the yahngorithm
← board
Stanford
ACC
@
Duke
ACC
Week 3 · 2026 · Sat, Sep 19, 3:00 PM CT · CW · Wallace Wade Stadium
Market spreadDUKE -9.5
Market total51
Model spreadDUKE -13.3
Edge+3.8

Spread models

SP+efficiencymodel DUKE -13.3
market DUKE -9.5  ·  edge +3.8 (home)  ·  STAN -5.8 vs DUKE +5 + 2.5 HFA
SRSscoring margin
no games played yet — empty in week 1, noisy through ~week 3
YahnSP+ + EPA + rostermodel DUKE -13.1
market DUKE -9.5  ·  edge +3.6 (home)  ·  STAN -6.3 vs DUKE +4.1 + 2.7 HFA

The spread signal is SP+ only right now. Market is the consensus of 9 books. Positive edge = model likes Duke. Guide §2.

How the Yahn number is built — not a separate model
SP+ base+ EPA+ roster= rating
STAN-5.8-0.50-6.3
DUKE+5-0.5-0.4+4.1

SP+ is the backbone. EPA nudges toward raw efficiency (weight grows through the season — near zero this early). Roster = talent + returning production + transfer-portal net, fading to zero by week 5. Venue home edge: 2.7 (2.7 base + altitude / hostile bump; SP+ and SRS use a flat 2.5). Weights un-calibrated.

Totals model

STAN expected
19.5
1.7 pts/poss
DUKE expected
32.8
2.9 pts/poss
Model total
52.3
market 51 · 22.8 poss, average pace
Edge
+1.3
under the 3.5-pt bar — noise

Only SP+ feeds this. Totals are noisier than spreads — the model's biggest UNDER edges cluster on blowouts and are mostly artifacts. Trust it most on games with a spread inside ~14. Guide §3.

Situational flags

letdownDukebeat Illinoishurts DUKE
Won an emotional game last week (rivalry or within 3 SP+ pts) and this week's opponent is 10+ SP+ pts weaker. Emotional hangover — fade this team / take the points.
off byeStanford14.8d resthelps STAN
A skipped week on the schedule (a true bye). Extra prep time — usually a small positive, especially for the better-coached team and for underdogs.
travelStanford2375 mi, 3hhurts STAN
Home stadium is 1,200+ miles from the venue, or the body-clock timezone shift is 2+ hours. A documented small drag on the traveling team.

Flags corroborate, they never drive a pick. A flag against a team the model already dislikes is a green light; against a team the model likes, it cancels out. Guide §5.

Team trends · 2026 season to date

STANDUKE
ATS overall1-12-0
ATS at home1-11-0
ATS on the road–1-0
ATS as favorite1-01-0
ATS as underdog0-11-0
ATS after a win0-11-0
ATS after a loss––
W–L at home1-11-0
W–L on the road–1-0
Over / Under2O–0U1O–1U

Amber = an outlier split (≥65% or ≤35% ATS on 8+ games). Trends are context, not a signal on their own.

Current lines

BookSpread (DUKE)Total
betmgm-9.5 -11051.5 -105
betonlineag-9.5 -10851 -110
betrivers-9.5 -11350.5 -114
betus-9.5 -11051 -110
bovada-9.5 -11051 -115
draftkings-9.5 -11251.5 -102
fanduel-9.5 -12051.5 -104
lowvig-9.5 -10451 -108
mybookieag-9.5 -11051 -110

Best over 50.5 · best under 51.5. Latest snapshot Sep 15, 8:01 AM. Odds shown where the source provides them (live pulls only). Spread is from DUKE's side.

Line movement

Showing the 5 most recent snapshots of 7.

Prediction market · Kalshi

STAN25%
DUKE75%
Implied lineDUKE -9.2
Volume3.5k
Open interest3.1k
Traded 24h847

Kalshi's market prices DUKE to win 75% — about DUKE -9.2 vs the book's DUKE -9.5. They agree. Volume and open interest are total contracts on the game (each settles at $1), not a per-side split — they tell you how much to trust the price. Guide §9.

Showing the 5 most recent of 12 snapshots.

Weather

Forecast for the venue at kickoff time (nearest hour). Each row is a snapshot — expect it to firm up as the game nears; the timestamp is when that snapshot was taken.

Showing the 3 most recent of 5 readings.

Sustained wind ≥ 15 mph is the one weather factor that reliably moves a total (UNDER). Everything else is minor. Guide §4.

Injuries

No impact-player injuries listed. ESPN's CFB feed is thin — an empty report means "unknown," not "clean." Guide §4.

Picks

PICKDUKE -9.5spread · sp_plus · logged at +4.3 edge
model 13.8logged 9.5close –CLV –
travel

Decision support, not a guarantee. Read the interpretation guide.