Skip to content

Always on: keeps you signed in and saves choices. Accept all helps us see which pages you use and fix errors.

Settings
$1 first week
CURRENT
▲-1 → 1 (+2) since open DAL @ GB GB -1 → +1 · Bovada — open DAL @ GB matchup▲6 → 5 (-1) since open BUF @ LV LV +6 → +5 · BetMGM · 2h ago — open BUF @ LV matchup▲-8.5 → -7.5 (+1) since open LAC @ KC KC -8.5 → -7.5 · Bovada — open LAC @ KC matchup
Access: Anonymous access. Content follows.

Angles

Every game has more than one truth.

Put the voices around one table, browse their records, compare their disagreements, or return to the angles you follow.

Angle families

73 angles across 11 families · updated as of 2026-10-18

19 angles are not yet mapped to a family: Hype Fade, Injury Whisper Confirm, Model Edge, Oline Whisper Fade, Player Backup Elevated, Player Buzz Vs Board, Player Funnel Defense, Player Pace Up Matchup, Player Prop Line Lag, Player Red Zone Role, Player Rest Travel, Player Snap Share Breakout, Player Target Air Yards Divergence, Player Weather Suppressed Passing, Primetime Dog, Sharp Vs Public, Starters Out, Travel Lag, Wind Under.

Angle lifecycle

73 angles tracked · gate as of 2026-10-18
  1. Surfaced mined from graded rows · n ≥ 100 · FDR-corrected 49 angles
  2. Candidate auto-written formula angle, or yours, or a pundit's calls 19 angles
  3. Backfill layer A · 2019–25 · every sport no automated stage yet
  4. Grading public, provisional · Feed only 1 angle
  5. Proven ≥1,000 · Brier ≤ .25 · lift > 0 → Module · on Tonight's why 0 angles
  6. Decaying drift → demoted, explained · retired at 60 d 0 angles

All angles by job

Graded record over all graded games. Early sample means fewer than 100 graded plays.

Market reads

AngleRecordGraded playsAs of
Line value421-376-8 ATS (52.8%)805Oct 18, 2:41 PM ET
Steam movenot graded· early sample77Oct 18, 2:41 PM ET
Public fade46-23-5 ATS (66.7%)· early sample74Oct 18, 2:41 PM ET
Contrarian31-24-5 ATS (56.4%)· early sample60Oct 18, 2:41 PM ET
Sharp24-31-5 ATS (43.6%)· early sample60Oct 18, 2:41 PM ET
Splits divergence21-26-4 ATS (44.7%)· early sample51Oct 18, 2:41 PM ET
Total key number26-23 O/U (53.1%)· early sample49Oct 18, 2:41 PM ET
Key number cross20-22-5 ATS (47.6%)· early sample47Oct 18, 2:41 PM ET
Favorite traprecord not shown· early sample30Oct 18, 2:41 PM ET
Reverse line move8-3-2 ATS (72.7%)· early sample13Oct 18, 2:41 PM ET
Ml value dogrecord not shown· early sample10Oct 18, 2:41 PM ET

Model reads

AngleRecordGraded playsAs of
Model market gaprecord not shown· early sample61Oct 18, 2:41 PM ET
Modeler30-25-5 ATS (54.5%)· early sample60Oct 18, 2:41 PM ET
Ensemble disagreement5-21-2 ATS (19.2%)· early sample28Oct 18, 2:41 PM ET
Model conviction pickem0-1 ATS (0.0%)· early sample1Oct 18, 2:41 PM ET

Situation reads

AngleRecordGraded playsAs of
Ats streak regression442-411-24 ATS (51.8%)877Oct 18, 2:41 PM ET
Divisional dog386-322-13 ATS (54.5%)721Oct 18, 2:41 PM ET
Prime time under240-197-7 O/U (54.9%)444Oct 18, 2:41 PM ET
Rest edge217-208-10 ATS (51.1%)435Oct 18, 2:41 PM ET
Ml divisional dogrecord not shown342Oct 18, 2:41 PM ET
Home dog primetime95-76-3 ATS (55.6%)174Oct 18, 2:41 PM ET
Divisional primetime combo86-63-3 ATS (57.7%)152Oct 18, 2:41 PM ET
Short week both sides total66-62 O/U (51.6%)128Oct 18, 2:41 PM ET
Cold weather late season total40-44-2 O/U (47.6%)· early sample86Oct 18, 2:41 PM ET
Short week fade1-2 ATS (33.3%)· early sample3Oct 18, 2:41 PM ET

