Aurix Exchange

Optimizing period for an EA

maven

Member
Can anyone suggest the best method of optimizing an EA in terms of the period to cover?

I have recently optimized an EA for 2010/ 2011 and got absolutely fantastic results. However, when I applied it to this year the results were poor.

Re-optimizing for this year gets things to look a lot better, but it is rather like fudging it for what has been.

Your suggestions would be most appreciated.
 

stovedude

Active member
My experience with optimizing is this: you can optimize for eternity, but it will never prepare you for the future. I think the optimizer just tells us what we want to hear, basically because we work it to death. As soon as you apply those settings, it is guaranteed that the market will suddenly change at that moment. And yet, we must optimize to find decent settings. It is like trying to read a woman's mind. As soon as she notices what you're doing, she will change the rules yet again:D. Unless you can follow the rhythm of the market, optimization will always fail. That is why those shiny new EA's win a little when they are first marketed. They are "curve-fitted" for the present market, but the market makes and changes its own rules continuously.
 

maximo

Member
Hi Maven,

I am against optimization as it is as SD says. History is unlikely to happen again in the context of those wins for the reasons that they occured. Like betting on a horse because it won it's last race, does not mean it's going to happen again, more often than not, winning is just a temporary statistic as it is with most games, gambling or not. The fact that a string of wins happened within a period of time with a set of indicators is usally coincidence.

After many years.. I have found my 'holy grail' lol and it is this. Use a longer term MA to simply confirm your trading system. Does not matter if you are scalping or trend trading, the purpose is to have more winners than losers right? So it should become obvious that the amount of time price action stays above or below a longer term moving average has more trades going your way. This is the key to statistically robust trading systems. At this point you may go ahhh.. but your losing trades will require larger stops than a 1:1 risk reward if you are using a longer term average as the stop. This is often true, but you have to admit that larger wins or a higher frequency of wins for scalping must also follow with this. It's just statistical fact.

You must also base your longer term average just outside the noise zone of the currency pair you are trading, since the range of each pair is different due to their individual volatility characteristics. This can be found by testing, but it will be a non optimised parameter. Like 130MA may work as well as 140MA, since they are rounded by 10 you should regard this as statistically significant, particularly if, 'all values inbetween' work just as good :D

Currently I make only 0.8% per day average consistently. May not seem much, but with 5% risk it averages to more than 125% over 5 months live trading with compounding added. Never forget the power of compounding!

Cheers!
 

grood

Member
Currently I make only 0.8% per day average consistently. May not seem much, but with 5% risk it averages to more than 125% over 5 months live trading with compounding added. Never forget the power of compounding!

Cheers!

Which system (and their timeframes) are you actually trading?

Looking at upper trend using MA is indeed easy and a very additional filter. I also take a look at ADR, i.e. if a certain pair has moved already 2/3+ of its ADR, no trade for the day anymore (if trading M15-H1).
 

maximo

Member
The system I trade is my own and not found here, it does not have a name. It simply follows the trend using a long average. I like M30 charts the best. On average I get 3 good trend trades per month, but they are on average 8 times larger than the more frequent losses during ranging periods.

This is my analogy of the market:
It's like a battlefield. Those that fight close in using small averages with small stops and optimisations are like those using swords and knives. Their enemy are the big guys standing far away with their guns and some smaller ones with arrows (ones using a large average). Often there is an Achilles or a Ghengis who leaps out and cuts you with a loss, but on average if you are standing afar with your longer range weaponry you will prevail.

The point I want to make is that price bars spend a lot more time beyond the larger averages and this should be an edge you can exploit with any system.
 

Aurix Exchange
Top