Can a slope which exceeds a value be considered a pattern?

kmiklas

kmiklas

Assume two price points are taken at time t; say, at USD29.43 at time 0, and 32.78 at time 1.

The slope of these lines, given by the good-ole' fashioned rise/run method is:

Code: 
     y2 - y1     32.78 - 29.43     3.35
m = --------- = --------------- = ------ = 3.35
     x2 - x1         1 - 0           1

If we recognize as a signal a slope greater than 3 between two price points in a time period, is this a pattern? Generally stated: for any two price points, if slope m >= some limit l, in time t, does this constitute a "pattern?"
 
I did a similar analysis using regression. I plot the change in m distribution by sector. This distribution profile is similar the daily price change.
slope.jpg
 
While at it, you may want to read zigzag which simply smooths the brownian model into a st line by connecting maxima and minima at major rotations. Now you have many little lines to deal with and apply your regression indicators or whatever you are thinking of to find potential future entry / exit.
 
Can a slope which exceeds a value be considered a pattern?

No. Think about it, price is a highly random variable, and the lookback period is randomly chosen as well. Why would a 10 day lookback be any better than an 11 day, or 27 day, or 2 day?

You may luck into a lookback period that provides positive expectancy, but that would require lots of careful backtesting to prove so. Probably not worth the time or effort on your part.
 
panzerman said:
Can a slope which exceeds a value be considered a pattern?

No. Think about it, price is a highly random variable, and the lookback period is randomly chosen as well. Why would a 10 day lookback be any better than an 11 day, or 27 day, or 2 day?

You may luck into a lookback period that provides positive expectancy, but that would require lots of careful backtesting to prove so. Probably not worth the time or effort on your part.
More...
AFAIK trend followers will constantly review and test lookback periods to measure the strength of the signal. They will typically apply a kalman filter on the initial data to reduce noise. I worked with a team that did this (a handful of PhDs and traders).
 
You must log in or register to reply here.
Back
Top