College football · Powered by R
Find the gap
before kickoff.
Understand the likely score, the underdog’s chances, and the price on offer. Every forecast comes with its data age and its limitations.
The underdog board
Your next matchups.
Research matchups are ordered by kickoff, not ranked as bets. Historical betting performance has not been evaluated with prices available at prediction time.
Massachusetts at Rutgers
Winning and covering are different.
Cover: finish ahead after the spread is applied. A push returns the stake. Fair odds are a model estimate, not an available offer.
- Cover / push
- Stale quote — withheld
- Model fair moneyline
- —
- Estimated spread return · unverified
- —
Expected return is a model calculation per unit staked, not achieved profit. Large disagreements with sportsbooks may reflect model errors or missing information, not a bargain. Positive point gaps can still have negative expected returns at the offered price.
- Trained on 3155 prior completed FBS matchups, with more weight on recent games.
- Includes pregame rolling PPA, success rate and play volume.
- Experimental baseline; no betting advantage has been demonstrated.
- Injuries, quarterback changes, roster turnover and weather are not included yet.
Compare available books
| Book | Rutgers spread / price | Total / over price | Rutgers moneyline |
|---|---|---|---|
| -29.5 / -110Stale / unavailable · Aug 31, 12:13 AM PDT | 52.5 / -112Stale / unavailable · Aug 31, 12:13 AM PDT | -10000Stale / unavailable · Aug 31, 12:13 AM PDT | |
| -29.5 / -105Stale / unavailable · Aug 31, 12:13 AM PDT | 53.0 / -110Stale / unavailable · Aug 31, 12:13 AM PDT | —Stale / unavailable · Not available | |
| -29.5 / -101Stale / unavailable · Aug 31, 12:13 AM PDT | 53.0 / -108Stale / unavailable · Aug 31, 12:13 AM PDT | —Stale / unavailable · Not available | |
| -29.5 / -105Stale / unavailable · Aug 31, 12:13 AM PDT | 53.0 / -110Stale / unavailable · Aug 31, 12:13 AM PDT | —Stale / unavailable · Not available | |
| -29.5 / -109Stale / unavailable · Aug 31, 12:06 AM PDT | 53.5 / -108Stale / unavailable · Aug 31, 12:06 AM PDT | —Stale / unavailable · Not available | |
| -29.5 / -110Stale / unavailable · Aug 31, 12:12 AM PDT | 53.5 / -108Stale / unavailable · Aug 31, 12:12 AM PDT | -10000Stale / unavailable · Aug 31, 12:12 AM PDT | |
| -30.5 / -114Stale / unavailable · Aug 31, 12:02 AM PDT | 53.5 / -114Stale / unavailable · Aug 31, 12:02 AM PDT | —Stale / unavailable · Not available | |
| — / —Stale / unavailable · Not available | 53.5 / -110Stale / unavailable · Aug 31, 12:11 AM PDT | —Stale / unavailable · Not available | |
| -29.0 / -110Stale / unavailable · Aug 31, 12:13 AM PDT | 53.0 / -110Stale / unavailable · Aug 31, 12:13 AM PDT | -10000Stale / unavailable · Aug 31, 12:13 AM PDT | |
| -29.0 / -110Stale / unavailable · Aug 31, 12:11 AM PDT | 53.0 / -110Stale / unavailable · Aug 31, 12:11 AM PDT | —Stale / unavailable · Not available |
Line movement · DraftKings
Waiting for a second captured quote. No line movement is invented.
Show the work
Results, not reputation.
2025 chronological evaluation. Lower score error and Brier score are better.
CFBD historical revisions are not point-in-time archives; final results/stats are assumed available 24 hours after kickoff. Betting ROI, against-the-spread results and closing-line performance have not been established.
Baselines, probability calibration & limitations
Home-only margin baseline error: 15.96 points. Average-total baseline error: 13.32 points. Actual coverage of the nominal 80% margin range: 82.1%.
| Home win probability band | Games | Mean forecast | Actual win rate |
|---|---|---|---|
| 0.0%–10.0% | 9 | 6.9% | 0.0% |
| 10.0%–20.0% | 29 | 14.7% | 20.7% |
| 20.0%–30.0% | 54 | 26.0% | 24.1% |
| 30.0%–40.0% | 75 | 34.8% | 32.0% |
| 40.0%–50.0% | 94 | 45.3% | 40.4% |
| 50.0%–60.0% | 125 | 54.9% | 58.4% |
| 60.0%–70.0% | 134 | 64.8% | 67.9% |
| 70.0%–80.0% | 122 | 74.7% | 71.3% |
| 80.0%–90.0% | 93 | 84.4% | 91.4% |
| 90.0%–100.0% | 57 | 94.3% | 98.2% |
- No demonstrated betting advantage.
- No timestamp-aligned historical betting evaluation yet.
- Fixed starting hyperparameters; not optimized.
- Injuries, transfers, quarterback changes and weather are not modeled yet.
- Gaussian marginal outcome approximation; football key-number and joint-score structure are not modeled.
Built for evidence, not hype
Every number
has a timestamp.
The R model learns team strength, home-field advantage and scoring patterns from prior games. Available advanced stats add efficiency and pace. Whole weeks are held back before testing. Missing inputs stay missing.