backtesting the original turtles

Discussion in 'Trading' started by mind, Apr 20, 2004.

  1. mind


    i tested the concept of the website on data from 1993 until now and it turned out that the system was not profitable. i was pretty aware that it would not be a great thing - because it was public for quite a while - but honestly speaking i did not think it would not work at all.

    did anyone here come to the same conclusion on this?
  2. Trying to use the turtles on a single market is worthless.
    Dont know how it performs the way it was meant to be used these days.
    It sucked on the Dow/SPY when I tested it.


  3. rwk


    I did a fair amount of testing on channel breakout (CBO) back in the late 1990's. My conclusion then was that the concept has merit, but that it is very hard for most people (myself included) to implement. What I like about the system is that when a trend develops, I am guaranteed to get an entry signal (which is definitely not true for all systems). What I don't like about the system is that it tends to experience VERY large drawdowns.
  4. DVB


    I did test it as well and I do agree that the performance is disappointing. However I don't think it's due to the fact that the system's been public; I think it's because the cycle has changed. I am sure that the system of buying 20 or 55 days breakouts will work again sometime in the future. But by the time we realize it, the cycle will change again. There will be some lucky traders who will make fortunes by chance, but you are not successful in trading until you quit, because your largest drawdown is possibly sometime in the future.

  5. T-REX


    There is MORE than just 20 day breakouts and the like.
    Markets are NOT static so NO static played out rules will continue to work built on a static premise.

    A breakout system will work during those types of market setups.
    othertimes when the mkt is not trending you need to switch strategies and trade your range bound system while simultaneously running the trend following mechanism.

    This way you won't be late entering into the trend once it develops.

    IMHO this is the best way.

  6. mind


    i did it on the whole original portfolio, including all the portfolioCorrelatioMoneyManagement-stuff. single markets are impossible.
  7. mind


    i would not mind the draw downs, if the thing overall made money, but it does not. i would see a strategy of this kind tradeable, if test sharpe ratio are near or above 1, and i would consider it good with sharpes above 1.5. our test since 1993 had negative sharpe.
  8. mind


    it is not just the 20 and 55 variable, the concept as such seems outdated.
  9. mind


    i am building automatic strategies. each of them is by definition as a "system" to some extent static. i am constantly trying to add new stuff in order to replace things that have passed the peak of their lifeCycle as a system. long term trend following has some nice features and usually makes money when shit happens. if overall expectation value is positive i do not worry about the draw downs, but this concept has negative expValue.
  10. rwk


    When a trend develops, all good trend-following systems will issue a signal. Some systems will issue a signal sooner than others, which is usually a function of parameter settings and filters. If the trend is a good one, the sooner you enter, the more money you make. If the trend fizzles, the slower systems will save you money. That's the basic tradeoff.

    There are some approaches that can miss a good trend altogether. For example, volatility envelope systems require a price spike to signal an entry, and a creeping trend can go a long way with no spikes. Swing traders usually wait for a pullback, and sometimes the best trends just run like they're on fire.

    CBO is a good, simple system that is guaranteed to give you a signal on the right trend. The problem is that it will also signal a lot of losers, as will most other trend-following systems. Probably the biggest error is to think that key to profitability is some certain filter or parameter settings. There is no way to predict which filters or parameter settings will work best going forward. The 1990's were generally not very friendly toward long term trend following. I am hearing that the last couple years have been good. T-REX's approach of sitting out the bad periods has a lot of appeal. I just don't know anybody who has been able to do that. (T-REX: Lend me your crystal ball!:D)
    #10     Apr 20, 2004