Backtesting issue

[pmcd]

I currently use esignal with the Kwippop indicators. I am currently trying to back test a theory. I am using a 1 minute ES#F. I can load the chart but when I set the time template to 240 days - esignal crashes and closes. It appears there is some type of buffer limit that is reached and then crashes and has to close. I have determined that I can load only 18 days on a 1 minute chart to do my backtesting. Is there a setting in esignal that I can increase to allow the backtesting of a longer period even if it is not the complete 6 months. I get the same result if I use a Dynamic timeframe and continue to go back in time - when I reach that 18 day threshold esignal crashes and closes. Any help would be appreciated. Thanks
My Machine is a 4GHZ, with 2GB of RAM and plently of HDD space so I do not think it is related to my machine,

[demarcog]

Re: Backtesting issue

Originally posted by thepcmd
I currently use esignal with the Kwippop indicators. I am currently trying to back test a theory. I am using a 1 minute ES#F. I can load the chart but when I set the time template to 240 days - esignal crashes and closes. It appears there is some type of buffer limit that is reached and then crashes and has to close. I have determined that I can load only 18 days on a 1 minute chart to do my backtesting. Is there a setting in esignal that I can increase to allow the backtesting of a longer period even if it is not the complete 6 months. I get the same result if I use a Dynamic timeframe and continue to go back in time - when I reach that 18 day threshold esignal crashes and closes. Any help would be appreciated. Thanks
My Machine is a 4GHZ, with 2GB of RAM and plently of HDD space so I do not think it is related to my machine,

thepcmd,

I know exactly what you are experiencing and share your pain as I have the same problem.


What troubles me is I did notice that the amount of data I could load and back test was dependent to a degree on the power of the PC I was using. But a P4 4GHZ w/ 2GB!! I didn't even know that clock speed existed, nice box.

My box is not as fast as yours and I have never been able to backtest even 5 min data of the #ES for more then 90 days, not sure about my current 1 minute threshold and don't want to find out as if eSignal doesn't crash it hangs requiring a restart.

A few things to try:

First the only possible setting are accessed through the Tools-EFS-Setting. There is a stack and heap and increasing them did definately increase my "crash threshhold". I changed them so long ago I forgot what the defaults were by mine are 100000 and 50, yes some ridiculous multiple of these DOS days parameters but it did help and runs fine.

The maximum documented parameter for a time timeplate is supposted to be 120 days, so specifying 240 could be triggering some error in the code try setting it back to 120.

Not familiar with Kwippop, if you unload those studies can you load the maximum data, as it'spossible that the EFS you are using may be related to the problem?

In event, I haven't really pursued this issue as I thought that it was related to my "inferior" system and now apparently that's not the case and we both should do so now.

Not having the ability to test a system for the full 120 days is a major problem and constant nuisance for me and also a feature that is marketed and paid for by all eSignal users and should be addressed.


Glen

[Guest]

Re: Backtesting issue

thepcmd,

Each time you tab back in the chart, the efs reinitializes. Repeating this results in repeated re-loads. I believe your problems are likely related to your KwikPop studies and the particular studies' Time Template. My recommendation is that you take some time to step through and perform some basic troubleshooting.

I would start with loading a very simple backtest study that is on the forum and readily available to other users. I have links to some below in my signature block. Modify your time template to your desired duration. Once there, you should verify that this study can run successfully on your machine.

Once you have accomplished this, then move back to your study. Before running it, remove the KwikPop Studies from the quation. Trying to debug with a third party software tool adds a indeterminant level of comlexity that should be remove before you go any further. Don't worry, you can put them back in (one at a time) once you verify there are no other problems with your code.

If you run into problems on the way, please post your code as discussed in the forum posting guidelines.

Also remember that if you do crash, you should consider rebooting the machine. You may have corrupted memory when the application crashes that can only be recovered with a reboot.

I suspect there is a simple answer to your problem, but you will have to go through a similar process to what I described. Please let us know if this works out or if you have some problems on the way.

There are several people on the forum (like Glen) who are knowledgable in this area and routinely volunteer their own time and will provide you assistance, provided of course, that you are willing to work through a logical troubleshooting process.

Finally, here are some links you may have already seen or used that you may fine helpful.

http://kb.esignalcentral.com/article.asp?article=2790&p=4

http://kb.esignalcentral.com/article.asp?article=1284&p=4

http://forum.esignalcentral.com/showthread.php?s=&postid=93056&highlight=debug%2A#post93056

http://share.esignal.com/groupcontents.jsp?folder=debug%20tools&groupid=339

http://forum.esignalcentral.com/show...t%2A#post90315 (to see the formuaOutput file after a crash, look here first, it is overwritten if you re-start eSignal).

FWIW, I have several very complex strategies running (4 timeframes, liberal use of efsInternal and efsExternal), multiple charts, etc, with a 4 year old machine. Even though I have 1Gig of memory, I rarely have these kind of problems. When I do, invariably, the reason is either my code or something I have inadvertantly introduced into the system by installing other software.

Originally posted by thepcmd
I currently use esignal with the Kwippop indicators. I am currently trying to back test a theory. I am using a 1 minute ES#F. I can load the chart but when I set the time template to 240 days - esignal crashes and closes. It appears there is some type of buffer limit that is reached and then crashes and has to close. I have determined that I can load only 18 days on a 1 minute chart to do my backtesting. Is there a setting in esignal that I can increase to allow the backtesting of a longer period even if it is not the complete 6 months. I get the same result if I use a Dynamic timeframe and continue to go back in time - when I reach that 18 day threshold esignal crashes and closes. Any help would be appreciated. Thanks
My Machine is a 4GHZ, with 2GB of RAM and plently of HDD space so I do not think it is related to my machine,

[demarcog]

Re: Re: Backtesting issue

Originally posted by stevehare2003
thepcmd,



FWIW, I have several very complex strategies running (4 timeframes, liberal use of efsInternal and efsExternal), multiple charts, etc, with a 4 year old machine. Even though I have 1Gig of memory, I rarely have these kind of problems. When I do, invariably, the reason is either my code or something I have inadvertantly introduced into the system by installing other software.

Steve,

I'm curious what's the CPU(s) configuration, Processor Type/Model, features, clock speed, number of cpus's etc, on this "older" machine?

With all your studies loaded can you please try loading 120 days of 1 minute ES #F using a 24 hour template and let me know roughly how long it takes?

Are you able to run a fairly complicated backtest on this much data?

I really hope you right as I desperately want to run my strategies on the entire extent of intraday data.

Perhaps the problem I'm having is being caused by some programming issue within my EFS, but why does it only occur when more then approximately 90 days worth of 5 minute data is loaded regardless of the EFS being backtested as that seems to be my limitations.

I've given up on trying to load 1 more 30-60 days of 1 minute 24 hour ES #F as most of the time it's an Esignal restart to get out of it.

It would be interesting to see who can and cannot load this amount of data and run a back test and perhaps I open a "poll thread to find out.

Thanks

Glen




thepcmd,


Are you able and approximately how long does it take to load up the ES #F 1 minute data using a 24 hour time template?

Also can you try running one of the sample backtester EFS's in and let me know how long it takes.