Bayesian Trend Filter
Statistical probability filter for trend vs mean reversion
Overview
The Bayesian Trend Filter is a sophisticated statistical tool that uses Bayesian probability analysis to determine the likelihood that a market is in a genuine trend versus a mean-reverting state. Rather than using fixed indicators, it updates probability estimates dynamically based on observed price behavior and volatility patterns.
This filter applies Bayes' theorem to evaluate the posterior probability of trend continuation given current market data. A trader with information (prior beliefs) updates probability estimates as new price data arrives, allowing real-time assessment of market regime. High probability scores suggest strong trending conditions favoring momentum strategies, while low scores indicate choppy or mean-reverting conditions preferring range-trading tactics.
Quantitative and algorithmic traders use Bayesian filters to avoid whipsaw trades by switching between trend-following and mean-reversion strategies based on statistical evidence. It's particularly useful in portfolio strategies where different markets may be in different regimes simultaneously.
Formula
Bayesian probability is updated using observed price action:
Parameters
| Parameter | Type | Default | Description |
|---|---|---|---|
| Lookback Period | Integer | 20 | Number of bars to analyze for probability calculation |
| Volatility Period | Integer | 14 | Period for calculating standard deviation of returns |
| Trend Threshold | Decimal | 0.65 | Probability threshold for considering market in trend (0.5-0.95) |
Common Use Cases
1. Strategy Regime Switching
Automatically switch between trend-following strategies (when probability high) and mean-reversion strategies (when probability low) based on statistical evidence.
2. Whipsaw Protection
Avoid entering momentum trades in choppy markets by waiting for probability to confirm trend conditions; reduces false breakouts.
3. Portfolio Diversification
Monitor trend filters across multiple assets simultaneously; allocate capital to those with highest trend probability to optimize risk-adjusted returns.
4. Position Sizing Optimization
Increase position size when trend probability is high and confidence is maximum; reduce when probability drops to low/moderate levels.
Advantages & Limitations
✓ Advantages
- Quantified Confidence: Converts market assessment into probability scores enabling objective decision-making.
- Adaptive to Regime Changes: Continuously updates probability estimates as new data arrives, naturally detecting regime shifts.
- Multi-Asset Ranking: Compare trend probability across multiple charts to identify strongest opportunities.
- Algorithmically Tradeable: Clear numerical output enables systematic implementation in trading algorithms.
! Limitations
- Prior Assumption Dependent: Results depend on initial prior probability assumptions; wrong assumptions distort probability estimates.
- Lag During Breakouts: Probability accumulates slowly, causing lag during sharp trend initiations.
- Black Swan Vulnerability: Extreme moves violate underlying distribution assumptions causing unreliable probability estimates.
- Complex Calibration: Requires substantial historical data and statistical expertise to calibrate effectively.
Tips & Best Practices
⚡ Calibrate on Historical Data
Backtest filter parameters on 2+ years of historical data for your specific asset to find optimal threshold settings and lookback periods.
📊 Combine with Momentum
High trend probability + increasing momentum = strongest trade setup. Use RSI or MACD as secondary confirmation.
🔄 Watch Probability Transitions
Cross above threshold = potential trend initiation signal. Cross below = regime change warning. Transition points often precede large moves by 1-3 bars.
⚠️ Account for Market Context
Filter may be unreliable during major news events, earnings, or Fed announcements. Reduce position size or avoid trading during these high-volatility windows.
Example Strategy
1. Setup: Probability Crossover
Apply Bayesian filter to your chart with threshold set to 0.65. Monitor when probability crosses above/below threshold to identify regime changes.
2. Entry: High Probability Trades
Buy when probability crosses above 0.65 AND close > 20-period MA. Sell when probability drops below 0.35 AND close < 20-period MA.
3. Stop Loss: Probability Reversal
Set stop loss when probability drops below 0.55 in long trades or rises above 0.45 in short trades, signaling regime weakening.
4. Target: Probability Divergence
Exit when probability reaches 0.85+ (trend exhaustion) or bank profits at previous swing high + 5% with trailing stop at -2% from entry.