Our model has a pick on all 15 NFL Week 5 games, and this week it has gone on the road. Ten of the fifteen picks are the away side, having been the home side in ten of sixteen a week ago. Kansas City and Carolina are on bye.
Week 5 opens on Thursday 8 October with Tampa Bay at Dallas and closes on Monday 12 October with Buffalo at the Rams. Below is the full board with the projected winning margin and win probability behind every call, the game the model is strangest about, the four it cannot separate, and how last week went.
Every NFL Week 5 prediction
Four of the fifteen sit at 55% or below and four at 70% or above, so the middle of this board is unusually thin. The model is either fairly sure or barely committed, with little in between.
Houston at Tennessee, and the search for a first win
The biggest projected margin of the week is Houston by 8 points at Tennessee, at 77%. Both teams are 0-4. One of them will leave with a first win, which is the only reason a winless team carries the largest number on the board.
It is worth saying plainly what the model is doing here, because the two records look identical and the call is not close. Houston has scored 84 points and conceded 109. Tennessee has scored 55, third fewest in the league, and conceded 83. The model is not backing Houston because it rates them; it is backing them because Tennessee has been the worse team by some distance, and a points differential of minus 28 on 55 points scored is hard to argue with.
The other four winless sides are all picked to lose again. Tampa Bay at Dallas, Cincinnati at Miami and Denver at the Chargers all go against the 0-4 team. Only Houston is picked to win, and only because of the opponent.
The four the model cannot separate
Cleveland at the New York Jets is the closest call on the board at 0.5 points and 52%. The Browns are 3-1 and lead the AFC North, the Jets are 1-3 and have lost three in a row, and the model still has almost nothing in it.
Chicago at Green Bay, the Giants at Washington and Buffalo at the Rams are the other three at 55%, and all three are road picks. The Monday night game is the one to watch of those: Buffalo lost for the first time last week, and the Rams beat Philadelphia to reach 2-2.
One more worth flagging. San Francisco at Seattle is the only game this week where the model picks against an unbeaten team, and it does it by 1.7 points at 56%. San Francisco is 4-0 with the best points differential in the NFC. Seattle is 3-1 and at home.
How Week 4 went
Week 4 was the best week of the season on both measures: 10-5 straight up and a mean margin error of 6.4 points, down from 8.4 in Week 3 and 12.5 in Week 2. Four calls landed inside two points of the real result, with Seattle out by 0.4 and Kansas City by 0.7.
The week started badly. We had Pittsburgh at 68% on Thursday night and Cleveland won 27-24, a miss on the opening game. The Sunday that followed was the model’s best of the year.
Stated confidence still is not sorting results the way it should. The picks at 70% and above went 4-1 in Week 4, but those between 60% and 69% went 2-3 and those below 60% went 4-1. Over 63 picks the pattern has not settled, which is roughly what a sample this size should be expected to show.
How these predictions are made
The picks come from Squawka Signal, our in-house model. It produces a projected winning margin and a win probability for every game, and we publish both rather than only the pick, because a 52% call and a 77% call are not the same statement and should not look the same on the page.
The model is explained in full on the Squawka Signal methodology page, and every pick is logged on the NFL picks board before kickoff. For where these results lead, see the NFL playoff picture and our NFL power rankings. The NFL injury report carries every filed designation from Friday afternoon.
Our model picks Dallas, Jacksonville, Cleveland, Chicago, Cincinnati, New England, Minnesota, Houston, the New York Giants, Pittsburgh, Denver, Detroit, Seattle, Baltimore and Buffalo across the 15 Week 5 games. Ten of the fifteen are road picks. Kansas City and Carolina are on bye. Every pick carries a projected winning margin and a win probability on the board above.
Cleveland at the New York Jets. Our model gives Cleveland a 52% win probability and a projected margin of 0.5 points, the closest call on the board. Chicago at Green Bay, the Giants at Washington and Buffalo at the Rams are next, all at 55%.
Through four weeks the model is 38-25 straight up, a hit rate of 60.3%, with a mean margin error of 9.9 points. Week 4 was the best week so far at 10-5 and 6.4 points, counting the fifteen games played by Monday morning.
Tampa Bay at Dallas opens the week on Thursday 8 October at 8:15pm ET. Philadelphia at Jacksonville is the early kickoff at 9:30am ET on Sunday 11 October, eight games follow at 1:00pm ET, Denver at the Chargers is at 4:05pm ET and two more are at 4:25pm ET. Baltimore at Atlanta is the Sunday night game, and Buffalo at the Rams closes the week on Monday 12 October at 8:15pm ET.
19+ (18+ in AB/MB/QC) | Please play responsibly | Picks and projections accurate at the time of writing and may move before kickoff | ConnexOntario: 1-866-531-2600 (ON). See your province’s helpline for resources elsewhere.