range vs trend and how to predict it

Discussion in 'Automated Trading' started by travis, Sep 29, 2006.

  1. travis


    Here I would like to sum up all the ideas I talked about in the preceding threads, and come up with a conclusion.

    The system will weigh pre-open range, pre-open volume and weekday, estimate whether today is a range or a trend day, and then trade accordingly - within support and resistance on a range day, and beyond such levels on a trend day.

    Look at the attached picture and tell me if it makes sense to you that on a given day an intraday system should be permitted either trend or range trading and never both.

    According to this concept, this week would have seen trades within support and resistance on thursday and friday, and trades beyond S & R on the first three days of the week. (That is, it would have done so provided that the "pre-open weighing" method worked and predicted correctly which were trend and range days).

    On the picture, you see daily candlesticks of the Dow Jones for this week. You can also see "micro-seasonality" at work. Just as August and December, also Thursdays and Fridays tend to be range days, and they were on this occasion.

    By the way, it can't be seen in the picture, but this week confirms my assertions and tests that say that the larger is the range on a day, the larger is the volume (it happens on monday, tuesday, wednesday). Also, I have verified once more that pre-open volume is directly proportional to the final daily volume, and that the pre-open range is directly proportional to the final daily range. In other words, the "pre-open weighing" method predicted correctly the whole week.
    • snap.jpg
      File size:
      39.7 KB
  2. Wow, 62 downloads and no comments, huh?

    It's a great chart, and I know you understand what you're saying, but we aren't putting the pieces togther.

    If you want to break-it-on-down and show you analysis step-by-step it would probably help.


  3. Not enough info on your chart. I do not think there is any better way to predict direction than looking at a simple price chart of what ever timeframe you wish to trade and seeing if it seems to be going up, down or sideways. You need a few dozen bars at least.
  4. I trade both range bound and trend systems and have long held the belief that you first identify the type of market you are in and then apply the correct strategy to that market. That is you trade the strategy that is likely to produce the results you are after.

    Your approach is an interesting theory, echoing some of my thoughts and research on the subject. I think you are on the right track, but I would caution you that your data sample is no where large enough to draw conclusions and put money behind it.

    Good luck!
  5. travis


    Thanks to everyone for the comments.

    It's true, as you guessed correctly, I have not tested this system I am talking about. The reason is that my system uses such small timeframes and needs so much information for one day, that I couldn't find the data, nor the program capable of doing such tests. Still, this doesn't prevent me from using these concepts in my trading because first of all they seem to work and as soon as they stopped I would stop investing money into them. Second of all, I did do some tests, which I explained in my latest threads.

    These tests are not tests of my entire system with drawdown and so on, but are tests of the specific concepts I asserted, and they proved to me that:

    1) Daily volume has a positive correlation with daily range

    2) Pre-open volume has a positive correlation with daily volume (therefore I can use it to predict daily range)

    3) Pre-open range has a positive correlation with daily range (therefore I can use it to predict daily range)

    4) Days of the week are not all equally bullish: mon tue wed rise much more than thu fri. They are not equally inclined to be trend rather than range: thu and fri are more likely to be range, just like August is likely to have range days. Therefore I can use this as well to help me predict whether today will be range or trend.

    So I weigh all the above information and get a coefficient which forecasts pretty correctly whether or not today we will move beyond support and resistance. Of course now I should tell you how I get support and resistance, but I will stop here, or I would have to keep on writing for a few days.
  6. Although I am still testing my automated system but I thought to put in my 2 cents.

    in my experience to differentiate between trending day and range day you need to watch the market delta.

    I usually wait for the first 30 minutes (I am talking about stocks) to pass then I start counting the market delta, in range days you will usually find the mean of the delta to be round 0 (zero) and for trending days you will find otherwise.

    I hope this help.
  7. travis


    This makes a lot of sense - wait 30 minutes and see where the market is. I should test this as well, though, before using it. Right now I am kind of tired of all this testing, but I will keep it in mind for the future. Thank you.
  8. I know exactly what you mean by being tired of testing :)

    people think trading is easy and some think automation will just pour gold on you, ok think again :)

    it takes days and nights to just grasp the smallest concept and put it in action and still there are lots out there to learn

    Good luck.
  9. Interesting observations...although i think it is one thing to make observations of correlation, and another thing to create a tradable system out of them with enough average profit to offset commissions slippage, and other costs and make a profit to make it all worthwhile. It sounds like you are using these concepts as auxiliary filters to enhance systems based on other concepts perhaps. In that case, these observations can be quite useful. Another question is how you would be able to determine that your ideas have "stopped working"...keep up the good work
  10. A montecarlo analysis will help with that. You can get an idea of what the expected drawdown will be and if you actual drawdown is greater then would be expected, your system has stopped working.
    #10     Sep 30, 2006