NFL

Model record

Every game since Oct 10, 2026, settled against the first spread we captured. Simulated 1 unit a game, not real bets.

The model's side is the favorite when its margin is bigger than the line's, the underdog when smaller. Lines are DraftKings first. Pushes return the stake; postponed, cancelled games are no action.

5–6–1 against the spread (W–L–P) · since Oct 10

45.5% hit rate · break-even 52.2% at the prices taken

-1.3 units at 1 a game · ROI -10.8%

1–0–1 when the model disagreed by 3+ points · +0.9 units

11 settled games: at this size the hit rate could be off by about ±29 points either way, far too few to separate skill from luck.

0 no action · 2 pending · 0 with no position (the model matched the market, or no price was posted).

Model's side vs the line	Result	Units
Week 5, 2026 →
BAL +3.0BAL 30, ATL 23	✓ covered	+1.02
ARI +5.5DET 45, ARI 21	✗ didn't cover	-1.00
SEA -3.0SF 32, SEA 29	✗ didn't cover	-1.00
LAC +3.5DEN 23, LAC 17	✗ didn't cover	-1.00
GB +1.5CHI 7, GB 31	✓ covered	+0.95
MIA +7.0CIN 31, MIA 24	push	+0.00
NYJ -2.5CLE 12, NYJ 17	✓ covered	+0.91
TEN +7.0HOU 26, TEN 3	✗ didn't cover	-1.00
IND +3.0IND 22, PIT 34	✗ didn't cover	-1.00
NE -3.5LV 17, NE 45	✓ covered	+0.93
NO +2.5MIN 23, NO 13	✗ didn't cover	-1.00
PHI +7.5PHI 17, JAX 19	✓ covered	+0.89

Spread