Picks generated from the V4 model. Tap a card to see model reasoning.
Updated: --
Season Tracker
Full Model Performance
Complete season history. Every pick, every result, nothing hidden.
Rolling Win Rate
7-Day
30-Day
PSI+ Added
← scroll to view earlier dates
Daily Calendar
Over Segments
Under Segments
Methodology
How the Model Works
A clear, data-driven system for predicting MLB pitcher strikeouts.
01
Feature Engineering
We analyze over 30 key details for every start: the pitcher's strikeout history, pitch quality, speed and movement, and how opposing batters typically perform against those pitches.
30+ features
02
Six separate models (Binary Classifiers)
Each answers one straightforward question: "Will the pitcher exceed a specific strikeout total?" (covering common lines from 3.5 to 8.5 Ks).
LightGBM · 6 thresholds
03
Beta-Binomial Layer
We convert the raw predictions into reliable percentage probabilities that account for the natural variation in baseball.
Probabilistic · overdispersion
04
Agreement Filter
We only display a pick when all the models agree on the same direction (Over or Under). Disagreements are excluded entirely.