Thirteen games left on the Week 6 card, and the model separates only two of them by less than a field goal. Every pick below is the Squawka Signal’s call as it stands this morning, with the projected margin and the win probability it carries. Thursday night is already in: Denver was the pick and Denver won.
Every NFL Week 6 prediction
| Kick-off (ET) | Game | Signal pick | Projected margin | Win prob |
|---|---|---|---|---|
| Thu Oct 15 | Seahawks at Broncos | Broncos (final: Broncos 17-14) | 1.5 | 56% |
| Sun Oct 18 | Texans at Jaguars | Jaguars | 4.6 | 67% |
| Sun Oct 18 | Titans at Colts | Colts | 9.5 | 81% |
| Sun Oct 18 | Panthers at Eagles | Eagles | 5.4 | 69% |
| Sun Oct 18 | Steelers at Buccaneers | Steelers | 2.7 | 60% |
| Sun Oct 18 | Ravens at Browns | Ravens | 7.4 | 76% |
| Sun Oct 18 | Saints at Giants | Giants | 4.3 | 66% |
| Sun Oct 18 | Bears at Falcons | Bears | 1.5 | 56% |
| Sun Oct 18 | Jets at Patriots | Patriots | 11.1 | 85% |
| Sun Oct 18 | Cardinals at Rams | Rams | 10.4 | 83% |
| Sun Oct 18 | Chargers at Chiefs | Chiefs | 7.0 | 74% |
| Sun Oct 18 | Bills at Raiders | Bills | 8.9 | 79% |
| Sun Oct 18 | Cowboys at Packers | Packers | 3.9 | 65% |
| Mon Oct 19 | Commanders at 49ers | 49ers | 10.3 | 83% |
Margins are the model’s projected winning margin, not a market line. Prices move; these do not, and they are published before kickoff so they can be checked afterwards.
The two the model cannot separate
Chicago at Atlanta is the closest call on the board at 56%, with the Bears favoured by 1.5. Pittsburgh at Tampa Bay is next at 60%, the Steelers by 2.7. Those are the two where the model is saying little more than that it leans one way, and they are the two worth reading the team news on rather than the number.
At the other end, the NY Jets at New England is the most confident call of the week at 85%, a projected 11.1 points. The model has gone 20-7 on calls at 70% and above this season, so the confident end of the board has earned more trust than the close end.
Four games worth a second look
Baltimore at Cleveland, Ravens by 7.4 at 76%. The number assumes a quarterback. Lamar Jackson sprained his left ankle in Week 4 and missed the Week 5 win at Atlanta, where Tyler Huntley started and took it to overtime, and Baltimore has not confirmed which of them starts here. Huntley has two starts to Jackson’s four appearances this season. A 76% call on a 4-1 team is reasonable; a 76% call that turns out to rest on the backup is a different proposition, so this is the one to check the designations on.
Houston at Jacksonville, Jaguars by 4.6 at 67%. The records say this is a mismatch and the points say otherwise. Jacksonville is 4-1 with a points differential of plus 53, the best in the league. Houston is 1-4, but it has been outscored by only two points all season, which is the smallest deficit of any team with four defeats. One of those two numbers is lying, and the model has split the difference rather than believing either.
Tennessee at Indianapolis, Colts by 9.5 at 81%. Tennessee has scored 58 points in five games, the fewest in the league, and has not won. This is the second week running the model has made the Titans a heavy underdog, and last week it was right by 23 points.
Carolina at Philadelphia, Eagles by 5.4 at 69%. Worth flagging because our own desk disagreed with the model on this one last week and took Carolina. Philadelphia is 2-3 having scored 92 points in five games; Carolina is 2-2 and rested. The model has not moved off the Eagles.
Thursday night, already settled
Seattle at Denver was the one game of the week the model had almost level, and it picked the Broncos by 1.5 at 56%. Denver won 17-14, so the call was right and the margin was close to it. One game is one game, but it is the sort the model is supposed to find: a narrow home win in a game two good pass defences were always likely to keep low.
How Week 5 went
Week 5 was the best week of the season so far: 10-4-1 straight up from 15 picks, 71.4%, with a mean margin error of 8.4 pts. The Giants and Washington finished level, which is a push rather than a win or a loss and is counted that way. That leaves the season at 48-30-1 from 79 picks, 61.5%, with the margin error averaging 9.8 pts. Every week of it is on the picks page, including the weeks that went badly.
What the confidence number is worth
Across 78 graded picks the bands sort the way they should, which is the only real test of a stated probability.
| Stated confidence | Picks | Record | Hit rate |
|---|---|---|---|
| 70% and above | 27 | 20–7 | 74% |
| 60% to 69% | 27 | 17–10 | 63% |
| Below 60% | 24 | 11–13 | 46% |
The bottom row is the one to take seriously. When the model says it is barely leaning, it has been wrong slightly more often than not, and that is what a 55% call should look like over 24 attempts. Only one of this week’s remaining games sits in that band, with five more between 60 and 69. The top row is where it has earned its keep, and seven of the thirteen games still to play are in it.
How these predictions are made
The Signal is our own model. It rates every team on what it has actually done, projects a margin for each fixture and converts that into a win probability, and it is published before kickoff and graded afterwards with nothing removed. The full method is on the methodology page, the weekly board lives on the NFL picks page, and each game has its own preview from the NFL hub. Availability can change a call late, so the injury report is the thing to read beside it.
19+ (18+ in AB/MB/QC) | Please play responsibly | Projections accurate at time of writing and subject to change | ConnexOntario: 1-866-531-2600 (ON). See your province’s helpline for resources elsewhere.