Not logged in
Viewing 6 posts - 1 through 6 (of 6 total)

Forums>StrategyQuant>General Discussion>Predicting Trend Changes

  • #114375 |
    Participant
    81 Posts

    I purchased this program about 6 months ago and honestly just having a hard time using it. But my question is:

    Would this computer program be able to predict trend changes and be able to analyze them to determine the average duration of time that a specified trend should last, how long these trends last, big or little, last statistically on average based upon pattern recognition?

    #133696
    Customer
    484 Posts

    Systems don’t predict. Magic 8balls and CNBC analysts do. (just having fun)

    That said, you can mine some trend data more easily in SQ4 which I plan to do. You probably could with some tinkering even in SQ3 but its more or less for generating actual systems. Some analysis of the systems can help through forcing it to generate trend following strategies. (set it to h4/D1, even weekly, trend based indicators, no exit rules, low win rate with larger wins, etc). The software takes some time to learn. I still am and had it for a while now.

    A simple example of something I’ll be doing for SQ4 is randomly generating donchian systems that enters/reverse on X day or week new high or low, exit after Y days, profit target Z distance.
    Thats how you can find average trend lengths, distance, and significant turning points robustly. There will always be trends in the future that go much farther than any profit target or time-based exit is ready for. I don’t think systems can predict them or their length (did anyone predict how far down crude would fall and how much time the trend would last?), but trend following systems should always get in them and trail stop the entire way.
     

    #133706
    Participant
    81 Posts

    LOL Yea I get you, if your system could really predict I would love it to let me know the winning Lotto Numbers..

     

    And I guess I must have just wrote it just a little different than what my brain was actually thinking of.

     

    But as an example:  Say if I wanted to know from Jan. 1st 2015 until Oct. 30th 2015

                  Could it tell me say SMA 10 vs. SMA 20, what is the average time 10 > 20, what is the average time 10 20 & 10 20 & 10 < 20, etc…..  

     

    I think you see where I am going with this but just unsure how to either tell it or ask it to provide me those details and if not what do you advise?

    #133724
    Customer
    484 Posts

    I think EA Wizard is better for data mining since it is more flexible for building what you want to mine, until SQ4 arrives. For discretionary trading, you can data mine robust/simple ideas like testing the candle patterns for example. Daily timeframe, if hammer, enter long at market next bar open, close on the close. No stop loss, no target, default lot 0.1. See how much accuracy they have or find the pair where they have the most value. Don’t care about equity curve really, just win% since it has no money management. Stuff like that. A system as simple as that can actually be done in SQ3. You can take it even further and plug something like that into the “Improver” fuction, and have SQ try to add an additional entry criteria to give it more value, and since the system is so simple, its not being overly curve fitted and has some statisical value for discretionary trading. In defense of using EA Wizard- you can use its math functions to program custom candles if you don’t like the default one’s definitions, which wont work in SQ.

    For your moving average examples you can do it in either program and Quant Analyzer or SQ will be able to return # of bars in trade. Maybe use ADX to find if its trending. Like ADX>25 and Close1>50SMA The Enter/Reverse at market. Use no stop loss or take profit and optimize the moving average and exit rule should be- If close1<50sma then close position. Alternatively can add # of bars exit with optimized value. Then you can get the trade statistics that you want. The idea isn't to build a winning system (though that is also helpful) for data that will be used with discretionary trading, just to seek something very specific. In this case how many bars it spends above/below an MA when trending.

    Hope that helps.

    #133741
    Participant
    81 Posts

    Yes that does help. Is their a particular thread or operational manual that I can read that will further explain the complete processes you have described above? I asked because I am not fully versed in either programs but do say have a mediocre understanding of all of your programs. But if not I still appreciate your response and will try to figure it out as best as possible.

    #133746
    Customer
    484 Posts

    Pretty much trial and error and occasional moments of inspiration in the process.

Viewing 6 posts - 1 through 6 (of 6 total)

You must be logged in to reply to this topic.