The Record
Oct 13, 7:42 AM EDTgrades land within about 35 minutes of a final
ATS · early weekOur first number after the week's lines open: Sunday afternoon for college, Sunday night for the NFL. Line value is measured from here to the last capture.169–154–652.3%
Avg miss, pts12.4market 11.5
Line valueHow far the market moved toward our side by the last capture — from the first seen number, and from the early-week line. A signal, never a grade.+0.41pts · our way 49.7%
Straight upHow often our number picked the outright winner.77.2%winners picked
329 games · still noisy
New college model from Sep 30 · earlier games graded under the previous one
Week by week
| Week · games | First seenOur side chosen against the first DraftKings or FanDuel number we captured, and graded at it. In college that capture can be a look-ahead weeks out; in the NFL it is the Sunday re-open. | Early weekOur first number after the week's lines open: Sunday afternoon for college, Sunday night for the NFL. Line value is measured from here to the last capture. | Last captureOur last pregame odds pull, usually inside the hour before kickoff. A median of the books we track, not a sharp close. | ValueHow far the market moved toward our side by the last capture — from the first seen number, and from the early-week line. A signal, never a grade. |
|---|---|---|---|---|
| Wk 1 · 51 | 19–31–1 | 19–31–1 | 21–28–2 | +0.4 |
| Wk 2 · 49 | 22–26–1 | 22–25–2 | 24–24–1 | +0.1 |
| Wk 3 · 57 | 35–21–1 | 32–24–1 | 29–27–1 | +0.5 |
| Wk 4 · 58 | 29–29 | 30–28 | 24–32–1 | +0.9 |
| Wk 5 · 56 · new model | 30–26 | 36–19–1 | 29–26–1 | −0.0 |
| Wk 6 · 58 · new model | 29–29 | 30–27–1 | 33–25 | +0.6 |
| Season · 329 | 164–162–3 | 169–154–6 | 160–162–6 | +0.4 |
Did our percentages hold?
| We said | Games | Average said | Covered |
|---|---|---|---|
| 25 | 49.9% | 60.0% | |
| 230 | 50.9% | 47.6% | |
| 14 | 52.8% | 71.4% | |
| All games | 269 | 50.9% | 50.0% |
Every game at the early-week line. 52.4% breaks even at −110. 4 no-read games (college weeks 1–4, 7+ point gaps) left out. 166 of 269 were stated within a point of 50%, a coin-flip read. Brier 0.249 vs 0.250 coin.
Totals (tracking) 158–166–4 · miss 12.7 vs 12.0
How we grade · 2018–25 backtest · 2025 archive