Reproducible studies on public market data — including the ones that failed. Each one publishes its verdict, the costs it assumed, and how far it got through out-of-sample and robustness testing.
Filtering pullbacks using market efficiency (fractal dimension) and institutional limit order absorption turns a loss-making pullback ruleset into a robust, high-Sharpe trading strategy on BTCUSDT 1h.
Robust mean-reversion strategy entering longs on oversold conditions (RSI_14 < 30 and close < BB_Lower) and exiting on recovery. Proven highly robust with WFE = 2.01.
The definitive multivariable combo strategy on BTCUSDT 1h, out-of-sample validated under realistic commissions.
A textbook Bollinger-band fade looks tradeable in-sample - positive Sharpe, ~50% win rate over 1,291 trades. Walk-forward analysis kills it. A short case study in why out-of-sample validation is non-negotiable.
Symmetrical dual squeeze-release breakouts combined with Shannon price entropy exits cure the out-of-sample parameter decay seen in directional breakout strategies on BTCUSDT 1h.
A robust volatility breakout strategy on BTCUSDT 1h, gated by Hurst Exponent and Fractal Dimension Index, out-of-sample validated under real fees.
Short-only strategy based on custom squeeze-release indicators on BTCUSDT. Backtest shows Sharpe 1.64 and low drawdown, but WFE of 0.35 signals high potential for overfitting.
BTC 24h shapes before 08:00 UTC funding: holdout cluster C0 +3.3 bp, t=0.36 vs other clusters. NO_EDGE. Distinct from the already-killed funding-rate level.
A case study on why profitable trading patterns can still be rejected. Analysis of a 5-candle BTC capitulation bottom that returned +2.45% but severely underperformed the passive market baseline (+179.35%).
Crowded-long ∩ 15m vol filter: val n=38 mean +66 bp but t=1.41 — louder, not stronger than parent t=1.60. WATCH.
One of 276 crypto hour-pairs looked statistically compelling. Then realistic costs and chronological validation made it vanish across BTC, ETH and SOL.
Four honest negative results from selective ML, order-book, macro and on-chain research—and the reusable boundaries each failure added to the knowledge base.
A comprehensive study on 60,000 hourly bars (2019-2026) for BTC and ETH. Proves that 24h high breakouts have a robust, positive expectancy when aligned with the macro trend (+17.67% BTC, +10.88% ETH net of taker fees).
Pro DQN leftover+price-only both cancelled at 1800s (~ep 164/200), no model. Dummy overlapping val t=4.10; nonoverlap t=1.42. NO_EDGE.
Feature Lab marks lsratio/HIGH actually_tradable. Neural: no incremental vs price MLP. Overlapping val t=4.10; nonoverlap t=1.60; WFE=0.16. WATCH, not a book.
BTC top-trader L/S HIGH: val fade t=−2.17 but corr vs global 0.99; incremental γ t=−1.45. Same costume. NO_EDGE.
We preregistered the trapped-long sequence (OI expansion, aggressive buying, failed breakout, OI contraction) and tried to kill it. It died. Hyperliquid has no OI history; this is a Binance 5m negative result.
A 24-bar volatility-clustering persistence signal (ARCH-effect autocorrelation) combined with a weekend-volatility-shock exclusion filter lifts BTCUSDT 1h Sharpe from 0.85 to 1.20 (+114.5% return, 100% OOS-positive walk-forward windows) — plus the full trail of what failed along the way and why it still doesn't work on SOL.