I have experienced this backtester anamolie now on three different EA's. First the Gogetter EA, Then the multilotscalper EA. Now the cyberiatrader EA.

one of my tests on the multilotscalper EA went to $500 to $11.8 million. however the backtester only did this once and has not repeated this result again. Now the best it does is 4.8 million AFTER making significant changes to the code..When I was testing the identical code it would only do 4.1 milliion every time thereafter that it did the $11.8M. Mind you this is using the same settings for both trials and the same date ranges.

What this all boils down to is that the backtester doesn't return the same result each time on the same test. I can find no variable which I have changed to account for it's different results. I therefore conclude that it is an internal instability of the backtester itself which metaquotes themselves refuse to acknowledge.

The effect of this is that it destroys any confidence in any backtest result no matter how many times it 'does' repeat the test and return the same result. If the same test is repeated enough times the backtester will fail eventually to return the same result. It may take a week or so and repeating the same test on the same series a few dozen or more times but it WILL breakdown. At least it has broken down on my testing now with 3 different EA's.

It would be nice to get some acknowledgement of this problem from metaquotes, but they have deflected every attempt I have made to bring this to their attention. They simply do not want to address it and so alas the problem persists.

I was encouraged when the cyberiatrade seemed to be immune to this problem. It repeated the test with the same result more times than any of the other EA's however today it has shown me that it is just as vunerable to the backtester instability as the others. I am most disappointed to see this happen. I almost believed that the problem might have been in the EA code of the other programs.

Now I have seen this occur with three different EA's. I can hardly believe that it has anything to do with the code of any of them. No I think this proves that the problem resides in the backtester itself.

Furthermore it makes it impossible to stabalize any EA's performance when the backtester doesn't run the EA's reliably and returns differing results on the same criteria and data.

It's all just one big russian roulette game until this instability is resolved.