Roster reads

AngleRecordGraded playsAs of
Qb out51-59-3 ATS (46.4%)113Oct 18, 2:41 PM ET
Combined injury attrition total1-2 O/U (33.3%)· early sample3Oct 18, 2:41 PM ET
Injury cluster1-2 ATS (33.3%)· early sample3Oct 18, 2:41 PM ET

Voices

AngleRecordGraded playsAs of
Casual bettor7-8-1 ATS (46.7%)· early sample16Oct 18, 2:41 PM ET

Other angles

AngleRecordGraded playsAs of
Primetime dog238-198-8 ATS (54.6%)444Oct 18, 2:41 PM ET
Wind under78-64-2 O/U (54.9%)144Oct 18, 2:41 PM ET
Model edge17-29-6 ATS (37.0%)· early sample52Oct 18, 2:41 PM ET
Sharp vs public21-26-4 ATS (44.7%)· early sample51Oct 18, 2:41 PM ET
Starters out3-4 ATS (42.9%)· early sample7Oct 18, 2:41 PM ET
Travel lag2-0 ATS (100.0%)· early sample2Oct 18, 2:41 PM ET

Angle index

12 of 12 angles

Ranked by trailing-300 lift

  1. #1 · Divisional dog · n=300 · lift vs other lenses +3.1pt (CI -2.8pt to +8.8pt)
  2. #2 · Prime time under · n=300 · lift vs other lenses +2.9pt (CI -3.0pt to +8.7pt)
  3. #3 · Line value · n=300 · lift vs other lenses +2.2pt (CI -3.7pt to +8.1pt)
  4. #4 · Rest edge · n=300 · lift vs other lenses +1.9pt (CI -4.1pt to +7.7pt)
  5. #5 · Primetime dog · n=300 · lift vs other lenses +1.6pt (CI -4.3pt to +7.5pt)
  6. #6 · Ats streak regression · n=300 · lift vs other lenses -0.0pt (CI -5.9pt to +5.9pt)
  7. #7 · Ml divisional dog · n=300 · lift vs other lenses -11.0pt (CI -16.7pt to -5.2pt)
  8. unranked · Divisional primetime combo · n=152 · lift vs other lenses +6.6pt (CI -1.6pt to +14.4pt)
  9. unranked · Home dog primetime · n=174 · lift vs other lenses +4.4pt (CI -3.3pt to +11.8pt)
  10. unranked · Wind under · n=144 · lift vs other lenses +3.7pt (CI -4.7pt to +11.8pt)
  11. unranked · Short week both sides total · n=128 · lift vs other lenses +0.2pt (CI -8.5pt to +8.8pt)
  12. unranked · Qb out · n=113 · lift vs other lenses -5.2pt (CI -14.3pt to +4.3pt)
Not rated yet moneyline
Graded
unavailable
ATS record
unavailable
Lift vs other lenses
unavailable
Beat the closing line
unavailable
Brier (t300)
0.242
Rating
Not rated yet
As of
Oct 18, 6:25 PM ET

Last results · newest first

No graded results are shown for this angle.

Use in Studio

Discovery · find a way of seeing

The same discovery surface as the Bench library: search, facets, "for your model", trending, by record.

ATS spread break-even at -110 is 52.4%; every rate below names its graded n.

showing 12 of 12 · as of Oct 18, 6:25 PM ET
Sort by

Ranked by ATS hit rate

Graded results · n=10 graded lenses matching current facets · as of Oct 18, 6:25 PM ET

