Hello fellow traders and programmers,
I'm interessted in your opinion and would like to know how you evaluate a EA. What are the important requirements that a EA must fulfill?
Of course I want to know that because I need help with my own EA... Also, I dont know how much I can trust the backtests that I have made, due to some problems I had exporting from tickstory.
I have a EA that is meant to trade on anything! For now I have tested with some forex pairs, market indexes and metals. Indicator period and some values like TP, SL and risk have to be optimized for each symbol. The results are always similar and good enough for my standards... But while I was backtesting and optimizieng, I noticed that when I connect to the broker server, on some symbols the input values would not work. For example the TP and SL values were way to high (or low), or the lot sizes were just wrong. Now, my fear is that after I optimize my intial list of 30 symbols and go on to the demo account, all the optimization could be meaningless because of some mistake I made exporting from tickstory (entering the values for each symbol) or some other mistake that I surely made.
I also fear the so named "curve fitting". I always backtest the most time I can get out of tickstory, that is at least 3 years and sometimes 4 or 11 (eur/usd). Because of the nature of my EA, I can't always do "walk forward" optimization. That is because on some symbols my EA only opens a few orders a year. Also, if it would be necessary, I could re-optimize every couple of month...
Because I wanted to know how you evaluate EA's, I will post some information about mine (results are still not final):
Using only one symbol, the EA won't make to much profit. I tried very hard finding a way to open more orders and get rich in 2-3 years (lol), but then I realized that what I had was not trash, because I could use it on multiple symbols increasing dramatically the amount of orders without affecting the winning rate.
What still concerns me a lot is that when I change the period of my indicator, the results can turn around completely. So why would the indicator period have so much influence?
Please tell me how to evaluate this kind of EA and give me your opinions, I'll appreciate it very much.
What are the important requirements that a EA must fulfill?
Hey Marco, thank you for your answer.
Apart from confidence, what would motivate you to invest your time working on your EA? Do you keep spending your time working on a EA until it is profitable? For me it is a long way from one idea to the tests on real account :(
It's results that motivate you, these are gathered by a combination of 1. Programming and 2. Live trading = Gaining knowledge about financial markets, and developing software to eventually combine the two.
I started off writing simple robots, tested them on the back tester, then discovered that results on live markets were very different.
Then i realized you need to know or find the loop holes first only then you can digitize the model, it will not be a combination between signals from indicator x and y.
Make sure to trade a lot so you can try various things and at some point you find the Glitter.
By that time your programming skills will be developed up to a point where you don't even have to think about implementing what you just found, into a Robot.
A little bit of insanity, can also help.
@ Marco - Only a little bit? I am in trouble. :p
Basically Marco covered it in first response. That the EA be profitable (to me that means makes you more money than you lose from trades going bad, commission, drawdowns, and so on) over the long haul. A better EA would be one that can adapt to different market conditions, i.e. know when to trade and more importantly when NOT to trade. How you can tell is not so easy at first. Also, in my mind anyway, ideally, you should be able to use the same EA on different time frames, and different traded items (different currency pairs, and possibly different main markets, like commodities, stocks, etc.)
But, people have been looking for that holy grail for trading for a long time, and if it has been made, then the person or people who did it aren't likely to share, as it would mean everyone else could make as much money as they do, and being that trading is ultimately a zero-sum equation (one buys and one sells), then it would not be profitable to them for long. In theory, that would eventually flatten any money being able to be made.