How it's read
A price, not a prediction
A model that says “this guy gets a hit” tells you nothing you can act on. A model that says 63.1% tells you exactly what the number has to be before the bet is worth making.
The book prices the same event at −130, which implies 56.5%. We think it happens more often than that. The difference is 6.6 points, and that difference is the whole product.
Anything under a four-point gap never reaches the board.
Hits — over 0.5
Edge
+6.6%
Rating
Illustrative numbers, chosen to show the arithmetic. Live figures are on the board and in the record.
Six markets, every hitter
Priced independently, one model each
H
Hits
TB
Total bases
HRR
H+R+RBI
HR
Home runs
BB
Walks
K
Strikeouts
The scale
Rated in volts
Every posted pick carries a rating from two to five volts. It combines the size of the edge, how confident the model is in the projection, and how much the posted line differs from where we'd set it.
More lamps is a bigger disagreement with the book — not a promise. Five-volt picks lose too, and when they do you'll see it in the record.
- 6%+Top of the board
- 4%+Strong
- 3%+Playable
- 2%+Thin
- < 4%Not posted
Every day, in order
The board is set before you're awake
019:30 ET
Statcast
Overnight pitch data folded into every hitter's rolling form.
02
Model
An XGBoost model per market prices each hitter's line, calibrated against settled results.
03
Lines
Sportsbook numbers pulled for every hitter on the slate.
04
Edge
Model blended with the market price, minus that price. Under 2% is dropped.
05
Board
What survives is rated 1-5 volts and posted.
063:00 ET
Graded
Box scores settle every pick overnight. Wins and losses both stay up.
Steps two through five run straight after the 9:30 job. Nothing is added or edited by hand once the board is up.
Every pick we post, we grade
Wins and losses both stay on the record, with the edge we claimed at the time sitting next to each one. It's open — read it yourself, no account needed.