Trading Strategies
Four strategies running in parallel. One validated on live capital, three in paper testing.
Donchian Breakout
LIVE20-period channel breakout on BTC/ETH 4h, filtered by EMA 8/21 trend regime. OOS Sharpe 1.48, PF 2.19 across 815-day backtest. Trades only when the trend is clearly established.
Trend Pullback
LIVEEnters during pullbacks within confirmed trends. EMA(50) with 1% pullback trigger, 4× ATR take-profit. OOS Sharpe 0.45, PF 2.88 on ETH 1h.
Volatility Breakout
LIVEBollinger Band compression detector. Enters on expansion after low-volatility squeeze. OOS Sharpe 0.36, PF 2.38 on ETH 1h.
RSI Momentum
LIVERSI(35/65) with trend filter on BTC 1h. Catches oversold bounces in uptrends and overbought fades in downtrends. OOS Sharpe 0.25, PF 1.79.
Whale Copy (Scalper)
LIVEPolymarket-style copy trading. Tracks three proven Hyperliquid scalpers (BTC, BTC/ETH, diversified) with 58-62% measured win rates and 2.4-4.6 R:R. $11 copies, independent 2% hard stop, $10 lifetime loss kill.
How It Works
The 4H mean reversion signal, step by step.
Compute Z-Score
Calculate the deviation of current price from the 4-hour rolling mean, normalized by standard deviation. Extreme Z-scores indicate overextension.
Confirm Regime
Check regime model output. Mean reversion only fires in FLAT or TRANSITION regimes. In TRENDING regimes, the signal is suppressed to avoid fighting momentum.
Risk Gate
Pass through Kelly criterion sizing, circuit breaker check, and TP:SL validation. If any layer rejects, the trade is blocked. No overrides.
Execute + Monitor
Place maker order with ATR-based stop loss and take profit. Position monitor activates trailing stop at +0.5% profit, trailing 2% from peak.
Information Coefficient by Asset
Predictive power of the 4H reversal signal varies by asset. Higher IC = stronger edge.
Strategy Graveyard
I tried 12 strategies. I killed 6. Here's how they died.
94% win rate in backtesting. Looked incredible. Walk-forward test showed negative expectancy in ALL three out-of-sample periods. Average win: $0.013. Average loss: $0.371. A 29:1 loss-to-win ratio hiding behind a high win rate. Lesson: win rate is vanity. Expectancy is everything.
0% win rate. 153 trades. -$39.48. Root cause: the buy/sell logic was literally inverted on line 42. Every buy signal was actually a sell signal. Found during a P0 code audit. Lesson: always verify signal direction.
Signal threshold was set to 0.55. The mathematical maximum of the signal formula was 0.45. It was physically impossible for this strategy to generate a trade. Zero trades in 42.5 hours of runtime. Lesson: validate signal generation before deployment.
613 trades at 2.4% win rate. Lost $41.27. Transaction costs alone exceeded any possible edge. The SAR signals were overwhelmed by fees at this capital scale. Lesson: fees are the binding constraint at small capital.
Had random.random() in the signal generator. It was literally flipping a coin to decide trade direction. 6 trades, 33% win rate, -$0.29. Lesson: always audit your signal logic.
Copy 7 top whales' entries. 66 trades, 33% win rate, -$5.24. The fatal flaw: no stop-loss at all. Positions could drift from -3% to -50% with zero protection — the only exit was when the whale exited. Lesson: every position needs independent exit protection.