  • Divisional primetime combo

    situational · CANDIDATE

    57.7% ATS spread · n=152

    +6.6pt lift · CLV +0.25

  • Home dog primetime

    narrative · CANDIDATE

    55.6% ATS spread · n=174

    +4.4pt lift · CLV +0.55

  • Wind under

    situational · CANDIDATE

    54.9% ATS spread · n=144

    +3.7pt lift · CLV not graded

  • Divisional dog

    situational · CANDIDATE

    54.2% ATS spread · n=300

    +3.1pt lift · CLV +0.13

  • Prime time under

    narrative · CANDIDATE

    54.1% ATS spread · n=300

    +2.9pt lift · CLV -0.22

  • Line value

    market-price · CANDIDATE

    53.4% ATS spread · n=300

    +2.2pt lift · CLV +0.08

  • Rest edge

    situational · CANDIDATE

    53.1% ATS spread · n=300

    +1.9pt lift · CLV +0.08

  • Primetime dog

    situational · CANDIDATE

    52.9% ATS spread · n=300

    +1.6pt lift · CLV -0.06

  • Short week both sides total

    situational · CANDIDATE

    51.6% ATS spread · n=128

    +0.2pt lift · CLV -0.33

  • Ats streak regression

    historical-context · CANDIDATE

    51.4% ATS spread · n=300

    -0.0pt lift · CLV +0.52

ATS hit-rate board

Graded results · n=12 graded lenses · as of Oct 18, 6:25 PM ET

0.0%11.5%23.1%34.6%46.2%57.7%Divisional primetime combo: 57.7%Divisional primetime comboDivisional primetime…Home dog primetime: 55.6%Home dog primetimeHome dog primetimeWind under: 54.9%Wind underWind underDivisional dog: 54.2%Divisional dogDivisional dogPrime time under: 54.1%Prime time underPrime time underLine value: 53.4%Line valueLine valueRest edge: 53.1%Rest edgeRest edgePrimetime dog: 52.9%Primetime dogPrimetime dogShort week both sides total: 51.6%Short week both sides totalShort week both side…Ats streak regression: 51.4%Ats streak regressionAts streak regression

Sample size vs lift

Graded results · n=12 lift rows · as of Oct 18, 6:25 PM ET

Does the edge hold with sample?Each dot is one graded lens-12.1pt-8.1pt-4.2pt-0.3pt+3.7pt+7.6pt95.0139.6184.2228.8273.4318.0Divisional primetime combo: graded n152.0, lift vs break-even+6.6ptDivisional primetime comboHome dog primetime: graded n174.0, lift vs break-even+4.4ptHome dog primetimeWind under: graded n144.0, lift vs break-even+3.7ptWind underDivisional dog: graded n300.0, lift vs break-even+3.1ptDivisional dogPrime time under: graded n300.0, lift vs break-even+2.9ptPrime time underLine value: graded n300.0, lift vs break-even+2.2ptLine valueRest edge: graded n300.0, lift vs break-even+1.9ptRest edgePrimetime dog: graded n300.0, lift vs break-even+1.6ptPrimetime dogShort week both sides total: graded n128.0, lift vs break-even+0.2ptShort week both sides totalAts streak regression: graded n300.0, lift vs break-even-0.0ptAts streak regressionQb out: graded n113.0, lift vs break-even-5.2ptQb outMl divisional dog: graded n300.0, lift vs break-even-11.0ptMl divisional dogLIFT VS BREAK-EVENGRADED NGraded results · as of Oct 18, 6:25 PM ET
LensRatingRecord (ATS spread)Lift (ATS spread)CLVas-of
Ats streak regressionCANDIDATE51.4% ATS spread · n=300-0.0pt · n=300+0.52 · n=300Oct 18, 6:25 PM ET
Divisional dogCANDIDATE54.2% ATS spread · n=300+3.1pt · n=300+0.13 · n=300Oct 18, 6:25 PM ET
Line valueCANDIDATE53.4% ATS spread · n=300+2.2pt · n=300+0.08 · n=300Oct 18, 6:25 PM ET
Ml divisional dogGRADING41.1% ATS spread · n=300-11.0pt · n=300+0.01 · n=300Oct 18, 6:25 PM ET
Prime time underCANDIDATE54.1% ATS spread · n=300+2.9pt · n=300-0.22 · n=300Oct 18, 6:25 PM ET
Primetime dogCANDIDATE52.9% ATS spread · n=300+1.6pt · n=300-0.06 · n=300Oct 18, 6:25 PM ET
Rest edgeCANDIDATE53.1% ATS spread · n=300+1.9pt · n=300+0.08 · n=300Oct 18, 6:25 PM ET
Home dog primetimeCANDIDATE55.6% ATS spread · n=174+4.4pt · n=174+0.55 · n=174Oct 18, 6:25 PM ET
Divisional primetime comboCANDIDATE57.7% ATS spread · n=152+6.6pt · n=152+0.25 · n=152Oct 18, 6:25 PM ET
Wind underCANDIDATE54.9% ATS spread · n=144+3.7pt · n=144not graded yet · n=144Oct 18, 6:25 PM ET
Short week both sides totalCANDIDATE51.6% ATS spread · n=128+0.2pt · n=128-0.33 · n=128Oct 18, 6:25 PM ET
Qb outCANDIDATE46.4% ATS spread · n=113-5.2pt · n=113+0.25 · n=113Oct 18, 6:25 PM ET

