NFL Week 4: What the Model Sees in Every Game

Two weeks of NFL data. Fifteen games this weekend. Not one of them is a pick.

That is not a hedge, it is the rule. A pick requires 75 confidence, and 75 confidence requires evidence. Two games is not evidence. So every read below is a lean: a directional take we publish for transparency and do not count in the graded record.

Here is what the model sees.

Thursday night

Falcons @ Packers (-4.5), Lambeau Field

Green Bay's +2.2 yards per pass attempt and an 89.6-to-38.2 quarterback rating gap are the strongest edges on this card. Atlanta is also on a short week and sits at -6 in turnovers, a rate that regresses upward rather than down. The number is fair, which is why this is a lean and not more.

Sunday, 1pm

Seahawks (-7) @ Commanders

Seattle's +3.3 yards per pass attempt and +23 points per game of scoring margin are genuine. The problem is the price: 7 is the second key number in football, and you pay full retail for it. Take the efficiency edge without paying for the key number.

Panthers (-2.5) @ Browns

Carolina owns the largest scoring-margin gap on the slate (+14.5 per game) and converts red-zone trips into touchdowns at 70% against Cleveland's 20%. That red-zone split is the whole case. Both teams are 1-1, and the records hide a wide gap in how they got there.

Bills (-7) @ Highmark Stadium

Buffalo's +2.3 yards per pass attempt and +25.3 percentage points on third down give it the clearest sustained-drive advantage of any game this week. Again the number is the issue — a full touchdown plus the hook against a division opponent.

Chiefs (-11.5) @ Dolphins

The statistical profile is lopsided: Kansas City is +30 per game in scoring margin, Miami has scored 13 points a game and allowed 31. The model still calls this a lean. Laying 11.5 on the road in September, on two games of evidence, is how a good record becomes a bad one.

Bengals (-3.5) @ Steelers

Cincinnati's +2.0 yards per pass attempt and +24.8pp third-down edge are real. But -3.5 crosses the most important number in football. You are paying a premium to be on the wrong side of 3.

Jaguars (-3) @ Patriots

Jacksonville's 110.6 passer rating against New England's 63.1 is the largest single gap in the matchup. Getting a field goal rather than a field goal and a hook is the difference between a fair bet and a bad one.

Colts (+2.5) @ Texans

Indianapolis converts red-zone trips into touchdowns at 83.3% against Houston's 44.4% — a 39-point swing on the most decisive part of the field. Taking the points at home in a near coin-flip game is the value side.

Titans @ Giants (-2.5)

The Giants are better situationally (48% on third down against 40%, 66.7% in the red zone against 60%) and the line sits under the key number 3. Neither team has shown much, which is exactly why this is a lean.

Sunday, late

Vikings (-1.5) @ Buccaneers

Minnesota has a +16.5 per-game scoring margin, an 80% red-zone touchdown rate against Tampa Bay's 37.5%, and gives away four fewer penalties per game. The margin is the smallest on the board, and the underlying gap is not.

49ers (-8.5) @ Cardinals

San Francisco has decisive edges in pass efficiency (+3.1 per attempt), third down, red zone and scoring margin (+27 per game). The number asks you to lay more than a touchdown on the road, which is why this stays a lean.

Raiders (+3) @ Saints

A pass rush with eight sacks against a line that has allowed eight in two games is a nameable line-of-scrimmage mismatch. Getting the underdog at the key number 3 adds half a point of protection that matters more than it looks.

Ravens (-3.5) @ Cowboys — Maracanã Stadium, Rio de Janeiro

Dallas is the home team on paper. Dallas is not at home. Neutral-site games remove roughly 2.5 points of home-field advantage, and the NFL's international crowds are frequently split. Baltimore's +3.0 yards per pass attempt advantage is the strongest single predictive edge on the slate, and the venue makes the number softer than it appears.

Monday night

Rams (-2.5) @ Broncos

Los Angeles holds a +1.3 yards per pass attempt edge and a red-zone touchdown rate 37 points better than Denver's. Denver at altitude is a real factor, but it is already in the price.

What a lean is, and why we publish them

A pick is a bet we would make and grade. A lean is a directional read that did not clear 75 confidence. We publish leans anyway, for the same reason we publish losses: a record you only see part of is not a record.

Every one of these fifteen reads says the same thing in its own way — worth leaning, not worth betting yet. In three weeks the sample will be five games instead of two, the model's confidence ceiling lifts, and some of these become picks.

Until then the honest answer is the one above. You can read every pick and lean in full, graded, on the NFL page.