Dev Tools · 1h ago
ML Options Forecaster Scores Worse Than Coin Flip on AAPL Bets
A developer graded his own machine learning model's option forecasts against actual AAPL contract outcomes. The mean Brier loss was 0.3846 across two contracts, worse than a coin flip's 0.25. The model overestimated the probability of a call option finishing in the money and underestimated a put option's risk.
Meridian48 take
The post is more about personal accountability in ML forecasting than a breakthrough, but it highlights the gap between backtested models and real-world performance.
Read the full reporting
I graded my own ML option forecasts. Here's the Brier score. →
DEV Community
machine-learningoptions-trading