Over-fitting is a common issue in developing Expert Advisors (EAs) where the EA performs exceptionally well in historical backtests but fails to deliver similar results in live trading due to an over-optimization to past data. It often results from the EA being too tailored to historical market conditions, lacking generalizability for future scenarios.
Live trading performance: https://www.mql5.com/en/signals/1801317
Key Points Regarding Over-fitting:
- Common Problem: Many EAs suffer from over-fitting, either because developers are unaware of its existence or intentionally use it to present impressive backtest results.
- Avoiding Over-fitting: Traders are advised to avoid using EAs without live trading results for at least 5 months or 300 tracked trades. Live trading results provide insight into the EA’s performance in unforeseen market conditions.
- Generalizability: EAs should be designed to have generalizability, allowing them to adapt to different market conditions rather than being overly specific to historical data.
- Live Trading Monitoring: Monitoring an EA’s live trading performance is crucial to observe how it performs with data it has never encountered before.
Boring Pips EA’s Anti-Overfitting Process:
- Initial Optimization: The EA undergoes optimization using historical data from 2010 to 2019 to test the initial premise of the trading strategy and extract robust parameter values.
- Walk-forward: Parameters from the first stage are tested using entirely new data from 2019 to 2022 to ensure the system’s stability and evaluate its predictive power.
- Stress Testing: Robust parameter values undergo stress testing, introducing variables like noise and lag to evaluate the system’s tolerance to random factors.
Live Trading Performance: Boring Pips EA’s live trading performance has been monitored with a real account since October 10, 2022. The performance can be tracked here.
Boring Pips EA Algorithm Overview:
- Trading Strategies: Combines momentum, supply and demand zones, Fibonacci retracement, and artificial intelligence algorithms.
- Signal Scanning: Continuous scanning of potential supply and demand zones using an advanced algorithm.
- Momentum Analysis: Utilizes deep learning algorithms to measure momentum across four timeframes simultaneously.
- Automated Trading Decisions: Makes trading decisions based on changes in momentum at identified supply and demand zones.
- Risk Management: Manages trades based on probability distribution rules to maximize exploitation of the trading edge.
Installation and Recommendations:
- Pairs and Timeframe: Recommended for AUDCAD, AUDNZD, and NZDCAD on the M5 timeframe.
- Risk Management: Choose an appropriate risk mode (Boring, Low risk, Medium risk, or High risk).
- Base Balance: Allocate a specific amount of balance for trading.
- Personalization and Risk Management Settings: Follow detailed instructions for customization and risk management.