Transparency

Results, graded honestly

Every number on this page is the percentage of prop bets these models would have won, based on historical evaluation, with the sample size and period stated for each one. Here is exactly how SharpSlap has performed, at every layer of the model.

63.5%
Aggregate accuracy · 1,280/2,017 qualifying OOS predictions
84.4%
Top-5 SHARP accuracy · 76/90, 2025 historical evaluation
83.3%
2-Leg AI Slap hit rate · 25 of 30 evaluated weeks
70.0%
4-Leg AI Slap hit rate · 21 of 30 evaluated weeks
SHARP Rank

Accuracy concentrates near the top of the rank

SHARP ranks every qualified leg on the board each week. Across the full 2025 regular season (18 weeks), narrower slices of that rank hit at a noticeably higher rate — which is exactly what a working ranking signal should look like.

Top 3
83.3%45 / 54
Top 5
84.4%76 / 90
Top 10
77.8%140 / 180
Top 20
70.8%255 / 360
Top 50
60.1%541 / 900

Accuracy is the percentage of prop bets these picks would have won. Figures are a 2025 historical evaluation across all 18 regular-season weeks, aggregating the selected Top-K legs each week by SHARP score. SHARP rank is available on the Pro and Edge plans as SHARPER.

AI Slaps

The full 30-week parlay record, 2-Leg through 6-Leg

AI Slaps are model-built parlays of 2 to 6 legs, one per size per week. These are the same 30-week historical records shown inside the product's own AI Slaps history panel — nothing different is shown to marketing than to subscribers.

AI SlapHistorical recordHit rate
2-Leg25 / 30 weeks83.3%
3-Leg22 / 30 weeks73.3%
4-Leg21 / 30 weeks70.0%
5-Leg15 / 30 weeks50.0%
6-Leg13 / 30 weeks43.3%

Hit rate is the percentage of these parlays that would have won — every leg landing on the correct side of its line. As you'd expect, hit rate falls as leg count rises — a 2-Leg slip only needs two legs to land, a 6-Leg slip needs all six. Pick the size that matches how much variance you want. See the live AI Slaps history inside the product (Edge plan) for the week-by-week chart.

Production models

63.5% aggregate across 2,017 qualifying predictions

SharpSlap runs 14 active production models, one per prop market. Each is measured on historical, out-of-sample predictions at its live qualification threshold — the same bar a pick has to clear to reach your board.

ModelHistorical accuracyQualifying N
TE Receptions74.5%98
RB Rushing Attempts72.4%29
TE Receiving Yards70.9%110
RB Rushing Yards67.3%113
QB Passing TDs65.6%64
WR Receptions65.1%192
QB Pass Attempts65.1%63
WR Receiving Yards64.4%104
QB Fantasy Points63.8%152
WR Fantasy Points61.8%411
RB Receptions60.8%97
RB Receiving Yards60.5%238
RB Fantasy Points59.6%292
TE Fantasy Points53.7%54

"Out-of-sample" (OOS) means each model was scored on games it did not train on — the standard check for whether a model generalizes rather than having memorized its own training data. Read more on how the calibration and freezing process works on the methodology page, or how accuracy typically firms up in the days before kickoff on the blog.

Historical accuracy describes how a model has performed on past, graded predictions. Every model has losing stretches, which is exactly why the sample size and evaluation period are shown next to every number. Treat it as one input among several.

See this week's board for yourself

Every historical number above exists so you can check it against the live board yourself. Create a free account and look at today's qualifying legs.