MATCHDAY MODEL RECAP
Matchday's model graded 11 College Football picks this stretch, hitting on 9 of them (82%). Over the same 10 games with a market line, the betting market favorite hit 90% of the time.
On calibration: picks stated at 70-80% hit 100% of the time (3 games); picks stated at 80-90% hit 67% of the time (3 games); picks stated at 90+% hit 100% of the time (4 games). A well-calibrated model's stated confidence should roughly match its actual hit rate in each band.
The strongest signal lately has been talent or squad quality: when it favored a side, that side won 90% of the time (10 games).
Among this week's College Football storylines: the biggest upset was Colorado winning 13-14 against Georgia Tech (the market gave that result only about 32%); the model's best call was Utah in Utah vs Idaho at 92% confidence; its biggest miss was picking Rutgers in Rutgers vs Massachusetts, but Massachusetts won instead.
This is Matchday's own model reporting on itself for anyone to use. Every pick is locked before kickoff and graded automatically once the game finishes; none are rewritten after the fact. See the Q&A page for how the model works.
Matchday — pregame predictions, market context and postgame grading.