A closing spread a half point off a key number hands the underdog side extra backdoor- cover value that a spread sitting exactly on the key number does not.
The tell: The closing spread is within half a point of (but not exactly on) 3, 7, 10, 14, or 17.
390-357 ATS (52.1%) · all games · n=747
· as of Sep 22, 2026
Beat the closing line: not captured yet
Drift not measured
By window
No graded window split yet.
Fires this week on
ATL @ GB · spread · away
BAL @ DAL · spread · home
CAR @ CLE · spread · home
CIN @ PIT · spread · home
HOU @ IND · spread · home
KC @ MIA · spread · home
LA @ DEN · spread · home
LAC @ BUF · spread · away
LV @ NO · spread · away
MIN @ TB · spread · home
NE @ JAX · spread · away
NYJ @ DET · spread · away
PHI @ CHI · spread · home
SEA @ WAS · spread · home
historical-context
Ats Streak Regression
A team on an extended ATS streak (either direction) is due for mean reversion; the streak itself carries no forward-looking edge once it reaches a genuine length.
The tell: Fires against the more extreme of atsStreakHome/atsStreakAway once |streak| >= 3 games: fades a covering streak, rides a fading streak's regression.
418-397 ATS (51.3%) · all games · n=815
· as of Sep 22, 2026
Beat the closing line: not captured yet
Drift not measured
By window
No graded window split yet.
Fires this week on
ARI @ SF · spread · away
ATL @ GB · spread · home
BAL @ DAL · spread · home
CAR @ CLE · spread · away
HOU @ IND · spread · away
KC @ MIA · spread · home
LA @ DEN · spread · away
LAC @ BUF · spread · away
LV @ NO · spread · away
NYJ @ DET · spread · away
PHI @ CHI · spread · away
SEA @ WAS · spread · home
TEN @ NYG · spread · away
model-evidence
Model Market Gap
the model panel's mean win probability and the closing moneyline price usually agree closely; a persistent gap of >=8 points between them is informative about which one has not fully absorbed the game's real information
The tell: Model panel P(home wins) diverges >=8pt from the moneyline-implied P(home wins), panel >= 5 models.
8-19 ATS (29.6%) · all games · n=27
· as of Sep 22, 2026
Beat the closing line: not captured yet
Drift not measured
By window
No graded window split yet.
Fires this week on
ARI @ SF · moneyline · away
ATL @ GB · moneyline · away
CAR @ CLE · moneyline · home
CIN @ PIT · moneyline · home
HOU @ IND · moneyline · home
KC @ MIA · moneyline · home
LAC @ BUF · moneyline · away
NE @ JAX · moneyline · away
NYJ @ DET · moneyline · away
PHI @ CHI · moneyline · home
SEA @ WAS · moneyline · home
TEN @ NYG · moneyline · away
market-price
Total Key Number
A total that crosses an NFL key total number (37/41/44/47/51) between open and close reprices a meaningfully different fraction of NFL game totals than one that never crosses.
The tell: The closing total has crossed 37, 41, 44, 47, or 51 since the open.
2-11 ATS (15.4%) · all games · n=13
· as of Sep 22, 2026
Beat the closing line: not captured yet
Drift not measured
By window
No graded window split yet.
Fires this week on
ARI @ SF · total · over
ATL @ GB · total · under
CAR @ CLE · total · over
CIN @ PIT · total · under
HOU @ IND · total · under
LV @ NO · total · over
MIN @ TB · total · under
NYJ @ DET · total · over
PHI @ CHI · total · under
SEA @ WAS · total · under
TEN @ NYG · total · under
market-price
Key Number Cross
A spread that crosses a key number (3 or 7) between open and close reprices a meaningfully different fraction of NFL margins than a spread that never crosses one.
The tell: The closing spread has crossed the 3- or 7-point key number since the open.
6-4 ATS (60.0%) · all games · n=10
· as of Sep 22, 2026
Beat the closing line: not captured yet
Drift not measured
By window
No graded window split yet.
Fires this week on
ATL @ GB · spread · away
BAL @ DAL · spread · away
CAR @ CLE · spread · away
CIN @ PIT · spread · away
HOU @ IND · spread · away
LA @ DEN · spread · home
LAC @ BUF · spread · home
NE @ JAX · spread · home
NYJ @ DET · spread · away
PHI @ CHI · spread · away
SEA @ WAS · spread · home
model-evidence
Model Edge
A sufficiently deep model panel with a decisive win-probability lean identifies a side.
The tell: At least 5 models and home win probability at least 8 points from 50%.
5-5 ATS (50.0%) · all games · n=10
· as of Sep 22, 2026
Beat the closing line: not captured yet
Drift not measured
By window
No graded window split yet.
Fires this week on
BAL @ DAL · spread · away
CAR @ CLE · spread · home
HOU @ IND · spread · home
KC @ MIA · spread · home
LV @ NO · spread · home
situational
Divisional Dog
Familiar divisional opponents reduce a favorite’s informational advantage.
The tell: Divisional spread has a non-pick'em underdog.
373-325 ATS (53.4%) · all games · n=698
· as of Sep 22, 2026
Beat the closing line: not captured yet
Drift not measured
By window
No graded window split yet.
Fires this week on
ARI @ SF · spread · away
CIN @ PIT · spread · home
HOU @ IND · spread · home
moneyline
Ml Divisional Dog
Division rivals know each other well, keeping divisional games close; a moneyline underdog the market still prices as competitive (not a near-pickem, not a longshot) covers that implied probability at an elevated rate.
The tell: Divisional game AND the moneyline dog's implied win probability is in [0.3, 0.45].
132-201 ATS (39.6%) · all games · n=333
· as of Sep 22, 2026
