# How to extrapolate the optimization of an EA from a "Broker A" to a different "Broker B"

Hello everyone,

It is well known that, in the foreign exchange market, for any symbol and for the same period of time, the development of the candles is "significantly different" if we first observe it in a "Broker A" and then we observe it in a second " Broker B".

This implies that if the optimization of an EA is done with the histories of a "Broker A", that optimization is not necessarily valid in a different "Broker B" and "almost" you are forced to carry out the optimization of the EA in the "Broker B" if you want to operate on it in the future.

Once we have seen the "general approach", the following question arises:

Suppose I have performed the optimization of an EA for a Broker A.

Suppose I want to take advantage of the above optimization to trade on Broker B.

Is there any application or "smart trick" "trustworthy enough" that allows me to extrapolate the optimization done on Broker A to be able to trade on Broker B?

In other words, is there a way to avoid having to optimize again on Broker B?

Or... Is it going to turn out that optimization on Broker B is absolutely necessary and inescapable?

If there isn't an exact solution to this question, I'd also like to explore "closest solutions"...

Thank you so much!

Yes, there are reliable ways to make an EA do well on both brokers, but it is NOT by depending on parameter optimisation in the Strategy Tester, but to “optimise” the strategy itself.

However, that type of “optimisation” of the strategy is not done by the Strategy Tester, and many times is it done by the “human” developer. Observe why a certain parameter values work well and others not.

For example, instead of having a fixed S/L and T/P that need to be optimised, maybe the S/L and T/P needs to be based on a ratio the ATR or StdDev, or on fractals, or some other form of observable pattern, or maybe the EA has to monitor MAE and MFE and adjust accordingly. It can even be by a combination of multiple factors.

Another way is to do incremental optimisations in real-time inside the EA itself, where the EA would continuously analyse past data and monitor the outcome of variations of parameter values and choosing the best ones based on a set number of rules and conditions.

There are other ways, but it really is up to the ingenuity and skill of the developer.

It depends of your EA strategy and the kind of optimization you are talking about. An EA can behave almost the same on different brokers if it is not much dependent of changes in data feed and parameters.

If your EA give very different results on different brokers, that probably means it is over optimized and will not work live even with the broker on which it was optimized.

Thanks Fernando and Alain, but I would like to receive answers focused on "usability" and as far as possible from "abstraction".

For example, go down to the field of concrete examples... Ideas that can be applied in practice

adenauer692 #: Thanks Fernando and Alain, but I would like to receive answers focused on "usability" and as far as possible from "abstraction". For example, go down to the field of concrete examples... Ideas that can be applied in practice. Is it possible please?

It is not possible to give you "practical" answers because each case is different. It cannot be generalised.

We would need to have full access to source code, optimisation runs, log output, detailed information of strategy and broker contract specifications, etc.

In other words, it would require a intense analysis of your code and conditions, which is what is usually done by the author of the strategy and developer of the code.

None of us here are you, or in your "team", or can read your mind, so we are unable to give you "practical" answers.

Thanks Fernando, I wonder if everyone thinks like you...

Thanks Fernando, I wonder if everyone thinks like you...
Yes.

Thanks Fernando and Alain, I wonder if "everyone" thinks the same as you...

Because beyond talking about "strategies", "optimization" or "over-optimization", I imagine that if the development of the candles between Broker A and Broker B are "similar enough" we could say that the optimization for A would be valid for B.

In other words, is there something similar to a list of Brokers where you can see what similarity there is between them regarding the development of the candles?.

Or, an Application where you can enter the name of 2 brokers and it tells you the correlation between them?

adenauer692 #:Thanks Fernando and Alain, I wonder if "everyone" thinks the same as you... Because beyond talking about "strategies", "optimization" or "over-optimization", I imagine that if the development of the candles between Broker A and Broker B are "similar enough" we could say that the optimization for A would be valid for B. In other words, is there something similar to a list of Brokers where you can see what similarity there is between them regarding the development of the candles?. Or, an Application where you can enter the name of 2 brokers and it tells you the correlation between them?

Two experienced traders, experienced MQL code developers, proficient with both MQL4 and MQL5, both accepted as MetaQuotes forum moderators have answered you their similar opinions and explained their reasoning.

You are free to not accept those opinions and keep waiting for someone to offer what YOU want to hear even if it is not valid.

However, when are you going to do your own research and draw your own conclusion?

Do a search in the Articles and you will find several articles on optimisations and also self-optimisations. There are also forum threads about adaptive Indicators and EAs.

Thanks Fernando, as I already have experience in other forums, it is not my intention to challenge anyone. And since you are a moderator, you can close the topic.

adenauer692 #:Thanks Fernando, as I already have experience in other forums, it is not my intention to challenge anyone. And since you are a moderator, you can close the topic.

This site's forum threads currently cannot be closed. They remain open "forever". Anyone in the future can continue to debate the issue if they wish.

You are welcome to continue to wait for a different answer while you to do your own testing and research on the matter.