Graded angle records

narrative

Home Dog Primetime

House angle of the week

A home underdog in a primetime spot draws a spirited crowd and extra motivation, covering at a rate the closing line under-prices.

The tell: Fires home to cover whenever the game is primetime and spreadClose < 0 (home is the underdog, nflverse sign convention).

95-76-3 ATS (55.6%) · all games · n=174 · as of Oct 18, 2026

vs break-even: +3.2 pts at −110

Beat the closing line: +54.5 pts

Drift not measured

By window

WindowRecordnAs of
1-953.8% hitn=80as of Oct 18, 2026
10-1853.3% hitn=75as of Oct 18, 2026
playoffs72.7% hitn=11as of Oct 18, 2026

Fires this week on

  • DAL @ GB · spread · home

market-price

Steam Move

A large net line move between open and close — spread points, total points, or moneyline implied-probability shift — reflects sharp money and predicts value on the side the price moved toward.

The tell: Spread moved >=1.5pt, total moved >=2.5pt, or moneyline implied win probability moved >=3pt.

54.8% hit · all games · n=77 · as of Oct 18, 2026

Beat the closing line: −16.6 pts

Drift not measured

By window

WindowRecordnAs of
1-953.4% hitn=58as of Oct 18, 2026

Fires this week on

  • BUF @ LV · spread · home
  • BUF @ LV · total · over
  • DAL @ GB · spread · away
  • LAC @ KC · spread · home
  • LAC @ KC · total · under
  • WAS @ SF · spread · home

market-price

Line Value

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.

421-376-8 ATS (52.8%) · all games · n=805 · as of Oct 18, 2026

vs break-even: +0.4 pts at −110

Beat the closing line: +8.1 pts

Drift not measured

By window

WindowRecordnAs of
1-955.1% hitn=410as of Oct 18, 2026
10-1849.0% hitn=351as of Oct 18, 2026
playoffs56.3% hitn=32as of Oct 18, 2026

Fires this week on

  • BUF @ LV · spread · home
  • DAL @ GB · spread · home
  • LAC @ KC · spread · away
  • WAS @ SF · 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.

all games · n=61 · as of Oct 18, 2026

Beat the closing line: +0.9 pts

Drift not measured

By window

WindowRecordnAs of
1-934.8% hitn=66as of Oct 18, 2026

Fires this week on

  • BUF @ LV · moneyline · home
  • DAL @ GB · moneyline · away
  • LAC @ KC · moneyline · away
  • WAS @ SF · 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.

26-23 O/U (53.1%) · all games · n=49 · as of Oct 18, 2026

vs break-even: +0.7 pts at −110

Beat the closing line: −13.2 pts

Drift not measured

By window

WindowRecordnAs of
1-954.5% hitn=55as of Oct 18, 2026

Fires this week on

  • BUF @ LV · total · over
  • DAL @ GB · total · under
  • LAC @ KC · total · under
  • WAS @ SF · total · over

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.

442-411-24 ATS (51.8%) · all games · n=877 · as of Oct 18, 2026

vs break-even: −0.6 pts at −110

Beat the closing line: +51.6 pts

Drift not measured

By window

WindowRecordnAs of
1-954.4% hitn=417as of Oct 18, 2026
10-1848.8% hitn=377as of Oct 18, 2026
playoffs47.7% hitn=44as of Oct 18, 2026

Fires this week on

  • BUF @ LV · spread · away
  • DAL @ GB · spread · home
  • LAC @ KC · spread · away

market-sentiment

Public Fade

The betting public over-bets lopsided favorites and trendy sides; fading a lopsided ticket majority captures value the crowd is pricing in emotionally rather than analytically.

