Topic: Difference In backtesting on MT4 to EA Studio using same data set!

Hi guys,

So been doing some back testing on a number of strategies (about 50 in total). and so far have found that all of them have had different results in MT4 drastically different, number of trades, profit etc etc.

Bit of back story of this one. So I use TickStory for historical Dukascopy data to build my EA's on. So I i download the data from TickStory and load it into MT4 then into EA studio all my results are perfect and matching. So I figured why not open an account with Dukascopy since im using there data set.
So ive now opened my account with them, loaded there MT4 and have NOT imported any data from TickStory into the Dukascopy MT4 terminal.

I have then tested strategies on EA Studio which is using my TickStory Dukascopy data then tested on the Dukascopy MT4 platform using there default data which is already loaded in MT4. I did the standard practice of opening each time frame and pressing the home key etc.

I have gone through the indicator chart on a strategy and matched the charts with the MT4 platform and the price all looks good. However there are minute differneces in the indicator values at any given time for example: Upper bollinger band on MT4= 114.421 on EA Studio 114.253 on the exact same bar.

I have attached a couple of screenshots showing a trade that was opened on EA studio but not MT4 back tester.

I have spoken to Dukascopy and they assure me that there historical data feed is what is in MT4.

Has anyone experianced or trade with Dukascopy and have this same issue?

FYI I am only back testing on a 1hr time frame from december last year. Also both are set at fixed spreads of the same amount.

Re: Difference In backtesting on MT4 to EA Studio using same data set!

So after do some further analysis I have found that the days in backtesting were different by one day which set off a very different chain of trades. Goes to show just how un stable a strategy can be. Now to think of a way to not have such unstable strategy’s. I run Monte Carlo testing with all boxes ticked, historical price change to 50% randomise indicator parameters by 10%. Acceptance criteria set to min Return/drawdown 2 and 80% pass

Re: Difference In backtesting on MT4 to EA Studio using same data set!

Hi Michael,

yea, those are the kind of strategies you do NOT want :-) To avoid such strategies, just go to MC and use the "Randomize backtest starting bar".