Jump to content



Photo

For Those Who


  • Please log in to reply
4 replies to this topic

#1 U.F.O.

U.F.O.

    U.F.O.

  • TT Patron+
  • 5,605 posts

Posted 25 February 2009 - 07:56 PM

MaryAM and goldswinger, here's some additional info. This is a 15-min $SPX chart and I could see where one might find benefit for VST horizons. I couldn't find a formula, I was referred to the book "The New Technical Trader" by Tushar Chande and Stanley Kroll. A lengthy interpretation is below the chart. (1 chart)

U.F.O.

Posted Image

Interpretation

r-squared values show the percentage of movement that can be explained by linear regression. For example, if the r-squared value over 20 days is at 70%, this means that 70% of the movement of the security is explained by linear regression. The other 30% is unexplained random noise.
It is helpful to consider r-squared in relation to Slope (see Linear Regression Slope). While Slope gives you the general direction of the trend (positive or negative), r-squared gives you the strength of the trend. A high r-squared value can be associated with a high positive or negative Slope.

Although it is useful to know the r-squared value, ideally, you should use r-squared in tandem with Slope. High r-squared values accompanied by a small Slope may not interest short term traders. However, high r-squared values accompanied by a large Slope value may be of huge interest to traders.
One of the most useful way to use r-squared is as a confirming indicator. Momentum based indicators (e.g., Stochastics, RSI, CCI, etc.) and moving average systems require a confirmation of trend in order to be consistently effective. R-squared provides a means of quantifying the "trendiness" of prices. If r-squared is above its critical value and heading up, you can be 95% confident that a strong trend is present.

When using momentum based indicators, only trade overbought/oversold levels if you have determined that prices are trendless or weakening (i.e., a low or lowering r-squared value). Because in a strong trending market, prices can remain overbought or oversold for extended periods. Therefore, you may want to reconsider trading on strict overbought/oversold levels used by many indicators. An "overbought" market can remain overbought for extended periods in a trending market. However, a signal generated by a moving average crossover system may be worth following, since these systems work best in strong trending markets.

To determine if the trend is statistically significant for a given x-period linear regression line, plot the r-squared indicator and refer to the following table. This table shows the values of r-squared required for a 95% confidence level at various time periods. If the r-squared value is less than the critical values shown, you should assume that prices show no statistically significant trend.

Number ofPeriods r-squaredCritical Value(95%confidence)
5 0.77
10 0.40
14 0.27
20 0.20
25 0.16
30 0.13
50 0.08
60 0.06
120 0.03
You may even consider opening a short-term position opposite the prevailing trend when you observe r-squared rounding off at extreme levels. For example, if the slope is positive and r-squared is above 0.80 and begins to turn down, you may consider selling or opening a short position.
There are numerous ways to use the linear regression outputs of r-squared and Slope in trading systems. For more detailed coverage, refer to the book The New Technical Trader by Tushar Chande and Stanley Kroll.
"Democracy is two wolves and a lamb voting on what to have for lunch. Liberty is a well-armed lamb contesting the vote!"
~Benjamin Franklin~

#2 MaryAM

MaryAM

    Member

  • Traders-Talk User
  • 1,200 posts

Posted 25 February 2009 - 08:39 PM

I use linear regression a lot in my work - I have never used it in TA because the first requirement for the data is that they must fit a normal distribution. I would play with it an run R squared by first log transforming the data - a natural log should do it - if not try base 10. That should smooth the curves and trend lines and give good predictions over each cycle but it does not look like it will apply to multiple cycles just the cycle your in. Once the data take a turn - then a new data base for a specific trend in one direction should be predictable. Here is where I would apply a sine (sine*R2) function and look for an approach to 0 for another directional turn. I'm fascinated - always been a math nerd. Mary Anne

#3 U.F.O.

U.F.O.

    U.F.O.

  • TT Patron+
  • 5,605 posts

Posted 25 February 2009 - 08:46 PM

MaryAM, I'm a blue collar bond trader who took 2 times to get through college algebra. :blink: I'm afraid applying the tweaks to this indicator that you suggest would be above my pay grade. :P U.F.O.
"Democracy is two wolves and a lamb voting on what to have for lunch. Liberty is a well-armed lamb contesting the vote!"
~Benjamin Franklin~

#4 goldswinger

goldswinger

    Member

  • Traders-Talk User
  • 2,612 posts

Posted 26 February 2009 - 01:09 AM

UFO, Thanks for the feedback, I use Stockcharts.com for my trading, do you know any charting service that provides R Square charts and options.... Thanks, Goldswinger.

#5 U.F.O.

U.F.O.

    U.F.O.

  • TT Patron+
  • 5,605 posts

Posted 26 February 2009 - 08:21 AM

Only MetaStock gs. Do a Google, might be someone else out there. U.F.O.
"Democracy is two wolves and a lamb voting on what to have for lunch. Liberty is a well-armed lamb contesting the vote!"
~Benjamin Franklin~