Inaccurate Results across SQ, MT4 and backtesting

Dear All,

i been struggling for weeks with this problem, I created a EA in SQ, using the tickdata from SQ Tickdownloader.

However just after 2 days in testing and running it on a live demo account, all results are completely different from the backtesting in SQ and MT4 backtester

I have tried with other EAs created from SQ, and the results are all different too,

I attached the EA, SQ files, and Screencaps from MT4.  Could someone please help.  No matter how i try, i could never get results similar or even close to the back-testing.

Note : I using Trade on Open Bar and Exit at End of Day on this EA.

Timeframe : EURUSD M15

can you please open a bug task here: https://roadmap.strategyquant.com (for SQ4) and post your .str file there?

I’ll look at it. There must be some proble with your configuuration, trading on bar open should be very exact.

Hi Mark, tks for quick reply,

Done as requested !

So if I understand the problem it is that the strategy behaves in some way in SQ3/MT4 backtests, but then differently in live trading?

There could be numerous reasons, one of them is that the sample of just 3 finished trades is not enough to evaluate demo trading.

It could be that it opens the trade a little later than in MT4/SQ3 backtest – because backtest starts at 0:00, but you attach your startegy to a chart at a different time.

This could result in the trade closed on a different bar and all consecutive trades will be then opened later. It should sort itself out after a while, but it woilll take more than a few trades.

You could test if this behavior doesn’t influence your strategy by running Robustness Test simulation wiht Randomize Starting Bar.

In any case, you should let it run at least for a week or two, have at leat 30-100 trades, and then evaluate the performance.

@KC: What for a dataquality  you are using in MT4?

Do you use the data’s, in the backtest from MT4, what are MT4 are offering or do you use also Data’s from dukascopy (for example)?

I’m asking for that because you can not challenge the MT4 backtestdatas with the Data’s from Dukascopy.

MT4 Backtestdatas are much more worst from the quality as the Dukascopy data’s.

Marcel wrote:

@KC: What for a dataquality you are using in MT4? Do you use the data’s, in the backtest from MT4, what are MT4 are offering or do you use also Data’s from dukascopy (for example)? I’m asking for that because you can not challenge the MT4 backtestdatas with the Data’s from Dukascopy. MT4 Backtestdatas are much more worst from the quality as the Dukascopy data’s.

Tks for reply,  what i did was,  I run the EA on Live Demo account on Oanda. After which, i backtest and compared the Backtested data VS the Demo account (for the same date periods) and discover there is a different.

Mark Fric wrote:

So if I understand the problem it is that the strategy behaves in some way in SQ3/MT4 backtests, but then differently in live trading? There could be numerous reasons, one of them is that the sample of just 3 finished trades is not enough to evaluate demo trading. It could be that it opens the trade a little later than in MT4/SQ3 backtest – because backtest starts at 0:00, but you attach your startegy to a chart at a different time. This could result in the trade closed on a different bar and all consecutive trades will be then opened later. It should sort itself out after a while, but it woilll take more than a few trades. You could test if this behavior doesn’t influence your strategy by running Robustness Test simulation wiht Randomize Starting Bar. In any case, you should let it run at least for a week or two, have at leat 30-100 trades, and then evaluate the performance.

thanks Mark,  ok i let it run in meantime, i created a new EA using M15 timetable using Trade on Open Bar and put it on Oanda+5GMT.  i let it run for a week and  then i will re-compare the data again.

btw, if my broker using GMT+5,  when i Download the tick data in SQ Tickdownloader is it needed that convert to GMT+5 too ?

Hello,

you with TickDownloader data export you need to match your broker’s timezone