DegenPredicts — NFL

NFL spreads and totals · Week 1 · updated 2026-09-08
☕ Buy me a coffee
This model does not beat the NFL closing line. Out of sample across 6 seasons its margin error was 9.91 points against the line's 9.77, covering 53.1% of the time on 1660 games — about 1.23 points of standard error, so indistinguishable from a coin flip. The published number is therefore the line nudged 15% of the way toward the model, and the numbers below are worth tracking rather than backing.

Week 1 board

Sun Sep 13, 1:00 PM Δ5.4 · early season
CLEatJAX
new CLE QB (Deshaun Watson)
Total no play
Over 40.5
model 45.9 · +5.4 vs line
Spread no play
JAX -8.5
model margin 8.8 · +0.3 vs line
Over 40.5 pays 1.90x fair 1.94x -2%
JAX by over 7.5 pays 1.83x fair 1.87x -2%
CLE to win pays 4.13x fair 4.00x +3%
Sun Sep 13, 1:00 PM Δ5.1 · early season
NYJatTEN
new NYJ QB (Geno Smith)
Total no play
Over 39.5
model 42.2 · +2.7 vs line
Spread no play
NYJ +1.5
model margin -3.6 · -5.1 vs line
TEN by over 1.5 pays 1.93x fair 2.11x -8%
NYJ to win pays 2.01x fair 2.09x -4%
Sun Sep 13, 4:25 PM Δ3.4 · early season
GBatMIN
new MIN QB (Kyler Murray) · divisional
Total no play
Under 46.5
model 43.1 · -3.4 vs line
Spread no play
GB +1.5
model margin -1.2 · -2.7 vs line
Over 46.5 pays 1.93x fair 2.04x -5%
MIN by over 1.5 pays 1.93x fair 2.05x -6%
GB to win pays 2.05x fair 2.15x -4%
Sun Sep 13, 1:00 PM Δ3.3 · early season
BUFatHOU
Total no play
Under 44.5
model 41.7 · -2.8 vs line
Spread no play
HOU +1.5
model margin 1.8 · +3.3 vs line
Over 44.5 pays 1.83x fair 2.04x -10%
HOU by over 1.5 pays 2.09x fair 2.37x -12%
BUF to win pays 1.86x fair 1.88x -1%
Sun Sep 13, 8:20 PM Δ3.2 · early season
DALatNYG
divisional
Total no play
Under 48.5
model 45.3 · -3.2 vs line
Spread no play
DAL -3.0
model margin -4.4 · -1.4 vs line
NYG by over 2.5 pays 2.59x fair 3.05x -15%
NYG to win pays 2.34x fair 2.49x -6%
Sun Sep 13, 1:00 PM Δ3.0 · early season
CHIatCAR
Total no play
Under 47.5
model 46.9 · -0.6 vs line
Spread no play
CAR +3.0
model margin 0.0 · +3.0 vs line
Over 47.5 pays 2.05x fair 2.01x +2%
CHI by over 2.5 pays 1.76x fair 2.00x -12%
CAR to win pays 2.34x fair 2.37x -1%
Sun Sep 13, 1:00 PM Δ2.4 · early season
NOatDET
Total no play
Under 49.5
model 47.1 · -2.4 vs line
Spread no play
NO +7.0
model margin 6.0 · -1.0 vs line
Over 49.5 pays 1.86x fair 2.02x -8%
NO by over 7.5 pays 9.41x fair 7.71x +22%
NO to win pays 3.52x fair 3.40x +4%
Sun Sep 13, 1:00 PM Δ2.2 · early season
BALatIND
new IND QB (Daniel Jones)
Total no play
Under 47.5
model 46.6 · -0.9 vs line
Spread no play
BAL -3.5
model margin -5.7 · -2.2 vs line
Over 47.5 pays 1.83x fair 2.01x -9%
IND by over 3.5 pays 3.40x fair 3.52x -4%
BAL to win pays 1.57x fair 1.62x -3%
Wed Sep 09, 8:20 PM Δ1.8 · early season
NEatSEA
NE crosses 3 time zones · 3,998 km trip
Total no play
Under 44.5
model 42.7 · -1.8 vs line
Spread no play
NE +3.0
model margin 1.3 · -1.7 vs line
Over 44.5 pays 1.97x fair 2.02x -3%
SEA by over 2.5 pays 1.70x fair 1.98x -14%
NE to win pays 2.52x fair 2.40x +5%
Sun Sep 13, 1:00 PM Δ1.6 · early season
ATLatPIT
new ATL QB (Tua Tagovailoa)
Total no play
Under 42.5
model 40.9 · -1.6 vs line
Spread no play
