walk-forward optimization software?

Discussion in 'Trading Software' started by headcutter, Sep 16, 2003.

  1. The only good way to backtest is in my opinion the walk-forward optimization(which gives the real results you would really achieve) for example I optimize the most profitable strategy for 1993-1994 and then use the most profitable inputs for 1993-1994 in yesr 1995. I would do so in the years 1995-2002. Is there any software that can automate this process? Manually I would have to backtest 8 times and have 8 numbers multiplied in EXEL which is very primivite and I waste time.

    Thanks for any info
  2. Roscoe


    Agree entirely!

    I use PowerST to do walkforward testing (www.powertesting.com) in conjuction with some Access database charting to visually evalulate the results.

    Basis is:
    1. Set up a universe of market/systems;
    2. Define timeslices (95-98, 96-99, etc);
    3. Define parameters to be optimized;
    4. Use a macro to run the test;
    5. Output to MS Access;
    6. View resulting charts.

    The only way to see if your parms are robust IMO.
  3. if you could compare how hard it is too program a strategy in PowerST then in other programs like TradeStation, AmiBroker, wealth-lab would you say it is more difficult or not? Also is there anything else you could say that would help me decide if this is really a good software since it costs $2500
  4. Roscoe


    Difficulty is difficult ;-) to define - PowerST is VERY powerful and VERY flexible, but the payoff for all that power and flexibility is more code.

    I certainly prefer it to TradeStation due to the lack of coding limitations - PowerST uses a subset of C++ as its native language so pretty much anything is possible!

    As to the price: I consider it minimal when you consider the possibilities this package has. I was going down the road of developing my own app when I found this, and it has saved me thousands, probably more, over a DIY approach.

    Summary: PowerST is the cutting edge IMHO, but it is not a simple charting package, rather a monster of a testing package that covers everything from systems and money management development thru testing, portfolios and trade signals.

    If that what you are after and you are prepared to do some serious work to achieve it then buy it.

    OTOH if all you want is a pretty charting package with some testing capability on the side try something like NeoTicker. (Sorry Lawrence!)

    Feel free to email me privately here with any specific questions (and no, I am NOT selling anything, just a very happy user).
  5. Guys:
    I hope the people who are developing this package are reading the posts. One of the problems with getting a package into widespread current use is that people who might want to use it can't tell much when they go to the site. Another problem is that when you try to use one of these packages, the learning curve is so daunting that you feel like you will never get to the point where you are able to make real use of it. What generally makes this kind of package a non-starter is lack of customer support. If the folks who are developing the product are smart, they will realize that the ONLY way to get past these hurdles, is to reach out to the public, choose a user or two (the more the better) and make them into "projects". Put resources into training them to the point where they can really accomplish something with the product, and encourage them to make posts in chatrooms like ET talking about their experience. If Tradestation would have done that early on (its still not too late) think about what kind of business they would be doing now! Just a thought. Steve46
  6. "if all you want is a pretty charting package with some testing capability on the side try something like NeoTicker"

    Does NeoTicker have any more features then TradeStation2000i?
    If not then I am really not interested
  7. A quote from Larry Hite in Market Wizards(at the time he managed 800 Million)

    "Its Incredible how rich you can get by not being perfect"We are not looking for the Optimum method,we are looking for the hardiest method,Anyone can sit down a perfect system for the past.
  8. CalTrader

    CalTrader Guest

    I would be very careful with so-called "walk-forward" optimization techniques. More often than not the implementation of these is flawed in one way or another: Make sure you thoroughly understand exactly how this is being accomplished. If you dont, or cant get the details then my rec is to avoid the vendor/implementation.
  9. PowerST


    I like this post! It describes exactly the approach PowerST has been taking. For example, Roscoe, the first person who mentioned PowerST in this thread, started with PowerST in March, 2002.

    The www.powertesting.com web site is only for initial screening. Clearly, the intent of the site is only so that someone who might be interested in starting as an "Early Adopter" will email, and we can discuss one on one whether they would be appropriate as an early adopter. The sales effort is very low key at this point because even with the low key sales effort there are sufficient inquiries for adding new customers slow and steady. The point is to get each new customer up and running, make sure they are happy and working productively, then add more customers. Exactly as you say in your post.

    PowerST is not a non-starter. In fact it is already well underway. There is already a small community of users who (to paraphrase your words) have been trained to the point where they are already really accomplishing something with the product. Clearly Roscoe, the person who first posted, is saying that he is currently accomplishing something with the product. You can also ask Roscoe about lack or non-lack of customer support (obviously I am sure he will say the support has been very good).
  10. Roscoe


    Download the demo and find out: www.tickquest.com or www.neoticker.com.

    Very slick, uses VB for code.
    #10     Sep 17, 2003