New to these reports? Start here
- Dotted-underlined words have a plain-English definition — hover or tap them. Every term is also on the glossary page.
- "Null" means we found nothing, not that something broke. Most reports here are negative results, on purpose — knowing an idea doesn't work is the point.
- Two questions get asked separately. First, is the effect real? Second, is it already priced into the betting odds? An effect can be completely real and still useless to bet on.
- A "calibration" row is a self-check. It runs the same method on something already known to be true. If that fails, the whole report is unreliable — so it's reported alongside the findings.
- If a confidence interval includes zero, the real effect might be nothing at all, so no claim gets made.
NFL alt spread / alt total ladders vs the real margin distribution (2026-10-31)
Answer: no edge. Books price alt rungs the way recent history says they should, including the key numbers 3 and 7. Every rung carries about 5–6.6% hold, far more than any mispricing we found.
Data
- Books: alt spread and alt total ladders from 8 books (DraftKings, FanDuel, BetMGM, BetRivers, Fanatics, Caesars, Bovada, Pinnacle) for 108 games in 2026 weeks 1–8. This came to 31,835 spread rungs and 29,521 total rungs, captured once, 1.5–4 hours before kickoff.
- Truth: the distribution of nflverse final scores, conditioned on the closing spread and total with a kernel. There's no normal approximation, so key numbers keep their lumps. Spreads are mirrored to both teams' views.
- Pricing: book no-vig uses exact Shin. Whole-number rungs are compared on P(win | no push).
Results
Book minus model, in percentage pointspercentage pointsAn absolute difference between two percentages. Going from 50% to 53% is 3 percentage points, not 3%. of win probability:
| History used | Longshot rungs (rung more than 7 points above the main line) | Rungs crossing 3 | Rungs crossing 7 | Rungs on a key number |
|---|---|---|---|---|
| 2006–2025 | −1.3 to −1.45 | +0.44 ± 0.19 | −0.17 ± 0.24 | +0.08 ± 0.16 |
| 2015–2025 | −0.4 to −0.8 | +0.05 ± 0.15 | −0.12 ± 0.24 | −0.02 ± 0.17 |
| 2020–2025 | −0.4 to −0.6 | −0.01 ± 0.15 | −0.25 ± 0.31 | −0.07 ± 0.19 |
The longshot gap in the 2006+ model is era drift. The spread of results around the closing spread shrank from 13.8 points (2006–10) to 12.5 (2024–25), and 2026 so far is 11.9. Books price the recent distribution.
"+EV" rungs under the model (EV ≥ 3% at the best price) were mostly longshots at +380, and they didn't hold up on 2026 results:
| History used | Realised ROIreturn on investmentProfit as a percentage of the money wagered. +2% means $2 profit per $100 bet., spreads | Realised ROI, totals |
|---|---|---|
| 2006–2025 | +17% ± 21 | −19% ± 19 |
| 2015–2025 | +15% | −12% |
| 2020–2025 | −16% | −21% |
The model's apparent EV is roughly the size of its own uncertainty.
Totals rungs: books and model agree within about 0.7 points in every band. On 2026 results, books had slightly better log losslog lossA score for probability forecasts that punishes confident wrong answers harshly. Lower is better. than the model (0.507 vs 0.512).
Hold on alt rungs by distance from the main line: 5.0% at the main line, rising to 6.6% at 7–10 points away. A rung would need a mispricing of about 3 points before it pays, and none comes close.
Run: python3 backtests/nfl_alt_ladders/ladders.py [first_history_season]