Backtesting on multiple brokers

New comment
 
Hi! I am in the process of backtesting one of my EAs. I have done this a thousand times before. I have downloaded and fresh installed MT4 from three different brokers, downloaded history data from MetaQuotes in the history center. And set the GMT offset according to the different brokers server time. Everything is setup EXACTLY the same except for the GMT offset in the EA. Two of the brokers have the same gmt offset. Still I get three VERY different results???? I thought I was testing on the same data on the same MT4 exe, but no....Anyone else experienced the same thing? What could be the cause of this?

Thanks for any comments!
 
ziggy:
Hi! I am in the process of backtesting one of my EAs. I have done this a thousand times before. I have downloaded and fresh installed MT4 from three different brokers, downloaded history data from MetaQuotes in the history center. And set the GMT offset according to the different brokers server time. Everything is setup EXACTLY the same except for the GMT offset in the EA. Two of the brokers have the same gmt offset. Still I get three VERY different results???? I thought I was testing on the same data on the same MT4 exe, but no....Anyone else experienced the same thing? What could be the cause of this?

Thanks for any comments!

This has been discussed many many times... It's due to different market properties among brokers and even with the same broker (for example spread fluctuations...). In order to achieve consistent testing u must have exactly the same testing environment (expert, set file, history data, market properties, testing properties, etc....).


Please SEARCH. There's lots of other info on the subject.

 

A System: Test
 https://www.mql5.com/en/forum/112086

 
Thanks for answers guys! I have already done a lot of searching on this topic. I can see your points when forward testing the system. No problem. I have experienced that myself several times. But I had downloaded the same data from MetaQuotes and done backtests on the same data on three different brokers and the results are different. I thought the spreads etc came from the downloaded data and not from the broker. It must be manipulated during download?

Thanks!
 
ziggy:
Thanks for answers guys! I have already done a lot of searching on this topic. I can see your points when forward testing the system. No problem. I have experienced that myself several times. But I had downloaded the same data from MetaQuotes and done backtests on the same data on three different brokers and the results are different. I thought the spreads etc came from the downloaded data and not from the broker. It must be manipulated during download?

Thanks!

The Tester uses the LATEST market properties obtained from the server the Terminal is connected to when u press 'Start'. Solution: stay disconnected from server. Best done by setting a bogus proxy server (search for it).

 

Also may help:
1. turn off internet connection during the backtest;
2. save and check contract specifications for trading symbol used in the backtest, for example, "GOLD, SERVERNAME, 2010.03.12 12:00" :
2.1. "Spread = 100",
2.2. "Digits = 2",
2.3. "Stops Level = 400",
2.4. "Contract size = 250",
2.5. "Margin hedge = 400",
2.6. "Margin calculation mode = CFD",
2.7. "Profit calculation mode = CFD",
2.8. "Swap long = -3.5%",
2.9. "Swap short = +1.5%",
2.10. "Lot step = 0.1",
2.11. "Minimum lot size = 0.1",
2.12. "Maximum lot size = 100000".

 
Ahh, now we're getting somewhere! Excellent! Thanks guys!!!!!
 
Ais:

Also may help:
1. turn off internet connection during the backtest;

Really no reason to do this. Set MT4 to connect to a bogus proxy server and u'll stay disconnected. Just search for 'bogus proxy'.

 
I have found what was causing the difference in profit for my EA. Its the spread of course :) The spread used on backtests for Alpari is 0.9 and for FXCM 2.3. That makes a huge difference for a scalper EA. Thanks for your help guys!!!
New comment
Reason: