CryptoLab: Walk-Forward Crypto Research and Gated Paper Trading
A Binance USDT-M futures research system where only a strategy that survives stitched out-of-sample tests can be promoted.
Walk-forward with realistic fees and funding, a sealed holdout year, funding carry identified as the only edge, meme-coin and copy-trading studies that debunked popular ideas, academic replications, and an executor that refuses live until a gate file passes.
- Walk-forward with sealed holdout and modelled funding
- Meme breakout, Supertrend tuning and copy-trading leaderboards all debunked with data
- Academic strategies replicated for calibration
- Live executor hard-gated behind paper results
CryptoLab exists to answer "what actually works in crypto futures" with evidence before any capital moves. Strategies are evaluated walk-forward (twelve months train, three months test) with maker and taker fees, funding modelled, and a sealed holdout year that is never used for selection. Funding carry was the only strategy with a healthy out-of-sample Sharpe; cross-sectional momentum, trend, mean reversion and a LightGBM model all died to costs or turnover.
Several popular ideas were tested properly and retired. A descending-trendline meme-coin breakout produced expectancy indistinguishable from random entries across 46 variants. A 2.15-million-run Supertrend sweep showed the default setting as good as any and rank correlation between train and holdout near zero. A study of 8,889 copy-trading leaders found top-decile 90-day returns predicted a median of zero for the next 90 days. Academic Donchian, moving-average filter and time-series momentum papers were replicated with lower but directionally consistent results.
Operationally, scheduled tasks run the paper book hourly and a daily report; the executor refuses to trade live unless a gate file records four weeks of positive paper results under a drawdown cap, and only I flip that gate.
Research and paper trading; shared as engineering work, not financial advice.
How it works
- Hourly task pulls perpetual prices and funding rates
- Strategies are re-evaluated walk-forward with costs
- Portfolio weights update from stitched out-of-sample results
- Paper executor journals fills; daily report summarises
- Gate file must pass before the live path is enabled
Software and automation are infrastructure, not financial advice. No performance guarantees. Every live system needs staged testing, risk limits, and owner approval.