20260115_004041_breakout_fade_btc_micro.md

# AI Review: breakout_fade_btc_micro (1-Minute) **Date:** Thu Jan 15 00:40:41 CET 2026 **Strategy File:** /root/trade_1m/strategies/breakout_fade_btc_micro.py --- ```json { "score": 45, "flags": ["cost_sensitive", "inconsistent", "execution"], "review": "## Breakout Fade Strategy Review\n\n### Concept Validity ✓\nThe core idea is sound for 1-minute microstructure:\n- False breakouts are frequent at micro timeframes\n- Wick rejection + volume spike is a legitimate signal\n- Mean reversion to VWAP is appropriate for short holding periods\n- Round number parameters (15, 30, 20) suggest generalization over curve-fitting\n\n### Critical Issues\n\n#### 1. Cost Sensitivity - MAJOR CONCERN\n- **Avg trade: 0.012%** - This is catastrophically below the 0.05% cost floor\n- With 0.10% round-trip slippage (0.05% per side), every trade loses ~0.088% on average\n- Total trades: 4,200 → Slippage cost: 420% (!)\n- After costs, the strategy would lose hundreds of percent, not gain 3.83%\n- **This strategy cannot survive real execution costs at this trade frequency**\n\n#### 2. Monthly Inconsistency\n- Two losing months out of five (Jul: -1.6%, Nov: -3.5%)\n- Sharpe flips from +1.86 to -0.83 across months\n- August dominates profitability (5.3% of 3.83% total)\n- Suggests the edge is condition-dependent or unstable\n\n#### 3. Execution Reality at 1-Minute Scale\n- **Same-bar detection and entry**: The strategy detects a false breakout (wick + close back inside) and enters on that exact bar\n- At 1-minute resolution, by the time you observe the bar close and confirm the pattern, you're already entering the *next* bar\n- The +1 bar delay stress test shows 70% edge retention, but with avg trade at 0.012%, even 100% retention doesn't help\n- Volume spike (1.5x) may occur *because* of the failed breakout, not before your entry\n\n#### 4. Trade Size vs Edge\n- 4,200 trades over 5 months = 28 trades/day\n- At this frequency with 0.012% avg edge, you need **perfect execution**\n- Any adverse selection (getting filled on losers, missing winners) destroys the strategy\n- High-frequency mean reversion requires sub-basis-point edge, which this doesn't have\n\n### What Works\n- Simple, interpretable logic (not a black box)\n- ATR z-score filter prevents trading in abnormal volatility\n- Cooldown prevents overtrading the same setup\n- Exit logic uses VWAP and time-based, both reasonable\n- Win rate ~59% shows some predictive signal exists\n\n### What Doesn't Work\n- **Edge is too small for cost structure** - fundamental blocker\n- High trade frequency amplifies cost impact exponentially\n- Monthly inconsistency suggests regime sensitivity\n- Same-bar pattern detection → entry may not be realistic\n\n### Path to Viability (if redesigned)\nTo salvage this concept:\n1. **Reduce trade frequency by 90%** - filter for only extreme false breakouts\n2. **Increase avg trade to >0.15%** - need 3x cost buffer minimum\n3. **Add regime filter** - only trade in conditions where edge is proven\n4. **Test +1 bar entry** - ensure signal is observable before execution\n5. **Require wider wicks** - current 1-tick breakout-and-close-back is too sensitive\n\n### Verdict\nThe strategy has a theoretically valid microstructure edge (false breakout fading) but **fails to implement it profitably**. The average trade of 0.012% cannot overcome 0.10% round-trip costs, making this unviable for live trading despite positive backtest returns. The high trade count (4,200) amplifies losses rather than profits when costs are applied. Monthly inconsistency and same-bar execution further reduce confidence.\n\n**Not recommended for deployment.**", "rule_suggestion": "Strategies with >1000 trades over the test period must demonstrate avg_trade_pnl >= 0.15% (3x the round-trip slippage cost). High-frequency approaches require proportionally larger per-trade edge to overcome cumulative friction." } ```