How to distinguish a good EA from a luck period?

[Rafael Grecco]

Hello,

I've been developing EAs for some years now. I have always tried to make an EA that can work on any forex pair. I've always thought that a good EA should be able to make profit regardless of what forex pair it was running.

This is what I usually do: Have an idea, write EA, make some adjustments, run some backtests on different pairs and different periods just to see if it "works" and then I put it on a Demo Account on 10 pairs, which I choose based on liquidity and spread size.

That's what I did on my most recent creation. The problem was that the growth was too irregular. It could have an excellent growth during some days/weeks and then the next couple days or weeks have a huge drawdown, losing almost all profit of the last days/weeks.

Then I made a report of all pairs that it was running and found out that half of the pairs were responsible for 90% of wins while the other half were responsible for 90% of losses. Then what I did was choose the best 5 pairs and start the EA again only on those 5 pairs.

It's been some weeks now and the EA continues to make excelent profit with only a few losses.

Here is where I want your opinion:

Can a EA perform excelent on some pairs and be terrible at others? This seems very strange to me. I've always thought that a solid EA should be able to handle anything... are forex pairs that different from each other?

---------

I initially tested the EA on the following pairs: EURUSD, GBPUSD, AUDUSD, USDCAD, USDJPY, USDCHF, NZDUSD, EURGBP, EURJPY and GBPJPY

Now I am running only on these pairs: EURUSD, GBPUSD, NZDUSD, EURGBP and GBPJPY

[Keith Watford]

What you are Possibly experiencing is the Pareto Principle/Distribution which is generally 80/20. In trading it is often referred to as the 80/20 Rule.

This may help you to identify prospective pairs for the short term, but unfortunately this is unlikely to continue.

If you continue with your testing with all the pairs, you will find that the pairs that give you 80/90% of the wins will change.

[Anthony Garot]

Rafael Grecco:

Can a EA perform excelent on some pairs and be terrible at others? This seems very strange to me. I've always thought that a solid EA should be able to handle anything... are forex pairs that different from each other?

There was a recent, related post here:
https://www.mql5.com/en/forum/331746

[Marco vd Heijden]

Every pair moves differently so you would have to use an EA that can distinguish between pairs and run a formula to calculate the needed parameters to fine tune the strategy for that particular currency pair.

If your EA does not do that and just applies a basic setting for all pairs then the results will usually vary a lot.

So it all depends on how it is coded.

[Rafael Grecco]

Thank you all for your answers.

This seems a very complex topic, where each individual has a very different opinion (which is the case on almot any topic related to trading).

Keith Watford:

What you are Possibly experiencing is the Pareto Principle/Distribution which is generally 80/20. In trading it is often referred to as the 80/20 Rule.

This may help you to identify prospective pairs for the short term, but unfortunately this is unlikely to continue.

If you continue with your testing with all the pairs, you will find that the pairs that give you 80/90% of the wins will change.

The Pareto Principle seems too philosophical to me. It may or may not happen in the future. And if it happens, it could be in a day, in a week or in a century...

Anthony Garot:
There was a recent, related post here:
https://www.mql5.com/en/forum/331746

Thank you for letting me know about that thread. It was a interesting read, but opinions there also vary a lot.

Marco vd Heijden:

Every pair moves differently so you would have to use an EA that can distinguish between pairs and run a formula to calculate the needed parameters to fine tune the strategy for that particular currency pair.

If your EA does not do that and just applies a basic setting for all pairs then the results will usually vary a lot.

So it all depends on how it is coded.

The only paremeter that my EA uses is ATR. ATR is used to calculate the amplitude of price movements and adjust the necessary numbers.

You said that every pair moves differently and I agree.

But does this mean that pair X can move like pair Y was moving yesterday? Will pair Y move like pair Z in the future? Or each pair has their own "personality", and while they will change, they will always move in their own way?

For example: Pair X will always be choppy, Pair Y will always move in great amplitudes, Pair Z will always have shorts trends and long periods of ranging, etc (by always, I mean ususally, the "normal" behavior of that pair).