Why Good Strategies Suddenly Stop Working
Our Pipeline trend-following strategy backtested at 70%+ win rate. Paper Trading went well at first. Then, an entire week passed with zero signals.
No bugs. No data issues. The market shifted from Trending to Ranging.
Trend strategies need price to move in one direction. When markets consolidate sideways, that premise disappears. Forcing trades means losses.
This is the core concept of Market Regime: markets don’t move the same way forever, and your strategy shouldn’t trade the same way forever.
How We Detect Market State
We combine four indicators to classify market state:
1. ADX (Average Directional Index) — Trend Strength
ADX is the backbone. It doesn’t tell you direction — just how strong the trend is:
- ADX < 20: Ranging market, no clear trend
- ADX 20-25: Transition zone, trend forming or dying
- ADX > 25: Trending market, clear directional move
We set ADX 20 as the threshold. Below 20, trend strategies pause.
2. BB Width (Bollinger Band Width) — Volatility
Bollinger Band width reflects market volatility:
- Narrow (< 5%): Market is coiling, breakout may be coming
- Normal (5-15%): Average volatility
- Wide (> 15%): High volatility, big move or panic
Narrowing BB + rising ADX = early breakout signal.
3. ATR (Average True Range) — Absolute Volatility
ATR measures absolute price movement. Primarily used for stop-loss sizing and position sizing. Low ATR = tighter stops. High ATR = wider stops.
4. EMA Slope — Direction Confirmation
50-period EMA slope confirms trend direction: rising = bullish, falling = bearish. This pairs with ADX: ADX says “there’s a trend,” EMA says “which way.”
Four Market States
Combining these indicators, we define four states:
| State | ADX | BB Width | Characteristics | Best Strategy |
|---|---|---|---|---|
| TRENDING | > 25 | Widening | Clear trend | Pipeline trend-following |
| RANGING | < 20 | Normal | Sideways | Mean Reversion + BB Squeeze |
| HIGH_VOL | 20-25 | > 15% | Volatile, uncertain | Reduce position size |
| BREAKOUT | Rising | Narrow→Wide | Breakout | BB Squeeze |
Automatic Strategy Switching
With Regime detection, we built a strategy router:
| |
During TRENDING:
- Pipeline (trend-following) ✅
- BB Squeeze (breakout) ✅
- MACD Divergence (reversal) ✅
- Mean Reversion ❌ paused
During RANGING:
- Pipeline ❌ paused
- BB Squeeze ✅
- MACD Divergence ✅
- Mean Reversion ✅
The benefit: strategies never run in the wrong environment. Pipeline won’t force trades during consolidation. Mean Reversion won’t fight trends.
Transition Prediction: Know Before It Happens
Beyond real-time detection, we added ADX trend analysis. By tracking ADX rate of change over 48 hours, we can predict regime transitions:
- ADX rising from 15 → 18 → 19: Ranging is ending, trend strategies preparing to activate
- ADX dropping from 28: Trend is weakening, prepare to switch to ranging strategies
Our recent observation: SOL’s ADX rose from 14 to 20.5 over 48 hours, successfully predicting the RANGING → TRENDING transition. The system started preparing when ADX approached 20, and activated Pipeline immediately after the crossover.
Real Data
We scan six major coins (BTC, ETH, BNB, XRP, SOL, DOGE) for Regime state.
Full-market ranging period (late Feb — early Mar):
- All 6 coins ADX < 20
- Pipeline signals: 0 (correct behavior — don’t trade in the wrong environment)
- Mean Reversion + BB Squeeze filled the gap
SOL/DOGE entering trend (March 5):
- SOL ADX 20.5, DOGE ADX 20.4
- Pipeline LONG signal proximity at 85%
- Other coins also rising (BNB estimated to break through in ~52h)
Coverage Improvement
Before vs. after adding Regime detection and strategy switching:
| Metric | Before | After |
|---|---|---|
| Strategies available in ranging | Only Pipeline (no signals) | MR + BB + MACD |
| Symbol × Direction coverage | 2/12 | 10/12 |
| Signal frequency during ranging | ~0/day | 2-4/day |
Tips for Building Regime Detection
1. Start with ADX alone
You don’t need all four indicators. ADX alone distinguishes trending from ranging. Just implementing “pause trend strategies when ADX < 20” is a massive improvement.
2. Backtest by Regime
Many people look at overall win rate. But a strategy might be 90% WR in trends and 30% in ranging, averaging to 60% that “looks okay.” Segment by Regime to find the real issues.
3. Add a cooldown to switching
Don’t flip-flop when ADX hovers around 20. We require ADX to stay above/below the threshold for 2 candles as confirmation, preventing false signals from causing rapid strategy toggling.
4. Don’t fully trust automation
Regime detection is a tool, not gospel. Black swan events (major news, policy changes) can instantly shift market state, but ADX needs time to react. Keep human oversight.
What’s Next
- Develop finer Regime sub-categories (weak trend, strong trend, narrowing range, widening range)
- Use machine learning to predict Regime transitions
- Dynamically adjust position sizes per Regime (larger in trends, smaller in ranging)
- Backfill historical Regime data to validate strategy performance across different environments
This is the third article in our “Quant Trading in Practice” series. Previous articles: The 100% Win Rate Lesson, Position Sizing: The Hidden Engine.