The tell: Spread tickets split >=65/35 either way.

46-23-5 ATS (66.7%) · all games · n=74 · as of Oct 18, 2026

vs break-even: +14.3 pts at −110

Beat the closing line: +6.7 pts

Drift not measured

By window

WindowRecordnAs of
1-968.1% hitn=47as of Oct 18, 2026

Fires this week on

  • BUF @ LV · spread · home
  • LAC @ KC · spread · away
  • WAS @ SF · spread · away

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%.

17-29-6 ATS (37.0%) · all games · n=52 · as of Oct 18, 2026

vs break-even: −15.4 pts at −110

Beat the closing line: −4.5 pts

Drift not measured

By window

WindowRecordnAs of
1-940.0% hitn=30as of Oct 18, 2026

Fires this week on

  • BUF @ LV · spread · away
  • LAC @ KC · spread · home
  • WAS @ SF · spread · home

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.

20-22-5 ATS (47.6%) · all games · n=47 · as of Oct 18, 2026

vs break-even: −4.8 pts at −110

Beat the closing line: −20.5 pts

Drift not measured

By window

WindowRecordnAs of
1-948.9% hitn=47as of Oct 18, 2026

Fires this week on

  • DAL @ GB · spread · away
  • LAC @ KC · spread · home
  • WAS @ SF · spread · home

narrative

Prime Time Under

Primetime broadcasts skew toward conservative game plans and public over-money inflates the total, so the under carries a live edge.

The tell: Fires the under whenever the schedule marks this game primetime (Thu/Mon, or Sun >= 18:00 local).

240-197-7 O/U (54.9%) · all games · n=444 · as of Oct 18, 2026

vs break-even: +2.5 pts at −110

Beat the closing line: −22.2 pts

Drift not measured

By window

WindowRecordnAs of
1-957.9% hitn=216as of Oct 18, 2026
10-1852.5% hitn=181as of Oct 18, 2026
playoffs53.1% hitn=32as of Oct 18, 2026

Fires this week on

  • DAL @ GB · total · under
  • WAS @ SF · total · under

situational

Primetime Dog

Primetime attention can overprice favorites, leaving value on the underdog.

The tell: Primetime spread has a non-pick'em underdog.

238-198-8 ATS (54.6%) · all games · n=444 · as of Oct 18, 2026

vs break-even: +2.2 pts at −110

Beat the closing line: −5.6 pts

Drift not measured

By window

WindowRecordnAs of
1-954.7% hitn=214as of Oct 18, 2026
10-1853.6% hitn=181as of Oct 18, 2026
playoffs51.5% hitn=33as of Oct 18, 2026

Fires this week on

  • DAL @ GB · spread · away
  • WAS @ SF · spread · away

situational

Rest Edge

Rest advantage of at least three days identifies the better-rested side.

The tell: Absolute rest edge is at least 3 days.

217-208-10 ATS (51.1%) · all games · n=435 · as of Oct 18, 2026

vs break-even: −1.3 pts at −110

Beat the closing line: +8.3 pts

Drift not measured

By window

WindowRecordnAs of
1-951.9% hitn=185as of Oct 18, 2026
10-1850.0% hitn=222as of Oct 18, 2026
playoffs50.0% hitn=16as of Oct 18, 2026

Fires this week on

  • DAL @ GB · spread · away
  • LAC @ KC · spread · home

market-sentiment

Sharp Vs Public

A large gap between money and ticket shares identifies the side receiving larger wagers.

The tell: Home money share differs from home ticket share by at least 15 points.

21-26-4 ATS (44.7%) · all games · n=51 · as of Oct 18, 2026

vs break-even: −7.7 pts at −110

Beat the closing line: −6.6 pts

Drift not measured

By window

WindowRecordnAs of
1-941.2% hitn=34as of Oct 18, 2026

Fires this week on

  • BUF @ LV · spread · home
  • WAS @ SF · spread · home

market-sentiment

Splits Divergence

When the money-percentage on a side meaningfully exceeds its ticket-percentage, a few large bets are outweighing many small ones — the more reliable read on professional backing.

