Machine learning · Quantitative research · Oct 2026 – Present
Matchup DNA
Tennis win probabilities, locked in before the first serve.
A Python research system that rates every ATP and WTA player from their full match history and gives each upcoming match a win probability. Every forecast is stored before the match starts, and a Next.js site shows the forecasts, what drove them, and how they scored.
- pro matches in the history
- 361k+
- matches tested against the market
- 13,846
- winners picked, 31,791 matches
- 65.9%
- Built calibrated ATP and WTA prediction models from Elo, surface Elo, logistic regression, and opponent-adjusted serve and return features, each tested on later years than it was trained on.
- Back-tested the models against betting-market prices on 13,846 matches, measuring calibration, whether the model adds anything to the price, and simulated returns.
- Picked the winner in 65.9% of 31,791 matches from 2021 to October 2026, 1.7 percentage points more than always picking the higher-ranked player.
- Python
- scikit-learn
- LightGBM
- SQLite
- Next.js








