin, pick, DraftKings spread, raw edge, straight-up win probability, and moneyline. Jacksonville is the only outright side split where The Model is more confident than the market. Chicago is the biggest favorite-cover call on the board.

Over/Under Picks

Totals went 9-7 in Week 3 after 4-12 and 5-11. We are showing the numbers and not betting them. The biggest total edge is Kansas City at Las Vegas Under 47.5. The Model says 42.2. Still not a bet.

The Model over/under picks for Week 4, 2026: predicted team scores, model total, DraftKings total, and difference for all 16 games

Where we disagree with Vegas

  • JAX at CIN: CIN -2.5; model MOV -5.4. On the card. The only outright side split of the week where The Model is more confident than the market.
  • NYJ at CHI: CHI -3.5; model MOV +10.9. On the card. Chicago’s file quarterback is Keenum at 74.5 with matched history. Caveat above.
  • MIA at MIN: MIN -10.5; model MOV +4.0. On the card for the points.
  • DEN at SF: SF -3.0; model MOV +8.6. On the card. Purdy at 89.3 is doing the work.
  • LAC at SEA: SEA -7.0; model MOV +11.8. On the card, and the Herbert number is why. See the starter notes below.
  • GB at TB: GB -3.5; model MOV +0.5. Off the card at 3.99. The raw edge does not clear 4.0, and the Tampa Bay quarterback in the file has no matched QBR history, so it is riding a 45 default anyway.
  • PIT at CLE: CLE -2.5; model MOV +0.3. Off the card at 2.8. Rodgers is at 43.0, Watson at 58.4, and it plays tonight.

Two starter notes that do not move the card. Seattle’s row is Drew Lock, which matches the reported starter from last week. The Chargers’ row is Justin Herbert at 29.0. That is a steep drop for a healthy quarterback, and if it is an injury split in the QBR file, Seattle -7 is stale in the wrong direction. Chicago is still Case Keenum, now with matched history at 74.5. Miami’s row is Malik Willis at 40.7.

Remember, the best way to use this is as one input into your own thinking. The Model does not know who is holding out, and it only knows the quarterback nflverse listed.

These are our personal bets and opinions, not betting advice. Wager at your own risk and within your means.

Good luck everyone.

Leave a Reply

Your email address will not be published. Required fields are marked *