Trendiness of Days - assigning a value

Discussion in 'Technical Analysis' started by TGregg, Nov 20, 2003.

  1. acrary

    acrary

    Here's a example from 11/19. The regression line is in Yellow.
    Using 15 min. periods to fix the number of data point on the run and using the slope = rise/run formula you get slope = 1042.45 - 1035.32/27 = .2640 for the day. This could then be used to compare with other days.
     
    #11     Nov 21, 2003
  2. acrary

    acrary

    Here's another example using today. The slope = (1034.50 - 1034.03) / 27 = .0174. So you could say today was about 1/15 as trendy as 11/19. If you wanted to measure the trendiness for each 1/2 hour you'd use a regression line for each period and average the slope number for the day.
     
    #12     Nov 21, 2003
  3. man, there is NO WAY I could make money with that stuff. Tony Crable, yes, the rest no.

    Maybe I misinterpeted the orginal thought of the thread, which was not the idea of a trend day of moving from low/high at the open to the other extreme at the close. I thought the idea of the thread was more along the line of followthrough.

    In that case, the "trendiness" is simply, when the market breaks an obvious pivot point, does it keep going so I can make money, or does it quickly stall back into the chop. Thats all.
     
    #13     Nov 21, 2003
  4. TGregg

    TGregg

    That's an interesting idea, and a new one to me (unschooled in stats as I am). But let's take the day I posted this. We had a nice uptrend until 1 or so (as I recall), then a downtrend for the rest of the day. Would a least squares take that as a range day? I mean, compare that to today (a day I would call a Range Day).

    I haven't dug into this as well as I will over the next few days, so there is likely additional insights that await me. And, thanks for the ideas.
     
    #14     Nov 21, 2003
  5. TGregg

    TGregg

    Yeah, it's kinda that. Take today as an example of what I would call a Range Day. Market set it's high and low for the day early on, and respected that for the rest of the day - that's a Range Day, and should score very low on our Trendiness Scale.

    We've all seen days where the market just keeps going up all day, or keeps going down all day. That's a very high score on our Trendiness Scale.

    Sometimes (rarely?) we have a double trend day, like the day I started this thread. That'd score well (IMO) on a Trendiness Scale, two trends.

    But I don't know how to compute this value.
     
    #15     Nov 21, 2003
  6. It seems Crabel's trendiless concept above would cover basically four characteristics:
    1. Daily bar;
    2. Expanding range;
    3. Close - around its bottom/top; and
    4. Open - around its top/bottom.

    Possibly the daily range could be established mainly:
    A. In the first few hours of the day;
    B. During the mid day;
    C. In the last few hours of the day;
    D. Etc. :confused:
     
    #16     Nov 21, 2003
  7. ig0r

    ig0r

    You might be able to improve on this (a cool idea, btw) by maybe multiplying by the correlation coefficient? I guess this would be unneccessary if you just calculate the line from a higher TF, but if you're on say a 1 min chart, multiplying by cc will give you a value that's weighted by trend and chop
     
    #17     Nov 22, 2003
  8. acrary

    acrary

    Ok here's the chart from 11/20. The white line is the regression for the whole day. That would indicate the day wasn't very trendy.
    However if you split the day into 2 halves and do separate regressions you'd see the yellow lines. Then you'd get two separate slopes for each line. Then add the two together and divide by 2 for the average for the day. This would indicate it was a very trendy day. You can do this for any number of periods you think would be a acceptable. I think the overall value to this approach is to use it to measure one period say 2002 versus 2003 to see if the days are on average more or less trendy (by however number of separte regesssions you choose). It's not very tough. I'll be hanging around tomorrow. If I have time I'll run some tests for each year from 1996 - present to see what's going on with the intraday trends.
     
    #18     Nov 22, 2003
  9. acrary

    acrary

    I'm sorry, you lost me. What are the two variables that would be used to determine the correllation coeff?
     
    #19     Nov 22, 2003
  10. Hi All,

    It seems to me that trying to "assign a value" to "trendiness" is a rather fruitless undertaking. If somebody can express in mathematical terms what he means by "trend", I will quickly come up with a suitable metric for expressing this property.

    In fact a poster pointed out that a good measure of trendiness is for him the profitability of his trading strategy. This is probably the only sensible approach but is at the same time very subjective. It has only meaning for the trader familiar with his own strategy.

    :D The trend is your friend :D

    Be good,

    nononsense
     
    #20     Nov 22, 2003