The tell: Money% on a side leads ticket% on that side by >=15pt.

21-26-4 ATS (44.7%) · all games · n=51 · as of Oct 18, 2026

vs break-even: −7.7 pts at −110

Beat the closing line: −6.6 pts

Drift not measured

By window

WindowRecordnAs of
1-941.2% hitn=34as of Oct 18, 2026

Fires this week on

  • BUF @ LV · spread · home
  • WAS @ SF · spread · home

situational

Divisional Dog

Familiar divisional opponents reduce a favorite’s informational advantage.

The tell: Divisional spread has a non-pick'em underdog.

386-322-13 ATS (54.5%) · all games · n=721 · as of Oct 18, 2026

vs break-even: +2.1 pts at −110

Beat the closing line: +11.5 pts

Drift not measured

By window

WindowRecordnAs of
1-955.7% hitn=291as of Oct 18, 2026
10-1853.8% hitn=394as of Oct 18, 2026
playoffs46.7% hitn=15as of Oct 18, 2026

Fires this week on

  • LAC @ KC · spread · away

model-evidence

Ensemble Disagreement

A deep, tightly clustered model panel makes its consensus direction more credible.

The tell: At least 5 models, disagreement at most 0.10, and probability at least 5 points from 50%.

5-21-2 ATS (19.2%) · all games · n=28 · as of Oct 18, 2026

vs break-even: −33.2 pts at −110

Beat the closing line: −12.0 pts

Drift not measured

By window

No graded window split yet.

Fires this week on

  • DAL @ GB · spread · away

market-price

Reverse Line Move

When the ticket majority leans one way but the closing line moves the other way, the money behind the move (not the ticket count) is the better read on which side is live.

The tell: Public tickets lean >=10pt one way while the spread closes >=1pt toward the other side.

8-3-2 ATS (72.7%) · all games · n=13 · as of Oct 18, 2026

vs break-even: +20.3 pts at −110

Beat the closing line: +0.0 pts

Drift not measured

By window

WindowRecordnAs of
1-978.6% hitn=14as of Oct 18, 2026

Fires this week on

  • BUF @ LV · 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].

all games · n=342 · as of Oct 18, 2026

Beat the closing line: +1.1 pts

Drift not measured

By window

WindowRecordnAs of
1-940.7% hitn=150as of Oct 18, 2026
10-1841.2% hitn=187as of Oct 18, 2026
playoffs33.3% hitn=6as of Oct 18, 2026

Fires this week on

No current fire for this angle.

situational

Divisional Primetime Combo

divisional familiarity and primetime spotlight each independently suppress blowouts; a game with both factors present has two independent reasons to stay closer than the spread implies, not one

The tell: Divisional AND primetime both true — the underdog side, a combined familiarity+spotlight factor.

86-63-3 ATS (57.7%) · all games · n=152 · as of Oct 18, 2026

vs break-even: +5.3 pts at −110

Beat the closing line: +25.0 pts

Drift not measured

By window

WindowRecordnAs of
1-956.5% hitn=62as of Oct 18, 2026
10-1857.5% hitn=80as of Oct 18, 2026
playoffs66.7% hitn=6as of Oct 18, 2026

Fires this week on

No current fire for this angle.

situational

Wind Under

outdoor regular-season games with stored wind of at least 15 mph suppress combined scoring

The tell: nfl_schedules wind >= 15 mph — frozen WATCH hypothesis; missing weather is not zero wind.

78-64-2 O/U (54.9%) · all games · n=144 · as of Oct 18, 2026

vs break-even: +2.5 pts at −110

Beat the closing line: not captured yet

Drift not measured

By window

WindowRecordnAs of
1-943.1% hitn=58as of Oct 18, 2026
10-1862.8% hitn=78as of Oct 18, 2026
playoffs71.4% hitn=7as of Oct 18, 2026

Fires this week on

No current fire for this angle.

situational

Short Week Both Sides Total

a short turnaround degrades practice time and execution regardless of which team has the smaller relative edge; when both sides are equally short-rested there is no differential to exploit but there is a real absolute effect on combined scoring

The tell: Both restDaysHome and restDaysAway <= 5 — an absolute fatigue effect, not a rest-differential edge.

