SharpJuice

2026 season to date · 278 graded games

Season results

147–125–6

Against the spread, 54.0% (95% range 48–60%). Break-even at standard −110 pricing is 52.4%.

Closing line moved toward the model
58%
126 of 216 line moves
Average closing line value
0.22 pts
per pick, snapshot to close
Edges of 3+ points
29–26–2
52.7%
Average miss: model vs. market
11.6 vs 11.5
points per game

Win probability (games with a Kalshi price, n=7): Brier score model 0.158 vs. Kalshi 0.133 (lower is better).

Win % by weekAgainst Sunday's line · dark line = 52.4% break-even
0%20%40%60%80%Week 0Week 1Week 2Week 3Week 4Week 5Week 6
Win % by size of edgePoints between model and Sunday's line · dark line = 52.4% break-even
0%20%40%60%80%< 11–22–33–55+

By week

WeekATSWin %Model missMarket miss
Week 06–2–075.0%12.213.4
Week 116–26–138.1%13.913.6
Week 228–21–057.1%11.010.8
Week 334–22–160.7%8.58.7
Week 433–23–258.9%11.711.4
Week 527–28–149.1%13.513.1
Week 63–3–150.0%10.810.1

By edge size

EdgeATSWin %
< 1 pts54–40–257.4%
1–2 pts33–30–152.4%
2–3 pts31–29–151.7%
3–5 pts17–19–247.2%
5+ pts12–7–063.2%

Miss = average absolute difference between projected and actual margin. Closing line value = how far the closing spread moved toward the model's side after Sunday's snapshot; closes for Weeks 0–5 are DraftKings via ESPN, later weeks are our last capture before kickoff. Weeks 0–1 used a three-model average (Action Network added in Week 2).