What the error numbers say
Across FBS-versus-FBS games, the Matrix missed the final scoring margin by 8.98 points on average. Its average team-score error was 8.74 points, while the projected total missed the actual combined score by 14.77. The full 75-game slate—including cross-division games—finished at 11.84 margin MAE, 9.47 team-score MAE and 13.84 total MAE.
FBS margin accuracy
The 8.98-point FBS margin MAE is the clearest Week 3 indicator that the model described competitive FBS games more closely than the broader slate.
Totals were noisier
The full FBS model-total comparison went 27–30 against the stored line. Week 3 does not support claiming a broad totals advantage.
Market comparison vs published signals
Every FBS matchup can be compared with the stored DraftKings lines as a diagnostic. That full-board side comparison finished 35–21–1. The published Signal Board was much smaller: three sides and six totals selected and pregame before kickoff.
| Layer | Scope | Record | Meaning |
|---|---|---|---|
| Model side vs market | 57 FBS games | 35–21–1 | Diagnostic full-board comparison |
| Model total vs market | 57 FBS games | 27–30 | No Week 3 broad-market edge |
| Published sides | 3 pregame plays | 3–0 | Published before kickoff |
| Published totals | 6 pregame plays | 3–3 | Published before kickoff |
The nine pregame Week 3 signals
| Game | pregame signal | Final | Grade |
|---|---|---|---|
| Ball State at Liberty | Liberty -14 | 15–51 | W |
| UConn at Southern Miss | UConn -3 | 48–20 | W |
| Miami (OH) at Cincinnati | Miami (OH) +16.5 | 31–35 | W |
| NC State at Vanderbilt | OVER 50.5 | 31–35 | W |
| Marshall at Missouri State | OVER 52.5 | 30–24 | W |
| Western Kentucky at Indiana | UNDER 60.5 | 0–38 | W |
| Florida at Auburn | UNDER 53.5 | 44–39 | L |
| SMU at Louisville | UNDER 59.5 | 31–41 | L |
| Mississippi State at South Carolina | UNDER 58.5 | 41–34 | L |
Upset Watch accountability
Week 2’s six formal upset alerts finished 3–3 outright and 5–1 ATS. Ohio State and Jacksonville State missed outright by one and two points. Week 3 had no formal Upset Watch, which is recorded as a process failure; two of the six closest Matrix games—Temple–Toledo and Wyoming–Central Michigan—produced outright upsets.
| Week 2 watch team | Stored line | Final | Outright | ATS |
|---|---|---|---|---|
| Ohio State | +1.5 at Texas | 23–24 | L | W |
| Jacksonville State | +2.5 at Ohio | 27–29 | L | W |
| Michigan | +5.5 vs Oklahoma | 17–10 | W | W |
| California | +3.5 at Syracuse | 21–18 | W | W |
| UTSA | +2.5 at Texas State | 31–26 | W | W |
| Air Force | +2.5 vs North Dakota State | 32–38 | L | L |
What changes next
Error review is evidence, not an automatic ratings formula. The Week 3 review still required advanced efficiency, opponent-quality, turnover, quarterback, personnel and multi-week corroboration gates. Thirty-one teams received controlled MPR moves; every earlier pregame projection remains preserved.