EA Factory
In developmentIn development – currently not available.
From a trading idea to a testable trading system
The EA Factory is a research, development and testing infrastructure for algorithmic trading systems in MetaTrader 5, developed by me.
It is designed to combine strategy research, MQL5 development, automated backtests, parameter analysis, robustness testing and structured evaluation in one continuous process. What is automated so far: the test series in MetaTrader 5, the import of their results and their evaluation.
The Factory is not tied to a particular strategy type. Which markets and time levels are examined depends on the verified price data available; further markets are added as soon as the data supports it.
Strategy research
An infrastructure for strategy research
The EA Factory was not built to develop a single Expert Advisor.
It was built to research new trading ideas again and again, to derive variants from them and to test those against each other under reproducible conditions.
A first hypothesis can turn into different entry logics, filters, exit models, risk models and parameter combinations. These variants are then meant to be implemented, tested automatically, evaluated and filtered.
The goal is not a single backtest, but a repeatable research process.
Process
- 01
Hypothesis & research
It starts with a concrete trading idea or market hypothesis. AI is meant to help to develop new approaches, to question existing ideas critically and to derive different testable variants from one basic idea.
- 02
Strategy development
The hypothesis becomes a clearly defined trading logic. Entries, filters, exits, risk rules and further strategy building blocks can be combined systematically and examined as different variants.
- 03
MQL5 implementation
The defined rules are turned into executable Expert Advisors, compiled and checked technically. That is what turns a theoretical idea into a trading system that can actually be measured.
- 04
Automated backtesting
The Factory is designed to test strategy variants automatically across markets, timeframes, historical periods and parameter ranges. Which markets are enabled depends on the verified price data available; a pilot with a few markets is currently running.
- 05
Robustness testing
A single good backtest is not enough. Strategies that look interesting can therefore be examined further under changed conditions — with different parameters, for example, or with changed market and execution conditions.
- 06
Evaluation & selection
Results are collected centrally, made comparable and evaluated against defined criteria. Approaches that are unstable, heavily overfitted or convincing in only one single configuration should be sorted out as early as possible.
The interesting strategy is not the one with the prettiest backtest — it is the one whose edge holds up under changed conditions.
Open research architecture
Open to new markets, timeframes and strategies.
The EA Factory is not tied to one strategy type. Which markets and time levels are enabled depends on the verified price data available. A pilot with a few markets is currently running.
The infrastructure is built so that new research questions can be taken up. Examples:
- Does the same logic work on different markets?
- How does an effect change between different timeframes?
- Which variants of a trading idea stay stable?
- Which filters actually improve a strategy?
- In which market phases does a logic work — and when does it not?
- How sensitively does a system react to changed conditions?
- Is an observed edge structural — or only the result of a single optimised configuration?
So the Factory is not only there to test existing strategies.
It is meant to be an infrastructure for researching new strategies systematically.
Scalable research
From one idea to a whole research space
A single trading idea can be turned into a very large number of possible variants.
- Different entries
- Different filters
- Different exits
- Different timeframes
- Different markets
- Different risk models
- Different parameters
- Different market phases
Done by hand, this research space quickly becomes too large.
The EA Factory is meant to automate large parts of this recurring technical process and to make the variants testable and comparable in a structured way. So far the MetaTrader 5 test series and their evaluation run automatically.
That is meant to let the research concentrate more on the question that actually matters:
Which trading logic is worth pursuing at all?
AI-assisted research
AI is meant to extend the research process — it does not replace it
AI is meant to be a tool inside the EA Factory for research, development and analysis. AI-assisted optimisation is planned and not running yet. The aim is that it helps, among other things, to:
- develop new strategy hypotheses
- vary existing approaches systematically
- examine new combinations of strategy building blocks
- develop and check MQL5 code
- analyse larger amounts of test results
- spot anomalies
- derive new research directions
The decision whether a strategy is actually interesting is not made by the AI. What counts for that are the measurable results from backtests, comparison tests and robustness checks.
AI is meant to increase the speed and the breadth of the research process. The verification stays data-based.
Technical highlights
Open market architecture
The architecture is designed to take in new markets as soon as verified price data is available.
Multi-timeframe research
Strategies and hypotheses can be examined across different time levels.
AI-assisted strategy research
AI is meant to support hypothesis building, strategy variation, development and analysis.
Automated MT5 pipeline
Development → automated test series → evaluation
Robustness
An edge has to survive more than one perfect setting
A system that only works with one particular parameter set, one single market or one single historical phase is considerably less interesting than a logic whose behaviour stays stable under change.
Strategies can therefore be examined across different markets, timeframes, historical periods, parameter ranges and changed market and execution conditions.
The central question
Is the observed edge reproducible — or was merely a good historical configuration found?
Why I built it
Manual strategy development turned into infrastructure
As the number of trading ideas, markets, timeframes and parameters grows, manual development quickly becomes the bottleneck.
Instead of running the same technical sequence by hand again and again, I built my own infrastructure out of it.
The Factory automates recurring parts of this process, above all test runs in MetaTrader 5 and their evaluation. It is still in development.
The aim: to examine more ideas, recognise weak approaches earlier and scale research more systematically.
Not investment advice and not an invitation to trade; past results are no guarantee of future results. Trading carries the risk of losing money.
Technologies
- MQL5
- Python
- MetaTrader 5
- Automated Backtesting
- Parameter Optimization
- Robustness Testing
- Automation
The EA Factory is not an Expert Advisor. It is the infrastructure behind one.
It was built so that trading ideas are not only programmed, but researched, tested, compared and checked under changed conditions in a systematic way.
The aim is to turn a manual development process into a reproducible research workflow.