College football · 2026

Track Record

How the picks held up, in public

2026 so far

pre-game predictions, scored once games are final
Winners picked
324-67
82.9% of 391 games
Beaver’s Picks: sides
195-191-5
50.5% · 52.4% breaks even
Beaver’s Picks: totals
200-191
51.2% · 52.4% breaks even
Average miss, margin
12.6
Vegas 11.8 · lower is better
Average miss, total
12.0
Vegas 11.9
Vegas gaps, against the spread
111-112
226 games with a 3+ pt gap

Are the win chances honest?

dots near the line mean yes
0%25%50%75%100%Beaver's win chance0%25%50%75%100%Predicted 7%, won 14% (7 games)Predicted 15%, won 0% (13 games)Predicted 25%, won 0% (9 games)Predicted 36%, won 25% (20 games)Predicted 45%, won 56% (25 games)Predicted 55%, won 48% (40 games)Predicted 65%, won 67% (24 games)Predicted 76%, won 81% (48 games)Predicted 85%, won 87% (46 games)Predicted 97%, won 98% (159 games)

Winners picked by week

0%25%50%75%100%WeekWeek 1: 87% of 991Week 2: 86% of 862Week 3: 81% of 753Week 4: 82% of 714Week 5: 75% of 595Week 6: 100% of 16

Sam’s Picks

Sam’s projected score against the Vegas line, graded the same way as Beaver’s
Sam’s Picks: sides
1-0
100.0% · 52.4% breaks even
Sam’s Picks: totals
0-1
0.0% · 52.4% breaks even
Games with a Sam’s Pick
58
1 graded so far

How Vegas gaps have done

backtest, model’s side vs. the closing line, pushes excluded
Spread
Gap sizeGamesRecordModel side won
0 to 3 pts3,8231,967-1,85651.4%
3 to 5 pts1,667864-80351.8%
5 to 7 pts1,012517-49551.1%
7 to 10 pts605311-29451.4%
10+ pts292135-15746.2%
Total
Gap sizeGamesRecordModel side won
0 to 2 pts2,6141,338-1,27651.2%
2 to 4 pts2,1201,084-1,03651.1%
4 to 6 pts1,358676-68249.8%
6 to 9 pts940493-44752.4%
9+ pts319167-15252.3%

52.4% is the break-even rate at standard odds. A gap means the model disagrees with Vegas, not that it is right; small samples swing a lot.

Backtest

7,582 games, 2017 to 2026, each season predicted using only earlier seasons
ModelAvg miss, marginAvg miss, totalWinners pickedBeaver sideBrier score
EnsembleIn use12.6713.0675.9%51.3%0.159
Ridge12.7513.0975.6%51.6%0.159
LightGBM12.8013.2475.6%51.1%0.160
Elo baseline13.6813.8273.8%48.9%0.170
Vegas closing line12.1912.67n/an/an/a

Average miss is points between the prediction and the real result. Brier score measures win-chance accuracy: 0.25 is a coin flip, lower is better.

Average miss by season

12.112.813.614.415.1201720182019202120222023202420252026Ensemble · 2017: 13.3Ensemble · 2018: 12.8Ensemble · 2019: 12.5Ensemble · 2021: 13.2Ensemble · 2022: 12.6Ensemble · 2023: 12.3Ensemble · 2024: 12.4Ensemble · 2025: 12.3Ensemble · 2026: 12.6Ridge · 2017: 13.4Ridge · 2018: 13.0Ridge · 2019: 12.5Ridge · 2021: 13.3Ridge · 2022: 12.7Ridge · 2023: 12.3Ridge · 2024: 12.5Ridge · 2025: 12.3Ridge · 2026: 12.7LightGBM · 2017: 13.3LightGBM · 2018: 12.9LightGBM · 2019: 12.9LightGBM · 2021: 13.3LightGBM · 2022: 12.7LightGBM · 2023: 12.5LightGBM · 2024: 12.4LightGBM · 2025: 12.4LightGBM · 2026: 12.7Elo baseline · 2017: 14.2Elo baseline · 2018: 13.8Elo baseline · 2019: 13.3Elo baseline · 2021: 13.7Elo baseline · 2022: 13.6Elo baseline · 2023: 13.3Elo baseline · 2024: 13.7Elo baseline · 2025: 13.4Elo baseline · 2026: 14.9
● Ensemble● Ridge● LightGBM● Elo baseline