Walk-Forward Analysis

What is Walk-Forward?

Walk-forward analysis is a robust optimization technique that:

1. Optimizes on a historical window (in-sample)
2. Tests on the next period (out-of-sample)
3. Rolls forward and repeats

This simulates how a strategy would perform if you re-optimized periodically.

How It Works

|-------- In-Sample --------|-- Out-of-Sample --|
|     Optimize here         |    Test here      |
         ↓ Roll forward ↓
         |-------- In-Sample --------|-- OOS --|
                    ↓ Roll forward ↓
                    |-------- In-Sample --------|-- OOS --|

Configuration

Parameter	Description	Typical Value
In-Sample Period	Optimization window	2-3 years
Out-of-Sample Period	Validation window	6-12 months
Step Size	How far to roll forward	Same as OOS

Running Walk-Forward

1. Go to Strategies > Optimize
2. Select Walk-Forward method
3. Configure windows
4. Run optimization

Tip

Use at least 3-4 walk-forward periods to get meaningful results.

Interpreting Results

Efficiency Ratio

Compares out-of-sample to in-sample performance.

• > 0.5 = Good (OOS performance is at least half of IS)
• > 0.7 = Excellent
• < 0.3 = Strategy may be overfit

Consistency

How often the strategy is profitable in OOS periods.

• 4/4 periods profitable = Very robust
• 2/4 periods profitable = Questionable

Degradation

How much performance drops from IS to OOS.

• < 30% drop = Normal

• 50% drop = Likely overfit

Benefits

• Realistic Performance Estimates - Closer to live trading
• Detects Overfitting - Overfit strategies fail OOS
• Parameter Stability - Shows if optimal params change

Next Steps

• Best Practices - Optimization guidelines
• Backtesting - Test final parameters