Gamestudio Links
Zorro Links
Newest Posts
IB - Stop order for options
by Petra. 01/19/21 10:52
Writing strategy DLL in C#
by AndrewAMD. 01/18/21 13:56
Training strategy - errors
by Petra. 01/18/21 12:14
question about View* Entity
by ratz. 01/16/21 23:18
Has anyone used LSTM?
by TraderTom. 01/16/21 22:41
NOFILTER for Level Blocks?
by Dooley. 01/15/21 20:30
AUM Magazine
Latest Screens
Scheherazade's Journey
Paradox Vector
The Space Between
Pogostuck: Rage With Your Friends
Who's Online Now
4 registered members (AndrewAMD, jbhunter, SaschaGanath, Dooley), 469 guests, and 4 spiders.
Key: Admin, Global Mod, Mod
Newest Members
minimeken, TraderTom, babyrazia, copernico, jari
18574 Registered Users
Previous Thread
Next Thread
Print Thread
Rate Thread
Zorro Slow Performance #482165
01/04/21 22:53
01/04/21 22:53
Joined: Dec 2020
Posts: 12
B
bot Offline OP
Newbie
bot  Offline OP
Newbie
B

Joined: Dec 2020
Posts: 12
Hi, I am trying to check the performance of Zorro but I do not see the performance claims made throughout the Zorro website:

"Test your strategies with the world's fastest tick-level backtester (2 seconds for 10 years) in high accuracy - including commissions, swaps, spreads, slippage, margins, interest, market hours, and holidays"

I tried Workshop 4 and some of the others with a 5 minute BarPeriod on EUR/USD over 6 years taking 60+ seconds to run.

Can you please describe why the performance seen is so slow?

I even tried the DLL Strategy but the performance is almost the same.

Re: Zorro Slow Performance [Re: bot] #482173
01/05/21 15:08
01/05/21 15:08
Joined: Apr 2008
Posts: 333
Austria
Petra Offline
Support
Petra  Offline
Support

Joined: Apr 2008
Posts: 333
Austria
I get in fact 0.7 seconds instead of 2 seconds, but the new version is a bit faster and I have a fast PC laugh. You should get 2 sec with the release. The benchmark script is included in the installation.

Re: Zorro Slow Performance [Re: bot] #482203
01/11/21 00:50
01/11/21 00:50
Joined: Dec 2020
Posts: 12
B
bot Offline OP
Newbie
bot  Offline OP
Newbie
B

Joined: Dec 2020
Posts: 12
I am not even getting 2 seconds, but 60 seconds for the whole test.

Mine is just basic install with running Workshop 4 - that's it.

Are you getting 0.7 seconds for whole test covering 10 years with BarPeriod of 1 second? What is the BarPeriod? What are the start and end dates? Can you share your script please?

Also, which version number is fastest?

Re: Zorro Slow Performance [Re: bot] #482207
01/12/21 01:48
01/12/21 01:48
Joined: Feb 2017
Posts: 1,160
Chicago
AndrewAMD Online
Serious User
AndrewAMD  Online
Serious User

Joined: Feb 2017
Posts: 1,160
Chicago
The Benchmark script is the basis of the Benchmark. Maybe you should run the Benchmark script to check the Benchmark.

1.015 seconds here, using an ordinary i7-2600 CPU.

Re: Zorro Slow Performance [Re: bot] #482216
01/13/21 00:10
01/13/21 00:10
Joined: Dec 2020
Posts: 12
B
bot Offline OP
Newbie
bot  Offline OP
Newbie
B

Joined: Dec 2020
Posts: 12
The following is with Zorro 2.34.5 beta:

With Benchmark script as is I get 1.317 sec.

When I change the BarPeriod from 60 to 1 minute in the same script, time goes up to 15.205 sec.

Also, the actual running time as per my stop watch is lot higher, so I assume the time is only capturing time spent for run() start and end.

"Test your strategies with the world's fastest tick-level backtester (2 seconds for 10 years) in high accuracy - including commissions, swaps, spreads, slippage, margins, interest, market hours, and holidays"

Does it mean the above is misleading as its not actually running a tick level tests but just once every hour for the purpose of the testing?

Last edited by bot; 01/13/21 00:27.
Re: Zorro Slow Performance [Re: bot] #482217
01/13/21 01:22
01/13/21 01:22
Joined: Feb 2017
Posts: 1,160
Chicago
AndrewAMD Online
Serious User
AndrewAMD  Online
Serious User

Joined: Feb 2017
Posts: 1,160
Chicago
No, bars are set to one-hour, but you still acquire every tick via the tick() function. This is because the TICKS flags was set at the top of run().

Flags are documented here:
https://manual.zorro-project.com/mode.htm

Bars vs ticks:
https://zorro-project.com/manual/en/bars.htm

tick() function:
https://zorro-project.com/manual/en/tick.htm

Pro tip: You can make the benchmark even faster by removing the TICKS flag. This brings it down from 1.0 seconds to about 0.25 seconds on my CPU.

Re: Zorro Slow Performance [Re: AndrewAMD] #482218
01/13/21 22:02
01/13/21 22:02
Joined: Dec 2020
Posts: 12
B
bot Offline OP
Newbie
bot  Offline OP
Newbie
B

Joined: Dec 2020
Posts: 12
I added a tick() function to count how often it is getting called and found its being called only once per minute. Ticks are more granular than just 1 every minute.

For EUR/USD the tick() was called 3,709,192 times for 10 years of data. In 10 years there are 3,744,000 minutes excluding weekends.

So it is not really tick level but simply 1 minutes. Also the benchmark does not even have any function for ticks - this test is entirely misleading especially with this claim about "tick-level" performance.

Overall its just an hourly bar test on RSI. There is no logic on TICK.

Last edited by bot; 01/13/21 22:04.
Re: Zorro Slow Performance [Re: bot] #482219
01/13/21 22:39
01/13/21 22:39
Joined: Feb 2017
Posts: 1,160
Chicago
AndrewAMD Online
Serious User
AndrewAMD  Online
Serious User

Joined: Feb 2017
Posts: 1,160
Chicago
In this case, the ticks being supplied are minutely T6 ticks. So every tick is being evaluated. You can change the History variable to load a T1 tick file instead. The Zorro website has some sample tick-by-tick data for EUR/USD that you can extract into the History folder.

History string variable:
https://zorro-project.com/manual/en/script.htm
Originally Posted by bot
There is no logic on TICK.
There is no user-supplied logic in particular. Nonetheless, Zorro is still checking for stop losses and take profit on every tick, so there is tick logic.

Re: Zorro Slow Performance [Re: bot] #482223
01/14/21 02:13
01/14/21 02:13
Joined: Jul 2017
Posts: 510
Z
Zheka Offline
User
Zheka  Offline
User
Z

Joined: Jul 2017
Posts: 510
...But the FAST flag - added to speed-up the benchmark - "produces" wrong results for this single-asset backtest (which it shouldn't);
Removing it makes both the speed and results more sensible.

Last edited by Zheka; 01/14/21 02:15.

Moderated by  Petra 

Gamestudio download | chip programmers | Zorro platform | shop | Data Protection Policy

oP group Germany GmbH | Birkenstr. 25-27 | 63549 Ronneburg / Germany | info (at) opgroup.de

Powered by UBB.threads™ PHP Forum Software 7.7.1