Method
The models are private. What's public is how picks are posted and graded, and how each model did on games it never saw before launch.
The rules
Posted before the game
Every pick locks before tip-off or puck drop, with a timestamp, the price and the book. Nothing is added after the fact.
Graded at the posted price
Results use the price that was posted, juice included. A prop is void if the player doesn't play. Nothing is edited or deleted.
Checked against the close
The closing price is logged before every game. Beating it over hundreds of picks is the clearest sign the edge is real.
NBA player props: the test
Points, rebounds and assists, 1 unit per pick. Tested on 2025-26 games the model never saw, against real sportsbook prices.
+7.4%ROI on 617 unseen bets, best price across books
55.4%Of those bets won, mostly near even money
+8.0%ROI on the 766 bets used to set the rules
292Unseen games, 2025-12 to 2026-01
| Where you bet | Bets | ROI | 95% range |
|---|---|---|---|
| Best price across 6 US books | 617 | +7.4% | -0.3% to +15.1% |
| DraftKings only | 195 | +0.2% | -13.4% to +13.8% |
| FanDuel only | 196 | +3.4% | -10.3% to +17.2% |
| BetMGM only | 277 | +6.6% | -4.7% to +17.8% |
NHL moneylines: the test
Four seasons, each priced only from games before it.
1,660Bets over 5,233 games
+128.2uUnits won
+5.0%ROI, 95% range -0.6% to +10.6%
-40.1uDeepest drop from a peak
| Season | Units | ROI |
|---|---|---|
| 2022-23 | +40.1 | +5.5% |
| 2023-24 | +10.3 | +1.6% |
| 2024-25 | +17.6 | +2.7% |
| 2025-26 | +60.3 | +10.8% |
Terms used on this site
- Unit
- 1% of bankroll. NBA props are 1 unit each. NHL picks are 1 or 2 units.
- ROI
- Units won divided by units staked.
- Closing-line value (CLV)
- How far the market moved toward the pick between posting and the last price before the game, in percentage points.
- Out of sample
- Games the model never saw while it was being built.