Market Regimes
No single strategy works in all market conditions. Understanding regimes is essential for building a robust trading system.
The Three Regimes
| Regime | Characteristics | Strategy impact |
|---|---|---|
| Bull (trending up) | Higher highs, higher lows, positive momentum | Trend-following profits, long bias works |
| Bear (trending down) | Lower highs, lower lows, negative momentum | Long-only strategies suffer, shorts are viable but risky in crypto |
| Sideways (range-bound) | No clear direction, mean-reverting | Trend-following gets whipsawed, worst environment |
Why This Matters for Gordon
Gordon is a trend-following system. It makes money when trends exist and gives back some when they don't. The key insight: you don't need to predict regimes — you need to survive them.
How Gordon Handles Regime Changes
Multi-strategy diversification — different strategies respond differently to regimes. Supertrend (D1-8h) and EWMAC (6h-1h) are partially uncorrelated.
Sentiment overlays — the Fear & Greed index and SSR provide early signals of regime shifts. Gordon uses these as sizing modifiers, not binary filters.
Volatility targeting — automatically reduces position size in volatile (often transitional) periods and increases in calm (often trending) periods.
Circuit breakers — when drawdown exceeds thresholds, trading halts. This prevents catastrophic losses during regime transitions.
Regime Detection Approaches
HMM (Hidden Markov Model)
Gordon-lab's v1 research used HMM to classify regimes. Results: strong statistical metrics (AUC 0.81) but negative equity in simulation. The lesson: detecting regimes doesn't automatically translate to profitable trading.
Sentiment-Based Detection
Gordon's current approach uses observable market data:
| Indicator | What it measures | Regime signal |
|---|---|---|
| Fear & Greed Index | Retail sentiment | Extreme fear = potential reversal, extreme greed = caution |
| Funding rates | Leverage positioning | High positive = crowded longs, high negative = crowded shorts |
| OI vs price | Positioning health | Divergence = fragile trend |
| Correlation density | Herd behavior | High correlation = crowded trade, vulnerable to unwind |
Volatility as a Regime Indicator
BTC realized volatility has declined structurally:
| Period | Annualized Vol |
|---|---|
| 2013 | 7.58% |
| 2020 | 3.41% |
| 2025 | 2.24% |
This compression is driven by ETF institutionalization. It affects all trend-following strategies because lower volatility means smaller moves to capture.
Multi-Strategy as Regime Insurance
Rather than trying to predict regimes, Gordon combines strategies that perform differently across conditions:
| Strategy | Best regime | Worst regime |
|---|---|---|
| Supertrend | Strong trends (any direction) | Choppy sideways |
| EWMAC | Sustained momentum | Quick reversals |
| PSAR | Trending with pullbacks | Ranging markets |
| Donchian | Long trends with clear breakouts | Sideways compression |
When combined into a portfolio, the losses from one strategy's bad regime are partially offset by another strategy's good regime. This is the fundamental argument for multi-strategy diversification.