Learn how to systematically improve strategy performance through parameter optimization, walk-forward testing, and avoiding overfitting.
Time required: 30-60 minutes
Prerequisites: Basic strategy with backtest results
Step-by-Step Optimization Process
- Identify the parameter to optimize
Start with one parameter (e.g., RSI oversold threshold). Never optimize multiple at once.
- Define test range
RSI oversold: Test from 20 to 40 in steps of 5. This gives you 5 backtests: 20, 25, 30, 35, 40
- Run backtests
Test each value on same symbol and date range. Record Sharpe ratio for each.
- Compare results
Which value gave best risk-adjusted returns? Not just highest return, but best Sharpe.
- Validate out-of-sample
Test optimal value on different time period or symbols. Does it still work?
Walk-Forward Testing
The professional way to optimize without overfitting:
Walk-Forward Method:
- In-Sample Period (Training): 2020-2022 - Optimize parameters here
- Out-of-Sample Period (Testing): 2023 - Test with optimal parameters
- Validation: If performance holds in 2023, strategy is robust
- Roll Forward: Repeat with 2021-2023 training, 2024 testing
Why this works: Tests if strategy works on unseen data, not just data you optimized on. Prevents curve-fitting.
Avoiding Overfitting
❌ Overfitting Warning Signs
- Perfect or near-perfect results on backtest
- Strategy has 10+ parameters all finely tuned
- Works great on one symbol, fails on all others
- Great 2020-2021, terrible 2022-2024
- You tested 100+ parameter combinations
✅ Robust Strategy Signs
- Works across multiple symbols
- Consistent across different time periods
- Simple with few parameters (<5)
- Logical, explainable edge
- Out-of-sample performance similar to in-sample
