Duke · Week 4 analysis

Where the team stands

Duke ranks No. 39 in the reconstructed EP field, with a rating of +4.86 points against the average FBS opponent.

The previous snapshot ranked it No. 56; this update moves it +17 places and +3.30 rating points.

The rating snapshot includes 4 eligible games, with an eligible-results record of 4–0. The 24-hour kickoff embargo is applied consistently in historical forecast snapshots.

What enters this model

Opponent-adjusted current margin: +4.581. Prior-season margin: +10.742.

Opponent-adjusted first-down rate: +0.0240; EPA/play net: +0.0959; success-rate net: +0.0121; explosiveness net: +0.1070. These are adjusted net signals, not raw offensive rates.

EP inputs cover 2 eligible games. Excluded games contribute no fabricated EP observation. Opponent results can move a rating even during a bye.

The result and the reconstructed expectation

Duke beat William & Mary by 55 points. The pre-result reconstruction projected a margin of +37.03; the result was +17.97 points relative to that expectation.

This is a newly reconstructed forecast under the frozen EP model. It is not represented as a prediction published before the game, and it does not replace the original published grade.

Next scheduled test

The next scheduled opponent at this snapshot is Georgia Tech on 2026-10-10. That fixture is schedule information; its later outcome is not used in this report.

Offense: production and workload

Total yards: 638 versus 170; Net passing yards: 352 versus 153; Rushing yards: 286 versus 17; First downs: 28 versus 7

Completions–attempts: 24-34 versus 12-21; Rush attempts: 37 versus 29; Yards per pass: 10.4 versus 7.3; Yards per rush: 7.7 versus 0.6

Read production alongside workload: more attempts can raise a yardage total without raising efficiency. The model uses opponent-adjusted rate signals, not this box score alone.

Defense and possession

The opponent scored 7 points and recorded 170 total yards. This describes the conceded production; it does not identify a particular player or coverage as its cause.

Recorded sacks: 4 versus 0; Tackles for loss: 9 versus 2; Third downs: 8-13 versus 3-13; Fourth downs: 1-2 versus 0-1

Turnovers: 0 versus 0; Fumbles lost: 0 versus 0; Interceptions thrown: 0 versus 0; Penalties–yards: 6-60 versus 7-55; Possession time: 32:20 versus 27:40

The season outlook at this snapshot

The recorded regular-season record is 4–0. The expected final record is 8.2–3.8 over 12 scheduled games.

Expected wins add recorded results to the estimated win probabilities of remaining games. They are not a list of certain winners. Probabilities use a symmetric approximation from historical development residuals; independent calibration has not been established.

The comparison around the ranking

Nearby in this same snapshot: No. 38 Toledo (+5.07); No. 40 North Dakota State (+4.86).

PSR compares predictive strength rather than sorting teams by wins. A team can win and move down, or lose and move up, as its performance and opponents change the adjusted evidence.

How to read this assessment

This is a fresh assessment under the frozen EP model, using this period's eligible inputs. Earlier published releases remain preserved separately; the reconstructed archive is not presented as a prediction published before those games.

Scoring margin, prior-season margin, first-down efficiency, EPA per play, success rate and explosiveness contribute together. Correlated signals should not be interpreted as independent proof of an advantage, and a model disagreement with a betting line is not proof of profit.