DAY 19: Update and Maintenance of EA — The Need for Continuous Refinement

Last time (DAY 18), we looked at a hybrid operation combining discretionary trading and EAs.
By not just running the EA, but also applying discretionary judgment as needed, the possibility of flexibly responding to sudden market changes and fundamental factors increases.

However, an EA is not something you can use forever once it is completed.
Today's (DAY 19) theme is,EA updates and maintenance.
Let's organize the perspective on modifying the logic to maintain and improve performance in response to changes in market conditions and advances in technology.

Why do EAs need maintenance

1. Changes in the market environment

   • The FX market is constantly changing. There are times when liquidity increases and spreads tighten, and there are also periods with sudden volatility increases or changes in monetary policy that cause price movements not seen in the past.

   • A logic or parameters that worked before are not necessarily effective in another market.

2. Detection and correction of overfitting

   • EAs that were overly tuned during backtesting tend to perform poorly in future markets. It is important to run while monitoring performance and adjust as needed.

3. Advances in technology

   • There are ongoing updates to MT4/MT5 versions, changes in OS environments, etc. If the EA's code remains old, issues can arise.

具体的なアップデート・メンテナンス方法

1. Regular backtesting and forward testing re-run

   • About every few months to half a year, backtest using the latest market data andcheck whether the performance has changed significantly.

   • Also verify the forward test results on real or demo accounts, and compare with backtesting to notice the aging of the logic.

2. Fine-tuning parameters (risk management, stop loss width, lot settings, etc.)

   • Has the market shifted to a trend-following regime, or does it stay in a range?

   • Even a small adjustment to stop loss width or risk-reward can improve performance.

   • However, if you tinker too frequently due to short-term market noise, it can introduce more variance. Aim for a reasonable frequency.

3. Adding new logic or filters

   • Adding filters such as news filters or volatility filters to adapt to external factors and market conditions makes it easier to avoid unfavorable markets.

   • It is also valuable to consider updating the EA by embedding insights gained from discretionary trading.

4. Code maintenance and version control

   • Regularly review the EA source code (MQL4/MQL5) to optimize and fix bugs.

   • Keep a record of changes (version control); it makes performance comparisons with previous versions smoother.

How to determine the timing of updates

• When drawdown exceeds an acceptable range
  Predefine a criterion such as “if maximum drawdown exceeds ○%, it requires verification,” and consider updates or shutdowns in case of rapid losses.

• Periodic checks every few months
  Even if there is no major deterioration, make a habit of reviewing backtest results and forward data quarterly or semi-annually to decide on maintenance.

• When market structure changes are detected
  If a long-term trend begins to change, there are major political/economic changes, or volatility changes significantly, consider whether the EA can continue to adapt.

Beware of excessive updates

Maintenance is important, butupdating too muchcan lead to a reproducibility-weak EA.
For example, making frequent adjustments to code or parameters after every small loss can lead to a patchwork state where there is little time to re-evaluate backtest results, increasing the risk of overfitting further.

• Validate data and periods separately (walk-forward analysis, etc.)
  In regular maintenance, using walk-forward analysis to verify parameters optimized in the past against future markets helps suppress overfitting risks somewhat.

• Divide into main version and experimental version
  Keep a stable version and test improvements in an experimental version; if successful, reflect the results in the main version.

Today's summary and next episode

• An EA is not something you set up once and forget for eternity;you need to update and maintain it in response to market conditions and technology changes.

• Excessive updates or frequent modifications can be counterproductive, soset quantitative criteria and validation rulesto feel secure.

Next (DAY 20) will discuss, "**How to identify suspicious EAs and bad EAs**" as the theme, and share concrete points on how to spot excessive advertising and disguised EAs in the world. For those considering purchasing or implementing an EA, this is content you should know.

Introduction to EAs I sell

If you're interested, please also take a look at the EAs I sell.
https://www.gogojungle.co.jp/users/147322/products

Choosing an EA with regular updates and clear maintenance policy may lead to long-term stable operation.

In the next article, we will整理チェックポイント to avoid being deceived by \"suspicious EAs\".
Pleasepress "Read more" to deepen your knowledge.