Record · 2026
Every prediction is stored before kickoff and graded against the final score. Nothing is edited after the fact. Wondering what a good number is? How accurate are NFL prediction models?
Our picks
93–40
69.9% of 133 games · 1 tie pushed
Benchmark pick
92–41
same games · 69.2%
Graded
Weeks 1–9
weeks graded so far
Confident picks
71%
hit rate on 75%+ picks (28)
12
longest streak · current 2
9.8
avg. margin error (pts)
11.1
avg. total error (pts)
9–8
when we differed from the market · the market 8–9
65.0%
backtest 2010–2025 · the market 66.5% →
Record by kickoff slot and by pick
The same record, cut the ways models usually show a bias. the market = the benchmark pick over the same games. Small samples early; judge after a month.
| Kickoff | Ours | % | the market |
|---|---|---|---|
| Thursday night | 6–2 | 75% | 4–4 |
| Sunday early (1:00) | 48–26 | 65% | 51–23 |
| Sunday late (4:00) | 27–6 | 82% | 27–6 |
| Sunday night | 7–2 | 78% | 5–4 |
| Monday night | 5–4 | 56% | 5–4 |
| Our pick was… | Ours | % | the market |
|---|---|---|---|
| picked the home team | 55–29 | 65% | 54–30 |
| picked the away team | 38–11 | 78% | 38–11 |
| picked the market pick | 84–32 | 72% | 84–32 |
| picked the other side | 9–8 | 53% | 8–9 |
Model vs. model + analyst layer
Since v0.4 the pick of record is the analyst-adjusted probability. The untouched model probability is stored and graded beside it, so the layer has to earn its keep. What it read and changed →
90–43
stats model alone
93–40
model + analyst (pick of record)
110
games the analyst moved
4–1
record when it flipped the pick
A 17-game season is a small sample: over the 272 games in a season the 95% margin on an accuracy figure is about ±6 points, so a season proves less than it looks. The backtest over 4,146 games is the fairer test of the method; this page is the live, pre-kickoff record.
Accuracy by week
Straight-up hit rate. Lighter bar = a week still being played. Hover a bar for the numbers.
Week 1Week 9
Does confidence mean anything?
Stated confidence vs. how often the pick was right. When we say 70%, are we right 70% of the time?
Hit rate by confidence × week
Darker = the model was right more often. A healthy board is dark at the top, lighter at the bottom, in every column.
W1
W2
W3
W4
W5
W6
W7
W8
W9
80%+
70–80%
60–70%
50–60%
Week by week
Click a week for its picks. Week 1 counts 14 of its 16 games: the two opening games (Wed 9/9 and Thu 9/10) kicked off before our first stored predictions, so they have no pick.
| Week | Games | Correct | Rate | Benchmark pick |
|---|---|---|---|---|
| Week 1 | 14 | 9 | 64% | 11/14 |
| Week 2 | 16 | 10 | 62% | 11/16 |
| Week 3 | 16 | 10 | 62% | 9/16 |
| Week 4 | 16 | 12 | 75% | 9/16 |
| Week 5 | 15 | 9 | 64% | 9/14 |
| Week 6 | 14 | 8 | 57% | 8/14 |
| Week 7 | 14 | 10 | 71% | 10/14 |
| Week 8 | 14 | 11 | 79% | 11/14 |
| Week 9 | 15 | 14 | 93% | 14/15 |
Follow your team. It's free.
Your team on top of every page, with its next game and our pick. Want more? Plus from US$7/mo, 7 days free.