Educational/Not betting advice. For modeling and research purposes only.
GP
GametimePicks
v0.4
sample results — demo data

54.8% sample hit rate

248 sample leans · 2026-01-15 to 2026-04-29

sample leans
248
wins
132
losses
109
pushes
7
sample hit rate
54.8%
break-even ~52.4%
by market
Points56-43-3 · n=102
56.6%
Rebounds41-35-2 · n=78
53.9%
Assists35-31-2 · n=68
53.0%
at or above break-evenbelow break-even
by confidence
High38-22-2 · n=62
63.3%
Medium65-56-3 · n=124
53.7%
Low29-31-2 · n=62
48.3%
at or above break-evenbelow break-even
model calibration
predicted vs actual
40%50%60%70%80%40%50%60%70%80%predictedactual
bucket size = sample count · dashed line = perfect calibration

Each circle is a probability bucket: x is what the model predicted on average, y is how often those props actually hit.

A perfectly calibrated model lies on the dashed line. Buckets above mean we under-predicted; below means we over-predicted. Bucket size scales with sample count.

Calibration matters more than raw hit rate. A 60% hit rate on props the model rated 65% likely is worse than a 55% hit rate on props rated 55% likely.

recent settled leans
DatePlayerLeanResultStatus
Apr 29Stephen CurryOver PTS 27.532Won
Apr 29Luka DoncicUnder PTS 33.528Won
Apr 29Joel EmbiidOver REB 11.514Won
Apr 29Tyrese HaliburtonOver AST 10.59Lost
Apr 28Nikola JokicOver AST 9.513Won
Apr 28Jayson TatumOver PTS 28.530Won
Apr 28Giannis AntetokounmpoUnder PTS 31.534Lost
Apr 27Jaylen BrownUnder PTS 24.522Won

Past performance does not guarantee future results. Hit rate alone does not equal profit — sportsbook vig means break-even is typically ~52.4% on -110 props. ROI calculations are intentionally not shown until the methodology supports them rigorously.