Forex Automated Testing Project
This is a project to find and run profitable forex expert advisors.
I already have a head start as I have been writing code for forex robots for quite a while and have been learning what works and what does not.
In an effort to keep this really simple the testing process is outlined below.
1.Back test - Back testing is fairly simple. i use Dukascopy tick data for backtesting to try to get a better quality result. While the quality is better I still dont trust a backtest to give a true indication of performance. It is just the best guide I can get.
2. Forward test - for those ea's that look promising I will forward test them. the first forward test will be simply putting them on a demo account and letting them run. I will usually use multiple currency pairs to see which perform.
These forward test posts will have Test 1 in the title
These forward test posts will have Test 1 in the title
3. More forward testing - taking the best performing currency pairs they will be forward tested with what i think are better parameters. I can use a control test with the original set to check their performance.
These forward test posts will have Test 2 in the title
For an ea to get to this stage it has to have performed consistently.
These forward test posts will have Test 2 in the title
For an ea to get to this stage it has to have performed consistently.
4. Analyzing the data - The analysis after the testing should give a good indication of the performance of the ea and what are the best settings.
Testing Time
pretty simple? Yes, just alot of time to test. which brings me to the next point, which is what time periods should the ea's be tested. Obviously the more time they can be tested the better, but there are a couple of problems with this.
1. the market can change - Changing market can make an ea not perform as it did in the past.
2. Time - It takes alot of time to forward test thoroughly.
3. Ea's - Ea's are all different. Some make lots of trades, some make only a few per year. Because they are all different, testing times will have to vary.
I will have to use my discretion in testing times. I want to test them rather quickly, but i need to get good results. I am not too worried about testing too quickly as even after they have been tested I can still put them on a demo account to see how they perform
Lot and Account Size
The forward tests generally use small lot sizes 0.01 and are in a $10,000 account.
I am not trying to optimize the earnings, as lot sizes can be scaled to suit an excepted drawdown.
Some of the tests look really bad. Why dont you stop them?
True some of the tests bomb out really badly. Some just get off to a bad start.Some tests look bad because some pairs perform really badly, but other pairs perform really well.
Sometimes it is worth analyzing the tests to see if the ea can be improved.
