Good Morning,
Over the past few months there have been a few comments regarding 2nd order polynomials (parabolas) as a means to approximate the price curve and sigma bands as a way to measure the chances of a stock moving higher. I was able to find a graphic example of this at http://sigmabands. blogspot. com/ . According to the author / blogger the center line in based on 250 bars worth of data and the sigma bands are each spaced one standard deviation apart. I am assuming that the standard deviation data is also based on 250 bars, although this is not mentioned specifically.
What I thought was quite interesting / impressive about these examples is that, visually, it appears that the price series does spend roughly 2/3rds of the time within the first sigma bands and roughly 2% of the time outside the third sigma bands. Exactly what you would like to see. Bollinger Bands, to me, have never shown this level of consistency.
I purchased Umits Trend Toolbox (I have no financial interest in this product at all) so that I could plot a second order polynomial fitted to the price curve. This curve on my charts was the same shape as the those on the web site, but roughly 10 pts or 1% higher; not really a big deal. Where my efforts began to stumble was with the Sigma bands.
To get the sigma bands equally spaced at all point I used to following construction:
SD:= LastValue(Stdev( C,60));
And then I added / subtracted this value to / from the centre line to create my first sigma bands.
As an aside, I am thinking about this to smooth out the Standard Deviation value:
SD:= LastValue(Mov( 20,Stdev( C,60),s);
The key difference is that my standard deviation value on the S&P 500 is near 80 and the website lists theirs at 25. Iʼm off by a mile. Additionally, as you would assume, nearly 100% of prices are within the first standard deviation; this should not be the case. This pattern repeated itself on a number of individual stocks that I tested this on.
Perhaps someone who has a better grasp of statistics might be willing to share why Metastockʼs Stdev value is so significantly different?
I know that I could fix this by just dividing the value by some number between 2 and 3; however, this is not very satisfying.
Also, I am looking at testing an long exit strategy whereby trades are closed when the 60-day parabola makes itʼs first move down.
Cheers,
Cameron
Stay in touch when you're away with Windows Live Messenger.
IM anytime you're online.