MLB Strikeout Predictor

Baseball
Python
Machine Learning
Streamlit
An interactive app that predicts pitcher strikeout totals using Statcast data, K/9, Whiff%, and opponent batting tendencies.
Published

May 5, 2026

Overview

This tool predicts how many strikeouts a starting pitcher will record against a given opponent in an upcoming game. Select any active 2026 MLB starter and an opposing team — the model returns a predicted K count with a transparent breakdown of the key factors driving the estimate.

Model inputs:

  • K/9 — The pitcher’s season strikeout rate per nine innings (primary signal)
  • Whiff% / SwStr% — Swinging-strike rate, a pitch-quality proxy from Statcast
  • Opponent K% — How often the opposing lineup strikes out, relative to the league average
  • Projected IP — The pitcher’s average innings per start (scales the opportunity)

Prediction formula:

\[\hat{K} = \frac{K/9}{9} \times \overline{IP_{GS}} \times \frac{Opp\ K\%}{Lg\ K\%} \times f_{whiff}\]

Confidence range uses Poisson variance scaled by model uncertainty.

Live App

Data Sources

Source What it provides
FanGraphs via pybaseball Pitcher K/9, SwStr%, GS, IP
Baseball Reference via pybaseball Team batting K% and plate appearances

Data refreshes every 12 hours via Streamlit’s cache layer. All stats reflect the current 2026 season; the app falls back to 2025 data for pitchers with fewer than 3 starts.

Deploy Your Own

The source code lives at projects/mlb-strikeout-predictor/ in this repo.

# Run locally
pip install -r requirements.txt
streamlit run app.py

To deploy on Streamlit Cloud:

  1. Go to share.streamlit.io and connect your GitHub account.
  2. Select this repository, set the App file path to projects/mlb-strikeout-predictor/app.py.
  3. Click Deploy — Streamlit Cloud installs requirements.txt automatically.
  4. Copy the generated URL and paste it into the src attribute of the <iframe> above.