Matrix Power Rating
Neutral-field underlying strength. MPR Rank is the primary GameDay Matrix national power ranking.
What matters first
Measure the miss.
Picking a heavy favorite straight up is not the lead performance claim. Forecast error, market-relative comparison and the record of explicitly published signals come first. Winner accuracy remains visible as a calibration check.
Model Performance
Week 8 Model Scorecard
Every completed final is compared with its pregame Matrix projection. The latest completed weekly closeout is shown below. Forecasts are preserved exactly as published before kickoff. Earlier weeks remain in Results.
FBS margin MAE12.18 pts56 Week 8 FBS games
Team-score MAE9.87 pts112 frozen FBS team scores
FBS total MAE15.79 pts56 FBS versus FBS games
Recorded spread comparison22–3456 dated public observations
Published Week 9 forecasts56 gamesOctober 26 model update
Full-slate winners35–21All 56 Week 8 games
Market-basis rule · Week 8 comparisons use the recorded per-game dated public quotes. The public edition combines original October 19 quotes for 12 earlier games and October 23 ET quotes for 44 Saturday games. These are dated observations, not verified opening or closing lines. All-game margin MAE is 12.18; the matched 56-game FBS margin MAE is 12.18 versus market 10.63. Three frozen challengers are evaluated separately; none is promoted from one week.
How the pieces connect
Strength → Schedule → Wins
MPR · Team StrengthHow good is the team?
Matrix Power Rating measures underlying team quality on a neutral field, independent of the schedule it happens to play.
Pure team strengthApply the Schedule→
Opponent strength. Venue. Timing. Context.
Projected Wins · Season OutlookHow many games should it win?
Team strength is run through the actual schedule to create an expected regular-season win total and range of outcomes.
Strength + schedule difficultyWhy both matter: a strong MPR team can project fewer wins because the path is difficult, while a lower-rated team can project a stronger record because its schedule is more favorable.
Definitions
What Every Number Means
Definitions and the season-strength update rules below explain how the outputs are produced and evaluated.
Projected Wins
Expected regular-season wins after MPR is applied to the team’s actual schedule. It is an expectation, not a guarantee of one exact record.
Projected Score
Expected scoring for both teams after strength and matchup context are applied. It produces the projected margin, total and win probability.
Win Probability
The model’s estimated chance of winning the matchup — not a promise that the higher-probability team must win.
Model vs Market
Market information is a separate comparison layer. It does not change MPR, projected wins, projected scores, margins or win probabilities.
Confidence & Range
Confidence, rating uncertainty and outcome distributions communicate that projections have error bands rather than false precision.
Accountability
How We Grade the Matrix
Margin Accuracy
Absolute difference between the projected winning margin and the actual result.
Score Accuracy
Average absolute error across the two projected team scores.
Total Accuracy
Absolute difference between the projected game total and the actual total points scored.
Full-board Market Comparison
Every eligible pregame model side and total can be compared with its recorded line as a diagnostic—not retroactively presented as a published bet.
Published Signals
Only explicit pregame side and total signals are graded against their recorded market lines.
Straight-Up Calibration
Winner accuracy remains available by confidence tier, but it is supporting context rather than the headline measure.
Pregame record rule: every result is graded against the projection and market update that existed before the game. A later model update is never used to make an earlier prediction look better.
Season Model Tracker
Win-Total Tracker
Preseason/current Matrix win projections, market totals and actual records stay visible together. Season signals remain OPEN until the final regular-season win count is known.
138 teams tracked12 strong preseason signals shownSeason grading: OPEN
| Team | Preseason Matrix | Current Matrix | Preseason O/U | Preseason Signal | Actual Record | Status |
|---|---|---|---|---|---|---|
| Texas StatePac-12 | 5.46 | 6.24 | 6.5 | UNDER | 4–3 | OPEN |
| Miami (OH)MAC | 8.54 | 7.67 | 7.5 | OVER | 5–3 | OPEN |
| WisconsinBig Ten | 5.32 | 7.69 | 6.5 | UNDER | 5–2 | OPEN |
| UNLVMountain West | 8.84 | 7.90 | 7.5 | OVER | 4–3 | OPEN |
| PittsburghACC | 6.87 | 7.01 | 7.5 | UNDER | 5–3 | OPEN |
| AkronMAC | 3.72 | 2.37 | 4.5 | UNDER | 1–7 | OPEN |
| Georgia SouthernSun Belt | 5.63 | 6.08 | 4.5 | OVER | 3–4 | OPEN |
| TulsaAmerican | 6.26 | 6.25 | 5.5 | OVER | 4–4 | OPEN |
| Virginia TechACC | 7.53 | 8.12 | 6.5 | OVER | 6–2 | OPEN |
| DukeACC | 6.40 | 7.10 | 5.5 | OVER | 5–2 | OPEN |
| Ole MissSEC | 7.14 | 8.37 | 7.5 | UNDER | 5–2 | OPEN |
| DelawareConference USA | 5.45 | 5.21 | 6.5 | UNDER | 3–4 | OPEN |
Update Discipline
Results update automatically. Ratings do not.
Live/final scores can update throughout the weekend. MPR, Projected Wins and future probabilities change only through the reviewed weekly model-update workflow after results, closing markets and personnel context are reviewed.
Protected Mechanics
Transparent outputs, protected recipe.
GameDay Matrix publishes definitions, updates, signals and performance. Proprietary weights, coefficients, tuning rules and calibration logic remain private.
FCS Cross-Division Calibration
FCS-CAL-v1.1 replaces generic FCS opponent proxies with independent team-specific FCS strength mapped onto the Matrix scale. The FBS side retains its GameDay Matrix MPR. Market lines are validation only, and FCS games are not eligible for market comparisons.
Updated after Week 8
Ratings that learn through the season
Our power ratings now blend the preseason starting point with opponent-adjusted game results. Recent FBS games carry more weight, and unusually large score gaps receive less weight. Teams can move more as season evidence builds.
We will evaluate the frozen forecasts weekly throughout our first season. The early comparison improved margin error by 0.34 points over Weeks 4–5, but the sample is short and missed two original improvement targets. Earlier forecasts and grades remain unchanged. Player availability and total forecasts appear with each matchup.
Active net power uses season strength v1. The original unit ratings remain diagnostics and scoring inputs; their difference no longer equals net power without the explicit season adjustment. Conference title chances use current standings and the active game probabilities; their update date appears on each conference page.
Weekly Model Discipline
Season Strength v1
Each reviewed weekly release refits net power from the season’s FBS results. We preserve every earlier forecast and compare accuracy week by week.
Preseason anchorEvidence changes the blend.
The preseason prior has an eight-game weight. Its share decreases as usable season evidence accumulates.
Opponent adjustmentStrength of opposition matters.
Game margins are evaluated against fitted opponents and the saved pregame venue, rest and travel context.
Recent gamesWeekly recency 0.90.
Older results retain influence, with less weight than recent results.
Large missesReduced influence.
Huber weighting limits the effect of residuals above 14 points. FCS results stay outside this net-strength fit.
Personnel and unitsReviewed separately.
Unit diagnostics and uncertainty retain the evidence review. A rating change does not establish player availability.
Weekly learningFrozen forecasts first.
We grade each saved model version before deciding on further changes. External ratings are comparisons only.