Tick Data causing late entry (1 bars late)

To add comments, please log in or register
gangsta1
1614
gangsta1  

I am testing initially with open prices only as my EA only uses open of bars and indicators. Now when I test results from open prices only testing on every tick I get the exact same entries and results which is great. The EA does not rely on scalping and uses large SL/TP so it makes sense open prices only & every tick results are the same. However my problem is when using Dukascopy tick data for 99% modelling quality - the results are way off! Now I have watched the system in visual mode and with tick data for some reason the EA is opening orders a few bars late and closing them a few bars late - e.g. rsi cross up would enter on open of next bar normally but with the dukascopy tick data it opens a few bars after the cross). I am thinking it may be due to the difference between Alpari (testing broker) and the Dukascopy data but how would this explain such a late entry/exit? There is now time issue here as the EA was not using a time filter. My thoughts are that it could be the indicators being on Alpari data and the ticks on Dukascopy data but that is far fetched and should not affect the EA entry as far as im concerned.

Any ideas?

Simon Gniadkowski
17271
Simon Gniadkowski  
gangsta1:

I am testing initially with open prices only as my EA only uses open of bars and indicators. Now when I test results from open prices only testing on every tick I get the exact same entries and results which is great. The EA does not rely on scalping and uses large SL/TP so it makes sense open prices only & every tick results are the same. However my problem is when using Dukascopy tick data for 99% modelling quality - the results are way off! Now I have watched the system in visual mode and with tick data for some reason the EA is opening orders a few bars late and closing them a few bars late - e.g. rsi cross up would enter on open of next bar normally but with the dukascopy tick data it opens a few bars after the cross). I am thinking it may be due to the difference between Alpari (testing broker) and the Dukascopy data but how would this explain such a late entry/exit? There is now time issue here as the EA was not using a time filter. My thoughts are that it could be the indicators being on Alpari data and the ticks on Dukascopy data but that is far fetched and should not affect the EA entry as far as im concerned.

Any ideas?

Yes,  what is the timezone of the Alpari data ?  what is the timezone of the Dukascopy data ?
gangsta1
1614
gangsta1  
Alpari is GMT+3 and the Dukascopy tick data is GMT+0. The EA runs 24 hours a day so I dont think it is a time related issue. If the rsi is over 70 the EA will sell when NOT using the tick data (using open prices or alpari data in every tick mode) but when using the dukascopy tick data (99% modelling quality) it will sell a few bars after. As mentioned the EA only uses open prices for entries and exits so I do not see how this could be happening unless the indicators are affected by the difference in broker data. On a side note I am beginning to think that 99% modelling quality is only required for EA's that scalp and can be affected by ticks within a 1 minute bar whereas my EA running on m15 and above will not be affected especially considering it uses open prices (not data within bars).
Simon Gniadkowski
17271
Simon Gniadkowski  
gangsta1:
Alpari is GMT+3 and the Dukascopy tick data is GMT+0. The EA runs 24 hours a day so I dont think it is a time related issue. If the rsi is over 70 the EA will sell when NOT using the tick data (using open prices or alpari data in every tick mode) but when using the dukascopy tick data (99% modelling quality) it will sell a few bars after.
A few bars,  what timeframe are you testing on ?  H1 ?  a few bars,  3 maybe ?
gangsta1
1614
gangsta1  
Yes seems to be 3 bars, it happens on the timeframes I have tested thus far - m15 and h1. I understand what you are trying to say about 3 bars difference because that is the difference in Alpari and the tick data time, but if the EA is entering on the condition rsi is overbought, then it should not matter what bar or time it is, the order should be placed. This is not the case, it will wait for a few bars after the overbought signal before entering. Perhaps I am missing something and indicators are not lining up with the bar data.
Simon Gniadkowski
17271
Simon Gniadkowski  
gangsta1:
Yes seems to be 3 bars, it happens on the timeframes I have tested thus far - m15 and h1. I understand what you are trying to say about 3 bars difference because that is the difference in Alpari and the tick data time, but if the EA is entering on the condition rsi is overbought, then it should not matter what bar or time it is, the order should be placed. This is not the case, it will wait for a few bars after the overbought signal before entering. Perhaps I am missing something and indicators are not lining up with the bar data.

You said . . .  "So I am testing my EA with the EXACT same parameters on the d1 timeframe and comparing 90% modelling quality using data from metquotes (history center) with 99% modelling quality using tick data from Dukascopy."

 

If you are using D1 for your Indicators then there will be a difference of 3 hours between the D1 bars of your 2 data sources . . . 

gangsta1
1614
gangsta1  
I understand there is a difference of 3 hours between the bars but that does not explain why the EA is not taking trades when the rsi indicator crosses into the overbought zone. I cannot get accurate results with 99% modelling quality if the EA does not enter on the cross regardless of the timezone of the data. It is not an issue with my code as it enters when the rsi crosses when not using the 99% modelling quality, maybe it is just that when I load the rsi indicator to test visually the gmt difference is causing the indicator to not line up with the entry in which case I should probably add the Alpari GMT offset difference to the Dukascopy tick data for better comparison of results.
To add comments, please log in or register