Sharp Sports
Track record
Every number below comes from the same backtest behind today's picks, scored only on games the model had never seen. 7642 games, 2023-07-01 to 2026-07-27. League NRFI rate over this span: 50.8%. NRFI and YRFI show up as two separate tables below, since each is graded on its own terms. The daily pick itself is just whichever side clears the bar that day. Want the day-by-day log instead? See every pick →
NRFI hit rate by confidence level, all years combined
| Model confidence | Games | Actual NRFI rate | Units |
|---|---|---|---|
| ≥ 55% | 3186 | 58.4% | +508.9u |
| ≥ 60% | 1645 | 61.0% | +299.6u |
| ≥ 65% | 764 | 64.3% | +188.2u |
| ≥ 70% | 371 | 66.0% | +103.9u |
| ≥ 75% | 135 | 68.1% | +43.3u |
This shows how often the outcome matched the model's stated probability. It's not a claim about beating any specific market's price. Compare it to whatever price you're actually offered before deciding a bet is worth it.
YRFI hit rate by confidence level, all years combined
| Model confidence | Games | Actual YRFI rate | Units |
|---|---|---|---|
| ≥ 55% | 1691 | 60.9% | +269.9u |
| ≥ 60% | 948 | 62.9% | +188.0u |
| ≥ 65% | 418 | 69.1% | +133.8u |
| ≥ 70% | 192 | 68.8% | +59.1u |
| ≥ 75% | 63 | 74.6% | +26.6u |
Excludes games where the model's probability was saturated (see the day-by-day log for how that's defined). Below that floor the model isn't really telling games apart, so we don't report it as a calibrated number.
NRFI year by year, every confidence level
| Year | ≥ 55% | ≥ 60% | ≥ 65% | ≥ 70% | ≥ 75% |
|---|---|---|---|---|---|
| 2023 | 390–312 (56%) +51.0u |
289–215 (57%) +54.0u |
189–142 (57%) +33.9u |
137–80 (63%) +47.5u |
56–29 (66%) +23.1u |
| 2024 | 647–431 (60%) +171.2u |
416–266 (61%) +121.2u |
170–78 (69%) +80.2u |
66–30 (69%) +31.4u |
22–6 (79%) +14.5u |
| 2025 | 432–308 (58%) +94.1u |
157–88 (64%) +58.1u |
59–28 (68%) +26.9u |
14–8 (64%) +5.0u |
4–5 (44%) -1.3u |
| 2026 | 391–275 (59%) +192.5u |
142–72 (66%) +66.2u |
73–25 (74%) +47.1u |
28–8 (78%) +19.9u |
10–3 (77%) +7.0u |
2023 starts mid-year, since the model needs a season of history before it can score anything, and 2026 is partial season-to-date. Higher-threshold columns have fewer games per year, so treat individual years as directional and the multi-year total as the real number.
YRFI year by year, every confidence level
| Year | ≥ 55% | ≥ 60% | ≥ 65% | ≥ 70% | ≥ 75% |
|---|---|---|---|---|---|
| 2023 | 82–52 (61%) +20.8u |
52–28 (65%) +18.2u |
27–6 (82%) +18.0u |
13–3 (81%) +8.5u |
10–0 (100%) +8.9u |
| 2024 | 292–201 (59%) +58.3u |
176–118 (60%) +38.3u |
83–42 (66%) +31.7u |
33–13 (72%) +16.3u |
4–3 (57%) +0.6u |
| 2025 | 363–250 (59%) +72.3u |
241–141 (63%) +73.0u |
116–59 (66%) +44.0u |
67–37 (64%) +22.5u |
23–10 (70%) +10.4u |
| 2026 | 292–159 (65%) +118.5u |
127–65 (66%) +58.6u |
63–22 (74%) +40.1u |
19–7 (73%) +11.8u |
10–3 (77%) +6.8u |
Same caveat as the NRFI table above: read individual years as directional, the multi-year total as the real number.