Most backtests look profitable because something was left out: a fee, an impossible fill, a split reused twice.

The complete working archive of a crypto-futures research laboratory built to stop exactly that: engine, validation, execution layer, and the recorded verdict of every campaign that failed.

Read this first: it did not find a profitable strategy.

profitable_strategy_demonstrated = false

That is the central finding, not fine print.

INSIDE
• 1,442 files, 626 Python files, 155 test modules, ~185,300 lines
• Causal backtest engine: fees, spread, slippage, impact and funding charged and reported separately
• Deflated Sharpe, PBO, walk-forward screens, multiple-testing adjustment
• 428 Markdown documents, Obsidian vault, 22 maps of content
• Twelve campaign families, none promoted, all documented with figures
• 231 redistributable tests, 90-check release self-check

ONE MINUTE, FULLY OFFLINE
python quickstart/first_run.py

No account, no API keys, no downloads. The cost identity closes to 0.000e+00; a +2.918353 gross edge still nets -2.097026 after costs; one candidate flips from +0.822698 to -0.766802 on cost assumptions alone.

NOT a bot, a signal service, a live track record or investment advice. Market data is not included. Closed archive: no updates, no support contract.

FOR quant researchers, Python developers and systematic traders who want infrastructure they can run and audit.

NOT FOR anyone wanting signals or a strategy that makes money out of the box.