Speed up TradeStation

Discussion in 'Trading Software' started by jboydston, Nov 1, 2002.

  1. Does anyone know how to speed up optimization on TradeStation 6?

    What is the determining factor of speed... CPU, RAM, Network, TradeStation servers?
     
  2. Works just fine for me.
     
  3. DaveN

    DaveN

    TS4 was really slow. TS2000i runs a bit quicker. I'd hope that TS6 is slightly faster still. Yes, optimizations are really slow. Compared to other software packages, it's painful to watch TS run through any size optimization. I've compared the same optimization on a 1.9G machine w/ 512M RAM with a 667M machine w/ 512 RAM. *Slightly* faster, but not by much.

    Some points that might help:

    Make sure your code is very clean. For example, instead of calling a function twice, create a variable that will store the value of the function. This helps speed up the code. An example: consider the Stochastic RSI. Instead of saying StoRSI=(RSI()-Lowest(RSI()))/(Highest(RSI())-Lowest(RSI())), use a varaible like LWST=Lowest(RSI()) and RelStrength=RSI(), then insert those into your code. You'll find these opportunities in many places.

    Manage your number of trials. Perhaps take a large step sample (step size = 10 instead of 1), then take a subset of interest using a smaller step size. Or you can code one variable as a function of another to limit your optimization steps. A simplistic example is a two MA crossover system. Instead of testing MA1 from 10 to 100 and MA2 from 10 to 100 (yielding 91*91 trials), you could say MA2=0.1*MA1. Then vary that 0.1 from 0.1 to 1.0 (yielding 91*10 trials).

    This is probably not the answer you are looking for, but I haven't found any hardware solutions that appreciably speed things up.

    If you hear otherwise, please let me know!
     
  4. One

    One

    Sarasota,

    I use TS4 and as you mentioned it can be deathly slow for more complex analysis. I did however find something that dramatically inproved speed, and though I am not sure whether it applies to TS6 or 2000i, this info. may help members using some TS versions.

    Using TS4 in Windows ME, I noticed that as soon as I opened the offline data server, CPU usage as measured by MemTurbo jumped to and stayed pegged at nearly 100%, and that nearly all programs ran much slower (P4 with 512K Ram). As long as the offline server was open CPU usage would not fall except with the following trick: in the server application, click on "file" and then "print setup". By leaving this dialog box open, CPU usage falls back to 0-10% and the user is free to toggle to different apps. including tradestation. The key is to open the box in the server application and to leave it open while back testing.



    O.
     
  5. Thanks everyone... do any of you use WealthLAB? I wonder if it's faster than TS6?

    I havn't given up on Tradestation yet... It's not an issue of cost, whatever the requirements are, I'll buy it! ;-)
     
  6. I understand TS7 now supports hyperthreading and dual processor. Has anyone been able to confirm this?

    I'd like to upgrade to a dual-xeon if it will really speed up testing.
     
  7. maxpi

    maxpi

    I don't think that optimizing code will make much difference with TS2000. I wrote a strategy that executed hundreds of lines of code every bar and it did not take much longer to optimize than a simple one. In my case (win98, AMD, 64 Meg, etc.) the normal Tradestation overhead was taking a lot more time than my strategy.

    Max
     
  8. A couple of folks on TradeStationWorld..com tried throwing additional hardware (i.e.: multi-processors) at TS7 optimization. It turns out that optimization in TS7 is still bound to a single processor. Radarscreen claims multi-processor support but I don't know anyone that has tried that yet.

    -eLindy

     
  9. rewriting code better, getting rid of redundant calcs and avoiding div0 checks definately improve speed. other than that, get a better machine. i dont see any speed improvement between ts6 and ts7 in backtesting
     
  10. that TS6 starts ghosting at around 6-7,000, but a quick fix is just to start a quick microspool down to Central. If that doesn't work, I can usually squeeze a few aezons out of the scrim-board for optimal code de-tangle on the 5-9 zhirbats.
     
    #10     Apr 9, 2003