66-62 O/U (51.6%) · all games · n=128 · as of Oct 18, 2026

vs break-even: −0.8 pts at −110

Beat the closing line: −33.3 pts

Drift not measured

By window

WindowRecordnAs of
1-951.7% hitn=60as of Oct 18, 2026
10-1850.8% hitn=65as of Oct 18, 2026

Fires this week on

No current fire for this angle.

availability

Qb Out

A team missing its starting QB has a materially different true win probability than the closing line has fully priced in; fade the QB-out team.

The tell: Fires against whichever team has homeQbOut/awayQbOut = true when the opponent does not.

51-59-3 ATS (46.4%) · all games · n=113 · as of Oct 18, 2026

vs break-even: −6.0 pts at −110

Beat the closing line: +25.0 pts

Drift not measured

By window

WindowRecordnAs of
1-948.0% hitn=50as of Oct 18, 2026
10-1849.1% hitn=57as of Oct 18, 2026
playoffs33.3% hitn=3as of Oct 18, 2026

Fires this week on

No current fire for this angle.

situational

Cold Weather Late Season Total

cold-weather outdoor games suppress passing efficiency and combined scoring; the effect is most reliable late in the season, away from early-season cold outliers and dome-scheduling confounds

The tell: tempF <= 32 in week >= 10 — a scoring-suppression factor distinct from wind or precipitation.

40-44-2 O/U (47.6%) · all games · n=86 · as of Oct 18, 2026

vs break-even: −4.8 pts at −110

Beat the closing line: not captured yet

Drift not measured

By window

WindowRecordnAs of
10-1846.8% hitn=62as of Oct 18, 2026
playoffs50.0% hitn=22as of Oct 18, 2026

Fires this week on

No current fire for this angle.

market-price

Contrarian

Popular sides become fade candidates when price outruns fair value.

The tell: Public pressure and the unpopular side's price.

31-24-5 ATS (56.4%) · all games · n=60 · as of Oct 18, 2026

vs break-even: +4.0 pts at −110

Beat the closing line: not captured yet

Drift not measured

By window

WindowRecordnAs of
1-971.4% hitn=7as of Oct 18, 2026

Fires this week on

No current fire for this angle.

model-evidence

Modeler

Model features and uncertainty explain repeatable edge.

The tell: Feature contribution, confidence, and ensemble disagreement.

30-25-5 ATS (54.5%) · all games · n=60 · as of Oct 18, 2026

vs break-even: +2.2 pts at −110

Beat the closing line: not captured yet

Drift not measured

By window

WindowRecordnAs of
1-952.7% hitn=74as of Oct 18, 2026

Fires this week on

No current fire for this angle.

market-price

Sharp

Line value and market movement identify mispriced sides.

The tell: Key numbers, closing-line value, and price movement.

24-31-5 ATS (43.6%) · all games · n=60 · as of Oct 18, 2026

vs break-even: −8.7 pts at −110

Beat the closing line: not captured yet

Drift not measured

By window

WindowRecordnAs of
1-938.5% hitn=13as of Oct 18, 2026

Fires this week on

No current fire for this angle.

moneyline

Favorite Trap

The market overvalues a heavy moneyline favorite that is rest-disadvantaged against its opponent — on a short week itself, or facing a dog off extra rest — because the price reacts to the teams' talent gap but not fully to the schedule context; fade the favorite.

The tell: Moneyline favorite implied probability >= 0.7 AND (the favorite is on a short week the dog is not, OR the dog is off a bye the favorite is not).

all games · n=30 · as of Oct 18, 2026

Beat the closing line: not captured yet

Drift not measured

By window

WindowRecordnAs of
1-933.3% hitn=9as of Oct 18, 2026
10-1825.0% hitn=20as of Oct 18, 2026

Fires this week on

No current fire for this angle.

market-sentiment

Casual Bettor

Public favorites and overs expose a measurable crowd lean.

The tell: The labeled public side of the market.

7-8-1 ATS (46.7%) · all games · n=16 · as of Oct 18, 2026

vs break-even: −5.7 pts at −110

Beat the closing line: not captured yet

Drift not measured

By window

WindowRecordnAs of
1-941.9% hitn=74as of Oct 18, 2026

Fires this week on

No current fire for this angle.

moneyline

Ml Value Dog

The spread market and the moneyline market usually price the same dog at close to the same implied win probability; when the moneyline market pays the dog a materially higher probability than the spread's own line implies, that gap is mispricing worth taking.

