Articles

Notes on machine learning, quantitative analysis, and the things I learn along the way.

Quantitative Finance2026-03-3020 min

Polymarket Earnings Prediction: Beating the Crowd with ML + Alternative Data

An end-to-end ML pipeline that predicts S&P 500 and Russell 2000 earnings beats, finds mispriced markets on Polymarket, and sizes positions using Diversified Kelly criterion with GPT-4o, FinBERT, insider trading, and options sentiment as risk filters. 81.5% win rate, +278% return on backtest.

pythonmachine-learningxgboost

Statistics2026-02-1910 min

Beyond the Bell Curve: Why Markets Don't Follow Dice Laws (and the Math Behind It)

Dice follow the Bell Curve; markets do not. The standard Normal model dangerously assumes constant volatility and mathematically erases the probability of extreme crashes. To capture reality, true quants combine GARCH to track how volatility clusters and the Student-t distribution to measure the devastating "fat tails" of the crowd.