Idea3 stated, explored Backtest52 in-sample + costs Out-of-sample22 unseen data Robustness8 walk-fwd, MC Replication4 independent Survived 17 studies Open 54 studies Failed 18 studies Survived 17 Institutional Alpha Architecture BTC 7-Year High-Sharpe Breakthrough The 1-Minute Liquidity Hunt Sovereign 1-Minute BTC Engine Optimized BTC 5m Oscillation Exhaustion BTC 1H Combined Gate Reversion RLXBT Agent Ingestion Guide No-Code Quant Research SOL 1M BTC-Triggered Spike Reversal BTC 1H Hurst Breakout BTC 4-Leg Combo Strategy Autopilot Strategy Creation & Evolution with AI Agents and Deep RL Curing the Falling Knife Curing Breakout Parameter Decay Robust RSI + Bollinger Mean Reversion Strategy on BTCUSDT 1h 7-Year Price Action Study Surviving Market Friction Open 54 Feature Lab calls leftover WATCH tradable. The neural and the book disagree I Asked an AI Agent to Find a Tradable Candlestick Pattern. It Refused to Pretend Three T… Volatility-Clustering Persistence + a Weekend Filter Squeeze-Release Short Custom Squeeze-Release Short Strategy with Custom Features on BTCUSDT 1h A 30-minute DQN does not finish, so it cannot beat the dummy Shapes before the 8am UTC funding print are not an edge Pattern Lab analogues do not retrieve the leftover A 4h trend filter does not save the 5m fade 5-minute volume leftover is real too, still not a trade A vol cluster is just three large bars wearing a name Classic pin bars are just close-location in a costume Crowded-long plus vol context is louder, not stronger A three-asset crowded-long book is weaker, not stronger Taker long/short volume ratio does not fade the next window After the leftover hunts, still no live book Position-weighted top-trader longs do not clear validation Crowded-long short does not survive a single-position book Top-trader longs are the same crowded-long dummy Crowded-long short clears validation Perp taker fade is not a new object beyond spot Crowded longs underperform on ETH and SOL too SOL 15-minute volume leftover is just a bigger shock ETH 15-minute volume leftover is not a BTC costume Crowded longs underperform; crowded shorts do not bounce Funding vs next perp is the same null as funding vs spot BTC vol memory length will not rank the same twice 15-minute volume still adds a little extra next-bar size Residualized OI is still just vol, and the fade does not travel What a week of falsification left standing ETH/BTC residual does not mean-revert on a frozen z-gate ETH and SOL copy the thin BTC fade Cross-sectional crypto momentum does not clear validation BTC 5-day extremes do not mean-revert BTC daily time-series momentum does not clear validation BTC open-interest swings mark vol, not the next direction Four thin BTC fades are one costume; fees still win BTC premium-index fade looks good until you shuffle it Taker-buy extremes fade a little, then fees eat it Extreme BTC funding does not fade the next window Donchian breakouts fade a little, then fees eat it BTC weekday is a vol calendar, not a return premium Large wicks do not fade the next BTC bar BTC vol clusters, but the half-life will not sit still Three same-sign bars fade a little, then fees eat it Close near the high fades a little, then fees eat it Volume mostly tags the vol regime, not the next move BTC open gaps vanished and never picked a side Range extremes are just yesterday's trend After a vol shock, size persists but the sign is a coin Comprehensive Price Action Study A Cross-Asset Signal That Survived Costs Optimizing RL Agents with Exit Rules in RLXBT Portfolio Regime Research with RLX Failed 18 The Momentum Mirage The regime reversed Sharpe 10.73, Still Untradeable Costs ate the edge The 0.033-Basis-Point Edge Costs ate the edge The Seven-Hour Ghost Overfit — broke out of sample Alt/BTC momentum is still just 2021 Overfit — broke out of sample The Kalman Hedge Ratio Made the Spread Noisier preregistered gates failed gross edge and breakeven The 2.3 Independent Bets Were a Measurement Artifact cost exceeds edge Volatility compression predicts persistence, not expansion Overfit — broke out of sample Vol-regime flagship fails on 4h and 1d too Costs ate the edge Hawkes intensity does not save the 1h vol-regime flagship Costs ate the edge The Sequence That Wasn't There Overfit — broke out of sample Four Trading Ideas That Failed Costs ate the edge BTC 1H Capitulation Bottom Reversal Rejected at validation The In-Sample Trap Rejected at validation UTC hour effects on BTC are vol, not a return premium Overfit — broke out of sample BTC shocks lift ETH size, not same-sign lead Overfit — broke out of sample Maker Execution Needs an 88% Fill Rate to Beat Taker breakeven fill rate above preregistered bar The Alt/BTC Pairs Are Not Cointegrated no cointegration and no economic reversion
Left to right: the gates a study has to pass. A study sits at the last gate it was tested at; ✕ marks where a failed one stopped. Hover for detail, click to open.