2026 season
College football
Every pick saved this season, graded against Vegas's closing line. 2 weeks, 110 of 110 games final.
63–45–1
Against the spread · 58.3% · Elo + eff, math alone
- Winners picked
- 78–31
- Vegas favorites
- 86–24
- CLV at pick
- +0.17 pts
- Brier, ours / Vegas
- 0.172 / 0.156
Against the spread, week by week
30
23
Wk 457%
33
22
Wk 560%
Elo + eff, math alone. Covers above the line, losses below; pushes left out. The percentage under each week is that week's cover rate; 52.4% breaks even.
Every week
- Week 5 Oct 1–3 · 56 games Final 33–22–1 ATS39–17 winnersVegas 41–15
- Week 4 Sep 25–26 · 54 games Final 30–23 ATS39–14 winnersVegas 45–9
Records are Elo + eff, math alone. Each week's sheet has every pick, when it was locked in, and the line then.
Every predictor, season so far
| Predictor | Winners | Brier | ATS | ATS, 3+ pts off | CLV at pick | CLV from open |
|---|---|---|---|---|---|---|
| Vegas favorite | 86–24 | 0.156 | – | – | – | – |
| Elo + eff | ||||||
| Math alone | 78–31 | 0.172 | 63–45–1 (58%) | 28–27–1 (51%) | +0.17 (38–32–39) | +1.03 (65–16–28) |
| + Haiku 4.5 | 80–29 | 0.172 | 59–49–1 (55%) | 24–27–1 (47%) | +0.18 (39–32–38) | +0.92 (62–16–31) |
| + Sonnet 5 | 78–31 | 0.175 | 61–47–1 (56%) | 28–31–1 (47%) | +0.02 (32–32–45) | +0.85 (60–16–33) |
| + Opus 5.5 | 79–30 | 0.173 | 62–46–1 (57%) | 27–30–1 (47%) | +0.17 (38–32–39) | +1.06 (67–16–26) |
| Combined | ||||||
| Math alone | 83–26 | 0.167 | 58–50–1 (54%) | 27–23–1 (54%) | +0.22 (43–32–34) | +0.94 (64–16–29) |
| + Haiku 4.5 | 87–22 | 0.167 | 55–53–1 (51%) | 32–25–1 (56%) | +0.25 (42–32–35) | +0.83 (61–16–32) |
| + Sonnet 5 | 81–28 | 0.169 | 62–46–1 (57%) | 33–24–1 (58%) | +0.14 (38–32–39) | +0.83 (61–16–32) |
| + Opus 5.5 | 83–26 | 0.167 | 61–47–1 (56%) | 27–23–1 (54%) | +0.21 (42–32–35) | +0.92 (63–16–30) |
Each game counts once, at its newest saved version.
Totals (over/under)
| Predictor | Avg miss | Vegas miss | O/U | O/U, 3+ pts off | CLV at pick | CLV from open |
|---|---|---|---|---|---|---|
| Totals model, math alone run9/v1 | 14.40 | 12.95 | 31–25 (55%) | 21–17 (55%) | +0.05 (20–18–18) | +0.77 (29–7–20) |
| Totals model + Haiku 4.5 run9/prompt-v7 | 14.46 | 12.95 | 32–24 (57%) | 18–17 (51%) | +0.05 (20–18–18) | +0.84 (30–7–19) |
| Totals model + Sonnet 5 run9/prompt-v7 | 14.61 | 12.95 | 31–25 (55%) | 16–15 (52%) | +0.20 (22–18–16) | +0.84 (30–7–19) |
| Totals model + Opus 5.5 run9/prompt-v7 | 14.42 | 12.95 | 31–25 (55%) | 20–17 (54%) | +0.05 (20–18–18) | +0.84 (30–7–19) |
Avg miss: points from the final total, ours and the closing total's. O/U: our side of the closing total.
Reading the numbers
- ATS
- Our side of Vegas's closing line. If Vegas has a team −14.5 and we have them by 9.9, our side is the other team +14.5. 52.4% breaks even at standard odds.
- 3+ pts off
- Only the games where we differ from the line by 3 or more: the ones you'd actually consider.
- CLV
- Closing line value: how many points the line moved toward our side after the line we took (then moved our way–didn't move–moved against). Positive CLV over many games is the strongest sign of a real edge. At pick uses the line when the pick was saved; from open is as if we'd bet the opening line.
- Brier
- How good the win percentages were. Lower is better; 0.25 is a coin flip.