The tell: Moneyline-implied dog win probability exceeds the spread-implied dog win probability by >= 5 points.

all games · n=10 · as of Oct 18, 2026

Beat the closing line: −2.4 pts

Drift not measured

By window

WindowRecordnAs of
1-90.0% hitn=9as of Oct 18, 2026
10-18100.0% hitn=4as of Oct 18, 2026

Fires this week on

No current fire for this angle.

situational

Starters Out

A material starter-availability imbalance favors the healthier team.

The tell: One side has at least two more starters out.

3-4 ATS (42.9%) · all games · n=7 · as of Oct 18, 2026

vs break-even: −9.5 pts at −110

Beat the closing line: not captured yet

Drift not measured

By window

WindowRecordnAs of
1-950.0% hitn=6as of Oct 18, 2026
10-180.0% hitn=1as of Oct 18, 2026

Fires this week on

No current fire for this angle.

situational

Combined Injury Attrition Total

offensive execution and combined scoring degrade with meaningful starter absences; when both teams are shorthanded at once the effect compounds on the total rather than cancelling out

The tell: Combined starters out across both teams >= 4 — aggregate talent suppression, not a one-sided edge.

1-2 O/U (33.3%) · all games · n=3 · as of Oct 18, 2026

vs break-even: −19.0 pts at −110

Beat the closing line: not captured yet

Drift not measured

By window

WindowRecordnAs of
1-933.3% hitn=3as of Oct 18, 2026

Fires this week on

No current fire for this angle.

availability

Injury Cluster

A team with several starters ruled out beyond just the QB is meaningfully short-handed in ways a single-injury view misses; fade the shorthanded team.

The tell: Fires against whichever team has >= 3 starters out when the opponent does not.

1-2 ATS (33.3%) · all games · n=3 · as of Oct 18, 2026

vs break-even: −19.0 pts at −110

Beat the closing line: not captured yet

Drift not measured

By window

WindowRecordnAs of
1-933.3% hitn=3as of Oct 18, 2026

Fires this week on

No current fire for this angle.

historical-context

Short Week Fade

A team playing on a short week has historically underperformed its closing line relative to a normally-rested opponent; fade the short-week team.

The tell: Fires against whichever team has <= 5 rest days when the opponent does not.

1-2 ATS (33.3%) · all games · n=3 · as of Oct 18, 2026

vs break-even: −19.0 pts at −110

Beat the closing line: not captured yet

Drift not measured

By window

WindowRecordnAs of
1-90.0% hitn=1as of Oct 18, 2026
10-1850.0% hitn=2as of Oct 18, 2026

Fires this week on

No current fire for this angle.

situational

Travel Lag

Long away travel can impair preparation and favors the home side.

The tell: Away travel is at least 2,000 miles.

2-0 ATS (100.0%) · all games · n=2 · as of Oct 18, 2026

vs break-even: +47.6 pts at −110

Beat the closing line: +0.0 pts

Drift not measured

By window

WindowRecordnAs of
1-9100.0% hitn=1as of Oct 18, 2026

Fires this week on

No current fire for this angle.

model-evidence

Model Conviction Pickem

a market spread near zero says the market thinks the game is close to a coinflip; a model panel that is simultaneously confident (conviction >=0.12) and internally agreed (disagreement <=0.15) is making the opposite statement, and that has not been priced the way it has been for lopsided spreads

The tell: Closing spread within 3pt of pick'em while the model panel (>=5 models) leans >=0.12 off a coinflip with <=0.15 disagreement.

0-1 ATS (0.0%) · all games · n=1 · as of Oct 18, 2026

vs break-even: −52.4 pts at −110

Beat the closing line: +0.0 pts

Drift not measured

By window

WindowRecordnAs of
1-90.0% hitn=1as of Oct 18, 2026

Fires this week on

No current fire for this angle.

Decaying

Angles whose recent record has slipped.

0 of 12 angles are marked Decaying

Grading

Angles still building a record; it stays grey until 1,000 graded plays.

1 of 12 angles are marked Grading· largest graded n 300 of the 1,000 target