How is the Crypto Overfitting Index computed?

Every month our engine backtests 10 classic strategy templates with default parameters across 50 coins on daily bars since 2021, splits each history chronologically 70/30, and measures every configuration twice: on the first 70% (in-sample) and the final 30% it never saw. The index is the percentage of configurations whose out-of-sample Sharpe turned negative — the share of historical promise that failed on unseen data.

Same engine, same rules, same universe every month, so the trend is comparable across readings. The index, its full history and the underlying monthly numbers are free to cite (CC BY 4.0) and available as JSON through the public API.