The Options Market Beat 97.6% of Participants in the M6 Financial Forecasting Contest

… and you can probably beat 97.6% of your model wielding business competitors if you think hard about the implications for data science pipelines.

As of today I have permission to make public my somewhat mischievous entry in the year-long M6 stock and ETF forecasting contest. The intent of my entry was to hijack the usual “model versus model” battle. I wanted to see a “market versus model” fight: one where the contest community faced off against a community of a very different kind.

Outline:

1. The hijack
2. Methodology for the options benchmark entry in more detail
3. Results and interpretation: skill versus luck in M6
4. Broader ramifications

Executive summary:

The contest was placed uncomfortably close to a powerful source of predictive power (the options market) and was ripped apart by tidal forces. That makes it a great experiment.

1. A Grand Cultural Battle

Modern terminology for applied mathematics and invented divisions (data science, statistics, machine learning, predictive modeling…) rather get in the way of a more important delineation in the smackdown I wanted to witness: one waged between cultures to some…