Is there any interest in EA performance statistics?

New comment
 
I'm wondering if there is any interest in collecting performance statistics on the collection of free / demo expert advisors available on metatrader. I have been looking around the internet for this kind of information and have not found it. My idea is that I could set up an automated system that would run the strategy tester in optimization mode and report all of the results for each EA. Optimization would be done over the past month of historic data for the major currency pairs and the result set would be optimizing on the account balance. The genetic algorithm would be turned off. This process would be repeated once all of the advisors have been tested, or two weeks have passed, whatever time is longer. I believe it would give everyone a good idea of which bots are performing well in current market conditions, and we could analyze result sets over time in order to find general strategies that work in various market conditions. With this data, we could then work on bots that have a statistically better chance of providing long term success, or bots that specialize in maximizing profits under specific market conditions.

I'm not kidding, I can create a system that does this. For my day job I am a software automation engineer. I write automated tests for Windows desktop applications and websites, that exercise the user interface and ensure that the applications are doing what the developers intended. I also create systems for automating the environment preparation for testing, cleaning up files, installing applications, configuring the applications under test, and creating the data that the application will use (sql databases, files, etc.). I also create job schedulers and monitoring applications, so that I can just describe the tests that I want the system to run, then wait patiently for results. Parsing results files and writing the results into a database is part of the process. I'm certain I can come up with a system for analyzing EA bots.

I'm going to analyze bots for myself, and it is going to be a lot of work to create this system. This is why I'm asking if anyone else is interested in this kind of data, I would like ideas on what people might want from a system like this before I start building it. My plan is to try finding resources for running the tests and hosting results, that won't cost me any money. I'll also host the source code for all parts of this system on github and allow contributions if anyone would like to help. Ideally the system will just run without any intervention, the virtual machines required to run the tests will be free, and web hosting for results will be free. If I can get all of that together, then the data reported by this system will be freely available to anyone. Of course if it takes some time to maintain the system every week, the hosting isn't free, or the deployment of virtual machines costs money, then I'd need some compensation before granting access to the data.

I can handle the hard part of putting it all together. Let me know if you're interested in this, have ideas on what kind of data you would like to see, or if you have any resources that you would like to donate to the project.

Thank you
 
Everyone is free to waste his time as he likes.
 
Alain Verleyen:
Everyone is free to waste his time as he likes.
How is it a waste of time? Most of the bots don't have any testing information, and I don't know of any analytical studies on automated strategies with respect to market conditions. This would facilitate that. Please tell me specifically how this is a waste of time, maybe I'm missing some information about the testing tools that need to be addressed before collecting this kind of information would be valuable?
 
Matthew Colter:
How is it a waste of time? Most of the bots don't have any testing information, and I don't know of any analytical studies on automated strategies with respect to market conditions. This would facilitate that. Please tell me specifically how this is a waste of time, maybe I'm missing some information about the testing tools that need to be addressed before collecting this kind of information would be valuable?

My opinion is you will not find any serious EA on Internet ("collection of free expert advisors available on metatrader"). So while your project is interesting from a technical point of view, it's waste of time on a practical side.

But I wish you good luck, and you will hopefully demonstrate I am wrong.

 
Alain Verleyen:

My opinion is you will not find any serious EA on Internet ("collection of free expert advisors available on metatrader"). So while your project is interesting from a technical point of view, it's waste of time on a practical side.

But I wish you good luck, and you will hopefully demonstrate I am wrong.

Oh yes, I do believe that the free bots will turn out to lose money all the time, so we do agree. There are paid bots that allow free demos too, so I'd include those as well, and some of them might do alright. I'm sure that there are market conditions favorable to some of the free bots as well, so in months that they show profit we could look at the data they ran on and determine those conditions, then take the algo into account for edge cases or multi strategy bots.
New comment