Exact same EA, different brokers, different results..............

New comment
 

Hi,


I find the following fact quite shocking:


I have an EA that I tested with exact same settings, exact same spread (chosen in the startegy tester) and exact same time period on 3 different brokers.


Believe it or not, all three results were different. Not only that, in one broker, the EA was performing very well, in the other one it totally lost, in the third one it was kind of flat.


How in this  world can that be?


I understand that some brokers change the slippage, spreads etc. But shouldn't the overall results still be kind of similiar?


Please explain.


Thx, FF

 

tested as in on live markets or in the tester ?

tester can be different.

 
fractalfreak:

Hi,


I find the following fact quite shocking:


I have an EA that I tested with exact same settings, exact same spread (chosen in the startegy tester) and exact same time period on 3 different brokers.


Believe it or not, all three results were different. Not only that, in one broker, the EA was performing very well, in the other one it totally lost, in the third one it was kind of flat.


How in this  world can that be?


I understand that some brokers change the slippage, spreads etc. But shouldn't the overall results still be kind of similiar?


Please explain.


Thx, FF

The answer is in the spreads, slippage, internet connection and the robustness of the EA. Some EAs have to many hidden filters the are not in the settings.

They are too sensitive to spread, and slippage. So, if a signal is triggered but the spread is not accepted, there is no trade. 

Then, there is slippage and internet. If a signal is triggered, the EAc reacts slowly, and when price changes 1 pip, the EA considers it as a risky trade, even if the price moved calmly. In those cases, it's better to pose a VPS service, because the raw reality is that in our own home we will always have more lag to the servers than with a VPS. Less signals are lost in a VPS.

The problem is, that when the EA losses many signals and only opens some of them... They may not be the better ones. Unfortunately, many EAs have these problems.
 

Hello,

let us concentrate on slippage and spread.


Ok, regarding the spread, I defined it as identical, so we could leave that one for now.


Regarding slippage; well, there are definitely difference among brokers, with some brokers abusing slippage to their advantage (which is illegal). Still, the result would then be rather small prices.


However, I think there is a different issue here, one that Felipe already pointed out. Indeed, one of the brokers trades more signals than the other.


I attached a screenshot which shows the comparison of 2 identical trades. Indeed, the overall result is roughly similiar (+2.4p vs +2.8p).


What differs though is the number of trades; as it can be seen, the broker on the right did some trades the broker on the left did not do.


This can also be seen in the charts.


The indi giving the entry gave additional signals on the right broker.


Would that actually also be a result of spread and slippage since on the left the entry indi is sometimes not triggered because of price differences?


Thx so much!

Files:
comparison.png  55 kb
 

In the example, the entry prices for both brokers are the same.

The exact prices differ slightly (I guess because of the slippage).


However, this is rather not the source for the vastly different strategy tester results of both brokers.


As it can be seen in the screenshot, the broker on the right did additional trades.


Is that because their pricing as such differs?


If yes, can I be sure that when the strategy was optimized for the broker I use the EA live on I will get what I see?

Or is there still the risk that one of the brokers shows me one thing but delivers another?


For example, could it be that the broker on the left shows me good results but once I use it live, I will suffer from a mismatch of backtest order execution and live order execution?

 
My feeling would be that the right broker is more realistic and gives more honest / less manipulated pricing and that I should optimize the strategy there.
 

"So, if a signal is triggered but the spread is not accepted, there is no trade. "


1. Could you elaborate a bit? That sounds right, too (however, why would a broker not accept a spread? Would that be the equivalent to a requote?).


2. However, in the case at hand, the broker on the right has indicator signals the one on the left does not have, so the indi appears on the right more often than on the left.

This would still be the result of different pricing in the first place (spread or slippage?), no?

 
fractalfreak:

If yes, can I be sure that when the strategy was optimized for the broker I use the EA live on I will get what I see?

o no no no not even close please try with 0.01

Please understand what you are optimizing for.

https://www.metatrader5.com/en/terminal/help/tick_generation

 

Hm...I think we are up to sth here....


so does that mean:


- Different spread LEADS TO different pricing LEADS TO different indicator signals?

OR

- Different spread in the history data LEADS TO different pricing (historical vs live) LEADS TO different indicator signals?

OR

- Different spread in the history data LEADS TO different pricing (historical vs live) LEADS TO different fill prices live?



So in other words, because of an unrealistic spread regarding the historical data, the live results will ALWAYS be different? Is the % of modelling quality shown in the strategy tester even relevant here?

 

By the way, both brokers shown are ECN brokers.


So since they should earn the commission and not the spread, why does the spread differ there?


Further, if the spread is defined via the strategy tester option, would both not be similiar?


So please be so kind to explain to me again:


1. Since the spread (and therefore prices as such) will differ among brokers, there will be a difference in indicator outputs among brokers.

Hence, different results regarding any EAs.

2. Can I trust that given the SAME broker and the backtests analyzed on the basis of historical pirces from THAT BROKER only, the backtest results are indicative for live results (both differeing just on the basis of slippage etc.?).

3. Do some brokers give different pricing between demo and live accounts? If yes, in which way? If only the live account shows the REAL spread, what do the demo accounts show?

 

fractalfreak:

the live results will ALWAYS be different?

Correct.

fractalfreak:

the backtest results are indicative for live results


Incorrect.

Maybe you can run another test on 3 demo feeds at the same time.

That would result in closer indications of what you seem to be looking for.

12345
New comment