Backtesting hangs NT

[kaywai]

Hi,

I've been backtesting a strategy using 6 months of 1 min data of Nikkei 225 futures. After running a backtest twice with changes to the parameters in the code, NT almost grounds to a halt. Even the pop-up window before backtesting is run takes ages to load. Is this a memory issue? If so, is there any workaround like allocating more virtual memory or something?

[NinjaTrader_Jason]

Hello kaywai,

I am not sure what if causing slowness after you performed 2 backtests. Does the issue persist if you restart NinjaTrader?

What happens if you backtest the SampleMACrossOver strategy twice; does NinjaTrader become slow as well once you ran the backtests?

Could it be the backtest was still running when you experienced the slowness?

[kaywai]

Hi Jason,

The problem doesnot happen after I restart NT.

I don't know whether the backtest was still running but the previous two results had already been produced and I wanted to do a third backtest. That was when it slowed...

It certainly processes a lot faster with the sample MA crossover strategy.

[NinjaTrader_Jason]

Hello kaywai,

Please let me know if the slowness persists after you backtested.

Generally, if you experience slowness of the software, please reduce resource request and check if this will make a difference.

Check if you are running workspaces in the background (these will request data & resources as well) and close them. Can you please restart NinjaTrader and go to File-->Workspaces and check what workspaces are listed below 'Save Workspace As'. The workspace with the checkmark is the workspace opened on your monitor. Any additional workspaces listed, are opened in the background.

Select a workspace without a checkmark and subsequently go to File-->Workspaces-->Close Workspace. Repeat this process until only one workspace is opened.

Close some charts, SuperDOM's or other NinjaTrader windows. Reduce the amount of indicators in your chart or set 'Calculate on bar close' to True. Reduce the amount of historical data in your charts by configuring the Format Data Series window.

More tips can be found at the following link.

http://www.ninjatrader-support.com/HelpGuideV6/PerformanceTips1.html

[kaywai]

cool. thx jason!

[kaywai]

Jason,
How do I reduce resource request? Is 6 months worth of 1 minute data too much if you use timeframes other than 1 minute?

I only had the strategy analyzer opened/running and I still had the same problem. Had to shut it down at least 10 times this evening.

[NinjaTrader_RyanO]

Hi,

Can you clarify which instruments you are testing because 6 months worth of 1 minute data should not be too much, but it may be that the data is not present for back testing and so it freezes.

You can check this by opening a chart for the same time period that you are back-testing and making sure the data is present.

[kaywai]

Hi Ryan,

The instrument is the Nikkei 225 futures listed on the Singapore Exchange. The data is present because it was imported to NT and I have been using the same data for indicator construction.

As I mentioned earlier, the first 2 times of backtesting, there doesn't seem to be any issues but by the third time of using the same data with different parameter (tweaking of the parameters), the whole thing freezes. I can't even get backtesting to run! This is because the window which opens before backtesting can commence can't even open!

[NinjaTrader_Jason]

Hello kaywai,

It seems to be related to the strategy you are backtesting, since it does not occur when backtesting the SampleMACrossOver strategy, correct? Does your strategy use many calculations and conditions? I suggest to reduce the amount of historical data in your backtest and check if this will make a difference.

In addition, NinjaTrader 7 has been improved in this regard. Please see the link below for more information.

http://www.ninjatrader.com/webnew/NT7/NinjaTrader7.html

General Application Performance
This is an area that we spent a lot of time on during the development of NinjaTrader 7. We re-wrote core sections of the application to be more efficient in terms of memory usage and CPU utilization.

[kaywai]

Hi Jason, Perhaps you are correct. There doesn't seem to be any issues when I use my strategy as an indicator. Works fine. 3 additional lines for a the entry, exit shouldn't hurt it, right? But it doesn't happen after the first backtesting...or the second...does that mean it can't flush out the memory or something because it has no reason to store past backtesting.

How reliable is NT7 now? how many more betas to go? will my existing indicators that i created be made redundant? is there a manual to come along with it for beta users?

[NinjaTrader_Jason]

Hello kaywai,

Unfortunately I am not sure if I follow. What do you mean by using your strategy as indicator?

NinjaTrader 7 is still in beta phase. I am not sure how many beta versions will be released before it is released as production version.

There are changes in regards to NinjaScript between version 6.5 and 7. Please see the following link for more information.

http://www.ninjatrader-support2.com/vb/showthread.php?t=21016

[kaywai]

There were a few methods within the indicator that I couldn't call up from the strategy...so i cut & pasted the whole indicator to the strategy.

[NinjaTrader_Jason]

Ok, so how did you backtest it after converting to an indicator?

Not sure how to interpret: "does that mean it can't flush out the memory or something because it has no reason to store past backtesting."

[kaywai]

it was originally an indicator which i tried to automate but not the conventional type of indicator...works fine as an indicator. backtesting it is the problem...i should say the hanging is the problem.

"does that mean it can't flush out the memory or something because it has no reason to store past backtesting." this means that by running the backtesting it is unable to clear out the previous backtesting and bloats itself till it has run out of memory and hangs....technical stuff that i'm not too sure about but came across before. system dump or memory dump?

[NinjaTrader_Jason]

I do not think this is the case; if this was a loop, it should occur on the first backtest as well.

Unfortunately I am not sure what is causing the issue. I suggest to reduce the historical data used in the backtest and check if this will avoid it.