skrecok/dax-day-types ? reverse-engineered prompt

Build me a small Python research repo that reproduces the analysis from the writeup about DAX day types and why retail trading is basically a coin flip.

I want it to take ten years of DAX one minute session data, clean it into regular cash session days, score how directional each day was, show that day behavior sits on a smooth continuum instead of neat buckets, and then test whether the morning gives any real edge for predicting the rest of the day. It should also include the simple fade and trend style backtests from the article, plus a sweep that shows how win rate can look good while expectancy still goes negative after costs.

Please make it easy to run end to end from a raw CSV, save the charts into a figures folder, and keep the scripts readable and reproducible. If the data is not bundled, just assume a straightforward input file and document the expected columns. A short README with how to run each script and what each figure means would be great. You can look up any current stats or plotting docs online if needed.