ATL +3.5
model margin 3.5 · +0.0 vs line
Over 42.5 pays 1.97x fair 2.02x -3%
PIT by over 3.5 pays 1.93x fair 2.00x -3%
ATL to win pays 2.59x fair 2.55x +2%
Mon Sep 14, 8:15 PM Δ1.6 · early season
DENatKC
new KC QB (Patrick Mahomes) · new DEN QB (Bo Nix) · divisional
Total no play
Over 43.5
model 43.6 · +0.1 vs line
Spread no play
DEN +3.0
model margin 1.4 · -1.6 vs line
Over 43.5 pays 1.97x fair 2.00x -2%
KC by over 2.5 pays 1.90x fair 1.96x -3%
KC to win pays 1.70x fair 1.70x -0%
Sun Sep 13, 1:00 PM Δ1.4 · early season
TBatCIN
Total no play
Over 50.5
model 51.9 · +1.4 vs line
Spread no play
CIN -3.5
model margin 4.5 · +1.0 vs line
Over 50.5 pays 2.01x fair 1.99x +1%
TB by over 3.5 pays 3.66x fair 3.49x +5%
TB to win pays 2.66x fair 2.59x +3%
Thu Sep 10, 8:35 PM Δ1.0 · early season
SFvsLA
SF crosses 6 time zones · 12,669 km trip · divisional
Total no play
Under 48.5
model 47.9 · -0.6 vs line
Spread no play
LA -3.5
model margin 4.5 · +1.0 vs line
SF by over 3.5 pays 3.80x fair 3.49x +9%
SF to win pays 2.66x fair 2.59x +3%
Sun Sep 13, 4:25 PM Δ1.0 · early season
MIAatLV
new LV QB (Kirk Cousins) · new MIA QB (Malik Willis) · MIA crosses 3 time zones · 3,493 km trip
Total no play
Over 41.5
model 42.5 · +1.0 vs line
Spread no play
MIA +3.5
model margin 3.3 · -0.2 vs line
Over 41.5 pays 2.01x fair 1.99x +1%
MIA by over 3.5 pays 3.52x fair 3.43x +3%
MIA to win pays 2.59x fair 2.55x +2%
Sun Sep 13, 4:25 PM Δ0.9 · early season
WASatPHI
new WAS QB (Jayden Daniels) · divisional
Total no play
Under 44.5
model 43.6 · -0.9 vs line
Spread no play
PHI -5.5
model margin 5.9 · +0.4 vs line
Over 44.5 pays 1.97x fair 2.01x -2%
WAS by over 5.5 pays 4.98x fair 5.20x -4%
WAS to win pays 2.98x fair 3.03x -1%
Sun Sep 13, 4:25 PM Δ0.6 · early season
ARIatLAC
Total no play
Over 47.5
model 48.1 · +0.6 vs line
Spread no play
LAC -9.5
model margin 10.1 · +0.6 vs line
ARI by over 9.5 pays 13.41x fair 14.84x -10%
ARI to win pays 4.98x fair 4.42x +13%

pays is what a Kalshi contract returns per unit risked after the taker fee. fair is what our model says it should return. Paying more than fair is the edge. The amber line under a matchup flags the situational facts the model weights and a glance at the line does not show: quarterback changes, rest mismatches, long trips, wind.

Cover rate by model disagreement — backtest

No games graded this season yet, so this is the walk-forward result over 1693 out-of-sample games from 2020–2025 — not live picks. Break-even is 51.75%.

|Model − line|SpreadsCoverTotalsCover
0-1 517 53.2% +0.7σ 488 49.6% -0.9σ
1-2 455 53.8% +0.9σ 420 51.0% -0.3σ
2-3 332 51.2% -0.2σ 320 50.3% -0.5σ
3-5 264 57.2% +1.8σ 340 48.8% -1.1σ
5-7 72 48.6% -0.5σ 83 55.4% +0.7σ

This report looked at many segments, so a bucket needs |σ| ≥ 3.11 — not 2 — before it means anything. No segment survives correction for the number of looks taken.

Model output, not advice. NFL spread and total markets are the most efficient in sports — every game is nationally televised, heavily bet and priced by people with better data than this. A public-data model beating those closing numbers is the exception, not the expectation. Bet only what you can afford to lose.