Player Lab
What it is
Player Lab is the model view for player props: points, rebounds, assists, threes, strikeouts, hits, etc. We project a distribution per player per stat, then compare it live against every book offering that prop. Hit rate and MAE tell you how trustworthy the model has been historically for that player and stat.
When to use it
- You want to fade soft books that haven't updated their prop after a lineup change or breaking news.
- You are shopping for the best line on a prop you already plan to bet.
- You want to find props where the model strongly disagrees with the book, even before sharp books start posting.
How to read it
| Field | Meaning |
|---|---|
| Player | The athlete and their team / opponent / game time. |
| Stat | Points, rebounds, assists, 3PM, K's, hits, total bases, etc. |
| Projection | The model's mean projection for that player in that game. |
| Hit Rate | Historical % of times the model's lean has cleared the line for this player / stat combo. Higher = the model has been right more often. |
| MAE | Mean absolute error of the model's projections vs actual outcomes for this player / stat. Lower = tighter projections. |
| Book Line | The current over/under line and price at each book. Updates live. |
| Edge | Projection minus line. Same direction as the model lean. |
| Live Compute | Player Lab projections recompute on demand as books move and as game state changes (in-play). |
Worked example
Example
Jalen Brunson points, line 27.5
Projection: 30.1 pts Hit rate: 61% (last 25 games for this stat) MAE: 4.2 (tight) Best book: DK Over 27.5 -114 Edge: +2.6 pts vs line Verdict: Over 27.5 looks live. Confirm starter status, check pace, size 1u.
The projection is 2.6 above the line, hit rate is north of 60%, and MAE is tight. That combination is what you want. If MAE were 7.0 you would size smaller because the model is noisy on this player.
Common mistakes
- Ignoring MAE. A 5-point edge on a stat with MAE 8 is noise. A 2-point edge with MAE 3 is signal.
- Betting before confirmed starters. Props are uniquely sensitive to scratches. Always re-check 30 minutes before tip.
- Stacking same-game props. Player props are correlated within a game. Two overs on the same team is one bet, not two.
- Chasing the edge after the line moves. If you saw 27.5 and now it is 29.5, the edge is gone. Move on.
- Forgetting to check role / minutes. Bench players are landmines. The model assumes baseline minutes that may not happen.