What is this code?
This code is acustom optimisation function for the MetaTrader 5 Strategy Tester. It is not an Expert Advisor, indicator or script in the classical sense, but a special script for analysing test results.
How the code works:
1. Data collection
-
Gets the history of trades from the tester
-
Checks minimum requirements (at least 50 trades)
-
Determines initial deposit and time periods
2. Splits the data
-
Splits trades into two periods:
-
In-Sample (IS) - the first 70% of the testing period
-
Out-of-Sample (OOS) - the last 30% of the period with a 1-day interval
-
3. Calculation of metrics
Calculates a set of metrics for both periods:
-
Profitability and drawdown
-
Sharpe and Sortino ratios
-
Profit factor and probability of profitable transactions
-
Statistical indicators (skewness, kurtosis)
-
Special metrics (Serenity Ratio)
4. Statistical analysis
-
Compares IS and OOS distributions using the Kolmogorov-Smirnov test
-
Checks normality of distributions using the Jarque-Bera test
5. Strategy evaluation
Creates a comprehensive strategy evaluation that takes into account:
-
Profitability (30%)
-
Consistency of results (30%)
-
Risk-adjusted performance (25%)
-
Statistical quality (15%)
Where to use this code:
1. strategy optimisation
-
Place the code in the MQL5/Scripts/ folder.
-
In the strategy tester, select "Custom optimisation criterion"
-
Use this script to evaluate the results of optimisation
2. Strategy validation
-
Use to validate the stability of the strategy
-
Analyse discrepancies between IS and OOS periods
-
Identification of over-optimised strategies
3. Comparison of strategies
-
To objectively compare different strategies
-
To rank strategies according to a comprehensive criterion
Advantages of the approach:
-
Minimising over-optimisation through data partitioning
-
Comprehensive evaluation takes into account múltiples aspects of performance
-
Statistical verification of the robustness of the results
-
Automatic screening of non-viable strategies
Important Notes:
-
Requires a significant number of trades (at least 50)
-
Need sufficient historical period to split into IS/OOS
-
Code returns -DBL_MAX for strategies that do not pass minimum requirements
This approach is particularly useful for traders and developers who want to take testing and optimising trading strategies seriously, minimising the risk of fitting to historical data.
