← Back to Reviews
20260114_142615_vol_control_btc_regime_switch.md
# AI Review: vol_control_btc_regime_switch (1-Minute)
**Date:** Wed Jan 14 14:26:15 CET 2026
**Strategy File:** /root/trade_1m/strategies_rejected/vol_control_btc_regime_switch.py
---
```json
{
"score": 45,
"flags": [
"overfitting",
"cost_sensitive",
"concentration",
"inconsistent"
],
"review": "## Vol Control Regime Switch Strategy Review\n\n### Critical Issues\n\n**1. Severe Validation Failure (BLOCKING)**\nThe strategy produced only 4 trades in validation, falling 60% short of the minimum 10-trade requirement for vol_control role. This is a fundamental failure - the regime it was trained to detect may not recur reliably.\n\n**2. Parameter Overfitting**\nThe strategy uses highly specific thresholds that smell of curve-fitting:\n- ATR z-score compression at exactly -1.0 → -0.3 transition\n- 0.25% minimum move (not 0.2% or 0.3%)\n- 0.45% take profit (why not 0.4% or 0.5%?)\n- 15-bar lookback for regime detection\n\nThese non-round parameters suggest optimization to historical data rather than robust microstructure logic.\n\n**3. Extremely Sparse Trading + Cost Sensitivity**\nWith ~3 trades/month frequency:\n- Each trade must carry 0.1% round-trip cost (0.05% slippage per side)\n- Strategy return of 1.77% across 18 trades = 0.10% avg per trade\n- After costs: near breakeven or negative\n- **The edge is too thin for the friction at this frequency**\n\n**4. Multi-Year Inconsistency**\nThe strategy shows wild variance across longer periods:\n- Returns range from -0.6% to +1.5% annually\n- Training period (last 5 months) shows 1.77% return\n- But 2021 was negative, 2023 was negative\n- Small sample + high variance = unstable edge\n\n**5. Concentration Risk**\nWith only 18 total trades:\n- Removing best 2-3 winners likely flips strategy negative\n- Not enough statistical power to distinguish skill from luck\n- Monte Carlo resampling would show massive uncertainty bands\n\n### Microstructure Logic Assessment\n\nThe core concept \"trade the exit from vol compression\" has theoretical merit:\n- Vol regimes do exist and transition\n- Early momentum after compression can be profitable\n\nHowever, the implementation fails because:\n- Detection is too specific (exact z-score thresholds)\n- 7-day cooldown creates artificial scarcity\n- No evidence this regime occurs frequently enough\n- 15-bar backward check may capture noise, not regime\n\n### Execution Concerns\n\nAt 1-minute resolution:\n- ATR calculations are stable (good)\n- Entry logic checks historical bars, not same-bar (good)\n- BUT: Vol spike exit happens on current bar's z-score, which uses current ATR\n - This could be lookahead if exiting intra-bar based on bar's final ATR\n - Not critical since it's exit-only, but worth noting\n\n### What Would Need to Change\n\nFor this to be viable:\n1. **Higher frequency**: 3 trades/month cannot amortize costs. Need 20+ trades/month minimum.\n2. **Simpler parameters**: Use round thresholds (z-score = -1 → 0, not -1 → -0.3)\n3. **Larger edge per trade**: 0.10% avg is too thin. Need 0.2%+ after costs.\n4. **Validation consistency**: Must produce 10+ trades in any test period.\n\n### Cost Reality Check\n\n- 18 trades × 0.1% cost = 1.8% total drag\n- Strategy return: 1.77%\n- **Net after costs: essentially zero**\n\nThis is not a viable 1-minute strategy.\n\n### Final Verdict\n\n**Score: 45/100 (Poor)**\n\nThe strategy fails the fundamental test for 1-minute trading: **edge must be large enough and frequent enough to overcome friction**. With only 3 trades/month averaging 0.10% profit, costs consume the entire edge. The validation failure (4 trades < 10 minimum) confirms the regime doesn't recur reliably.\n\nThis is a case where a potentially sound concept (vol regime transition) was over-optimized to historical data and doesn't generalize. The specific z-score thresholds and 7-day cooldown created artificial performance in training that collapsed in validation.",
"rule_suggestion": "Strategies with trade frequency below 10 trades/month must demonstrate average profit per trade of at least 0.25% (2.5x the round-trip cost) to survive friction. Additionally, if validation produces fewer than 50% of required minimum trades, the strategy regime may not be statistically reliable and should be rejected regardless of training performance."
}
```