The difference in backtesting on PRT Demo / PRT Live

Viewing 13 posts - 1 through 13 (of 13 total)

Author

Posts
01/20/2019 at 1:04 PM #89142 quote

Shitforbrains
Participant

Average

Hi,

Has anyone else experienced a big difference in backtesting the same code on PRT Demo vs PRT Live? Have a look at the pics below. The code is exactly the same, tested on the same amount of units and with the same spread. How can this be, and which one should I trust? Probably the losing one. 🙂

PRTLiveBacktest.png PRTDemoBacktest.png

01/20/2019 at 1:47 PM #89146 quote

GraHal
Participant

Master

You sure you have tick by tick mode selected on the winning curve?? 🙂

Shitforbrains thanked this post

01/20/2019 at 1:48 PM #89147 quote

Vonasi
Moderator

Master

My first thoughts are that it looks like one is tested with tick by tick on and the other with it off?

Shitforbrains thanked this post

01/20/2019 at 1:49 PM #89148 quote

Vonasi
Moderator

Master

Also is there any time conditions in the strategy and if so are both platforms set to the same time zone?

Shitforbrains thanked this post

01/20/2019 at 2:11 PM #89149 quote

Shitforbrains
Participant

Average

Ah you’re right, tick by tick was not “ticked” on the live one, time zones are the same on both. But it’s still behaving weirdly on the live version, banking profits that are clearly losses when I zoom in on the candlesticks. Any idea how to minimize these?

01/20/2019 at 2:19 PM #89150 quote

Vonasi
Moderator

Master

Shitforbrains wrote:

Any idea how to minimize these?

Not without the code so as to know what the strategy is!

01/20/2019 at 2:42 PM #89151 quote
Shitforbrains
Participant

Average
Oh yeah, the code. 🙂

It’s a simple Heiken Ashi strategy that opens a position based on the change of color.

This is used on 1h tiumeframe between 0900 and 1800. (Not the same code used from pics in first post)
```
DefParam FlatBefore = 090000

DefParam FlatAfter = 180000

Defparam cumulateorders=false

xClose = (Open+High+Low+Close)/4

if(barindex>2) then
xOpen = (xOpen[1] + xClose[1])/2
endif

c1 = xClose>xOpen AND xClose[1]<xOpen[1]
c2 = xClose<xOpen AND xClose[1]>xOpen[1]

if c1 AND Not LongOnMarket then
BUY 1 SHARE AT MARKET
EXITSHORT AT MARKET
endif

if c2 AND Not ShortOnMarket then
SELLSHORT 1 SHARE AT MARKET
SELL AT MARKET
Endif

SET STOP $LOSS 26

SET TARGET $PROFIT 24
```
01/20/2019 at 3:17 PM #89158 quote

Vonasi
Moderator

Master

Shitforbrains – please try to remember to use the ‘Insert PRT Code’ button when posting code as it makes it much easier for others to read. I have tidied up your post for you. 🙂

01/20/2019 at 3:35 PM #89159 quote

Vonasi
Moderator

Master

I can’t see any weird trades like you suggest – do you have any screen shots to share with us?

01/21/2019 at 4:30 PM #89263 quote

Jakub
Participant

Average

Dear all,

I beleive the difference is prompted by the new ProOrder engine (Multi Timeframe) which has been released only on Demo for now (it is available upon request on Live as well).

They have a different ProBacktest execution in which the new engine is always right.

Regards,

Jakub

Shitforbrains thanked this post

01/23/2019 at 2:18 PM #89439 quote

Shitforbrains
Participant

Average

Vonasi wrote:

I can’t see any weird trades like you suggest – do you have any screen shots to share with us?

Now that I do the backtest and look at it side by side with the system running, I can’t find the faulty trades. Hopefully I was just looking at the wrong trades.

03/26/2020 at 1:09 AM #123342 quote

VN
Participant

Average

Hi Jakub, is the difference in the Multi time frame engine you mention above still present today ie in version 10.3? I have been backtesting a Multi time frame strategy in demo and I want to know if there will be differences in how the same strategy performs with the same backtest parameters in prt live.

03/26/2020 at 9:20 AM #123351 quote

Nicolas
Keymaster

Master

No difference due to MTF. Jakub was talking of a time period where the two versions were existing for clients: one with new engine that embed the MTF and the old one, which is thing of the past now.
Author

Posts