Download MetaTrader 5

New Idea For EA opmization

To add comments, please log in or register
Nesimeye Oswald
54403
Nesimeye Oswald  
I have this idea I want the community to reason. The idea is for optimizing EAs and I call it trade SWITCH. Note this is unlike the contemporary trade reversal. It has come to my notice that every EA behaves in a certain way to every time frame and that behaviour in their number of consecutive losses and gains. So now supposing we decide to switch a trade in a role of consecutive losses so that it becomes profitable instead of ending in a loss, I believe this would optimize an EA greatly. For example, we know an EA is like to lose 3 times consecutively in a row, then immediately we record the first loss, we reverse the next two trades to opposite of their initial trade pattern so that they end in gains. The problems now is I don't think average consecutive loss from strategy tester result is sufficient in giving the actual consecutive losses. So my question is, how do I clearly define consecutive losses?

Thanks
JD4
1101
JD4  

I see a potential problem with this that might not happen immediately, but is likely to pop up sooner or later.  An EA needs to be constantly supervised, because unless it is self adapting (neural network style) then it will eventually need to be tweaked.  Not every EA can work with every market condition, and you will end up losing money.  The trick here would be to have good MM, and if your EA is constantly doing the 3 trade thing, then you need to either tweak your EA, or get another one.

IMO, a better solution would be to find the conditions that are causing something like that to happen to your EA, and adjust the EA.  Test it for the long haul to make sure it works (for $ value) making more than it loses, because you will lose some trades.  It is the only 100% guaranteed thing to happen in trading.  It would be better to plan for winning trades anyway, and have a good way to limit losses before they wipe out your account.

Enrico Lambino
4358
Enrico Lambino  

Strategy tester reports consecutive losses in terms of average and maximum streak.

In your example, if an EA likes to lose 3 times, that is just the average. It can lose 1, 2, 3, or even 5 trades consecutively. And if the EA reverses when it should not, then you get another losing streak. 

The idea you describe is somewhat similar to predicting at which point the ZigZag indicator will not repaint. You can use neural nets or statistical methods (such as Poisson) for this, but they have a margin of error. The only way I can think of that gives 100% accuracy when 'optimized' this way is data snooping, which is not good.

To add comments, please log in or register