Dukascopy Support Board
http://www.dukascopy.com/swiss/english/forex/jforex/forum/

Historical Tester question
http://www.dukascopy.com/swiss/english/forex/jforex/forum/viewtopic.php?f=21&t=54504
Page 1 of 1

Author:  hopeless [ Wed 30 Nov, 2016, 10:34 ]
Post subject:  Historical Tester question

I do not understand if I'm doing something wrong while using the historical tester.

I have taken some of the strategies at the top of the standing of the current month and try to run them (without any change) on historical tester selecting period from 01-Nov-2016 to 28-Nov-2016

I have set the account to 100k USD, processing all ticks and used the same instrument and default setting.

Well, what I'm getting is totally different from real data of the running strategies.

i.e.: this month's leader for strategy is bdpips with "superway" strategy.
I've dowloaded the strategy and run the in historical tester from 01-Nov-2016 to 28-Nov-2016 processing all ticks and with the account initially set to 100K USD.

Results I'm getting are totally different from what I see in the Strategy Contest page.
For the same period the strategy result having quite a high P/L.
Trying in historical tester the account balance reach 1$ in few days.

Can you explain to me if I'm doing something wrong in testing the strategy?

Author:  Illya [ Thu 01 Dec, 2016, 12:59 ]
Post subject:  Re: Historical Tester question

Greetings. The problem is with not a historical tester but with a strategy of bdpips. His strategy has no logic to open a crrect position. Looks like he got some bug while execution and used it in Strategy contest. So do not worry. That system works only with a sepcific terms. And can send a BUY orders only.

Author:  vadim_berezhnoj [ Fri 02 Dec, 2016, 14:46 ]
Post subject:  Re: Historical Tester question

Hello
Historical Tester can not replicate 100% conditions of contest account. Number of positions is limited by 1 on server side.
Some times strategy can stop for different reasons. And so on.

Author:  hopeless [ Thu 22 Dec, 2016, 12:14 ]
Post subject:  Re: Historical Tester question

Hi,

I may agree on the fact that historical tester cannot replicate 100% real market but so far I could not get even close to what is happening in real.
I have checked other strategies which open 1 position at a time and still what I get is different.

I.e.: strategy "dec" from fx211pips is quite simple and it only open one position at a time as there is a check on open position as first step.

When running with historical tester (I tried to run it for the period 1-Dec-2016 to 10-Dec-2016) I got very strange results, like orders are open more than once and equity becoming negative.

I have tried also with other strategies from other participants but in all test I have done the results are always quite different from what I see in the strategy contest result.

Is some of you able to reproduce the real performance on the same period with the historical tester for some of the strategy in the contest? If yes, can you please explain which contestant strategy you have used and the step you have taken to reproduce the same performance and in the contest?

Regards.

Author:  Illya [ Thu 22 Dec, 2016, 13:13 ]
Post subject:  Re: Historical Tester question

Please provide an example.

Author:  hopeless [ Thu 22 Dec, 2016, 19:12 ]
Post subject:  Re: Historical Tester question

Illya wrote:
Please provide an example.


Hi Illya,

as I was saying you can check the strategy from fx211pips that you can find here:

https://www.dukascopy.com/strategycontest/?action=strategyPage&trader=fx211pips&strategy_id=70942

Its strategy dec (ver.2) has been running since 1-Dec-2016 and the in the period 1-Dec-2016 to 10-Dec-2016 increased the equity from 100K to 348K.

If you take this strategy (dec ver.2) and you try to run it for the same period in the historical tester I cannot even get close to what happen in reality. Trying the same strategy for period 01-Dec-2016 to 10-Dec-2016 I got an final equity of 146K vs the real 348K.

Considering that strategy has been started on 01-Dec-2016 I tried also to try different period (i.e. 04-Dec-2016 to 10-Dec-2016) but still I cannot reproduce what happened in reality. Trying the period 04-Dec-2016 to 10-Dec-2016 I got a final equity of 21.2$ :?:

So again, is there a way to reproduce the real behavior of this strategy with historical tester?

By the way, I have to say that I was wrong in saying that there were 2 orders created at the same time. I was confused by the margin cut orders.

Author:  Illya [ Fri 23 Dec, 2016, 12:42 ]
Post subject:  Re: Historical Tester question

Hi Hopeless
I will check it out.

Author:  Illya [ Mon 26 Dec, 2016, 12:03 ]
Post subject:  Re: Historical Tester question

Greetings.

Sorry for delay.

I found whats wrong, but can't even imagine why it happens. Any way. The problem was in first If block where it check an ammount of all Open position. If to change this variable from Open.Amount to AllPostion.Amount, it gives a correct results. Otherwise by some unknown reason it opens more than one order at once.

Best regards.

Attachments:
dec.vfs [32.01 KiB]
Downloaded 1193 times

Author:  vadim_berezhnoj [ Tue 27 Dec, 2016, 10:56 ]
Post subject:  Re: Historical Tester question

Hello.

It is good to filter Periods and Instruments before everything else.

onCandle generates signals for all periods for all subscribed instruments.
For example at 12:00 there will be signals for 10sec, 1-5-10-15-30 min, 1 and 4 hours closed candles.

Attachments:
dec_2.vfs [38.79 KiB]
Downloaded 1179 times

Author:  hopeless [ Thu 29 Dec, 2016, 15:41 ]
Post subject:  Re: Historical Tester question

Hi,

thanks a lot for your clarifications.

I have retested both strategies (dec and dec2) as you both have modified and I have to say that the result looks better.
However it is not 100% accurate as it has happened in real.

But I cannot know if the strategy has been stopped and restarted so I don't want to investigate more on this issue.

Thanks a lot for your help.

Author:  vadim_berezhnoj [ Wed 11 Jan, 2017, 16:20 ]
Post subject:  Re: Historical Tester question

hopeless wrote:
Hi,
However it is not 100% accurate as it has happened in real.
Thanks a lot for your help.


And it will never be 100%. Historical Tester is a simulation.

  Page 1 of 1