NFL Weekly Record
How the leans on the Spread Dial and the Totals Dial did, week by week, once the games were played. Every game counts, graded against the closing line, and each week is scored on formulas fitted only on the games before it.
Informational only. This is a record of a formula against past closing lines; it is not betting advice, and no wagering is offered or arranged here.
Through week 8 of 2026, the Spread Dial's leans are 57–58–6 against the closing spread (49.6%), and the Totals Dial's are 54–67–0 against the closing total (44.6%). Break-even at standard −110 pricing is 52.38%. Over 115 decided games, a coin flip lands anywhere from about 40.9% to 59.1%, so this far into a season the record says more about luck than about the formula.
2026 by week
| Week | Spread (W–L–P) | Cover % | Total (W–L–P) | Hit % | Season, spread / total |
|---|---|---|---|---|---|
| Week 1 | 4–11–1 | 26.7% | 7–9–0 | 43.8% | 4–11–1 / 7–9–0 |
| Week 2 | 7–9–0 | 43.8% | 6–10–0 | 37.5% | 11–20–1 / 13–19–0 |
| Week 3 | 7–7–2 | 50.0% | 9–7–0 | 56.3% | 18–27–3 / 22–26–0 |
| Week 4 | 10–4–2 | 71.4% | 7–9–0 | 43.8% | 28–31–5 / 29–35–0 |
| Week 5 | 8–6–1 | 57.1% | 6–9–0 | 40.0% | 36–37–6 / 35–44–0 |
| Week 6 | 5–9–0 | 35.7% | 5–9–0 | 35.7% | 41–46–6 / 40–53–0 |
| Week 7 | 6–8–0 | 42.9% | 9–5–0 | 64.3% | 47–54–6 / 49–58–0 |
| Week 8 | 10–4–0 | 71.4% | 5–9–0 | 35.7% | 57–58–6 / 54–67–0 |
Regular-season weeks whose games have all been played. W–L–P: the side the formula leaned to (a team for the spread, over or under for the total) won, lost, or landed on the line. Percentages leave pushes out.
How each week is graded
For week N, both dials are rebuilt as they stood before that week: every game from week N on is treated as unplayed, so the power ratings, recent form, scoring pace, cover trends, play efficiency and the fitted coefficients all come from earlier games only. Each formula then scores every week N game at its fitted weights (every dial at 1×): the Spread Dial leans to the team it expects to beat the spread, the Totals Dial to the over or the under. After the games, each lean is graded against the final score.
- The line is the closing spread or total in nflverse's schedule file, not the line a dial showed earlier in the week.
- Every game counts, including leans of under a point. A game without a closing line would be listed and left ungraded.
- A push is a final margin or combined score exactly on the line; it counts for neither side.
- The fits are the same least-squares fits the two dials use, on every game since 2015 before the graded week. Each week page prints the coefficients it used.
For the spread this is the same test as the dial's out-of-sample record, refitted every week instead of every season. It is not a betting record: no one placed these leans, and a record near 50% is what both dials' histories say to expect.
Sources
- nflverse schedules and results (games.csv), CC-BY 4.0: final scores, closing spreads, rest days, sites and starting quarterbacks.
- nflverse weekly team stats, CC-BY 4.0: the EPA behind the play-efficiency factor.
Last reviewed: September 2026. Not affiliated with or endorsed by the NFL or any team.
Frequently asked questions
Is this what the Spread Dial showed during the week?
Almost, not exactly. Each week is rebuilt from the schedule file: both formulas are refitted on every game before that week, and each lean is scored against nflverse's closing spread or total. During the week the dials may have shown a median of sportsbook lines from an earlier refresh, and a late quarterback change could have moved a factor after you looked. Grading against the closing line is the standard way to keep a record, and it means anyone can reproduce these numbers from the public file.
Why does every game count, even the coin flips?
So the record can't be flattered by choosing which games to keep. Every game with a closing line gets graded on the side the formula leaned to, however slightly. Each dial's own record panel shows how leans of different sizes have done.
What record would be good?
At standard −110 pricing a side has to win 52.38% of the time just to break even. Over one season of about 270 games, a coin flip lands anywhere from roughly 44% to 56% about 95% of the time, so a single season says little. The dial's out-of-sample record since 2016, which is the longer test, is near a coin flip, and the site says so.
What about the playoffs?
Only regular-season weeks are graded here for now. Playoff games are still in both dials' fits and the Spread Dial's season-by-season record.
Related tools
Every game this week, scored against the point spread from seven public factors, with an out-of-sample record since 2016. Turn the dials to weight them your way and watch the back-test update.
Open tool →NFL Totals DialEvery game this week, scored against the over/under from five public factors: pace, roof, combined strength, division game. Turn the dials and watch the back-test update.
Open tool →NFL ATS RecordsEvery NFL team's record against the spread and the total since 2015, with home, road, favorite, underdog and off-a-bye splits — and how much of it is chance.
Open tool →