I released a dataset of 748 algorithmic crypto trading strategies (BTC/ETH/SOL and other pairs), each run through a strict 7-stage backtest validation pipeline on real Binance OHLCV data: code safety scan, in-sample, out-of-sample, rolling walk-forward, randomized starts, slippage/perturbation stress and a final holdout slice.

Link: github.com/CacheCarti/Crypto-Strategies

License: MIT. Free for any use, attribution appreciated!

Actively contributing weekly drops of 300-400 bots per week.

Dataset contents:

- strategies/: 700+ rejected strategy implementations in Python

- cards/: one JSON report card per bot: stage-by-stage verdicts, in-sample vs out-of-sample Sharpe/returns/drawdown, trade counts, failure class

- index.json: the whole corpus, sorted and filterable

- JUDGE_CHANGELOG.md: validator version history so older cards stay interpretable as the judge gets stricter

What makes it unusual: it publishes the failures, existing strategy repos ship code with no screen attached so you can't tell a…