## New Tool: Relative Strength

January 25, 2009 by dskills

I’m starting a new feature on the blog – I’ll be reviewing key sectors and the relative strength of each of them. I’ll be posting this using a new proprietary tool I’ve developed. This will be represented graphically by this new indicator over the prior 100 days. I may end up doing 50 days as well if people are interested – let me know.

No shocker this week – we’ve got strength in Gold and Bonds. We are seeing some relative strength in small cap stocks which, in my opinion, is key to the overall recovery of the market.

### Like this:

Like Loading...

*Related*

on January 26, 2009 at 4:46 pm |John DyerRelative strength is very important in my trading. Most use a 126 week period. Have you considered 6 months?

Thank you, jonedyer

on January 26, 2009 at 4:50 pm |dskillsJohn – in my own systems I use a combination of 1 year, 6 months, 3 months and 1 month RS. I just use this to give me an overview of where the market is at over a relatively short period. I may end up publishing this for those periods if people are interested – what would be useful to you?

on January 27, 2009 at 3:44 pm |John DyerYes, your relative strength maybe a most usefull method to evaluate where the market is right now. Michael J Carr found that he had best results useing alpha. Is alpha part of you calculation?

on January 27, 2009 at 3:56 pm |dskillsIt’s not – how does he define alpha in this case?

on January 28, 2009 at 4:42 pm |John DyerHe uses linear regression to calculate alpha and beta coefficients. He said that alpha provides a measure of how much a stock’s return is derived from factors unique to the company. He also quoted Robert Pierce who defined alpha as volatility-adjusted RS.

If anyone knows how to use Excel to calculate alpha, I would appreciate the information.

on January 28, 2009 at 5:38 pm |dskillsI hate the terms used by Carr in the book – because he creates new definitions for existing terms. First, he defines alpha and beta as the y-intercept and slope (respectively) of a linear regression of returns for two instruments.

Then he uses a definition created by Robert Pierce which is as follows:

- Calculate RS of all instruments over a 99-week moving average of price changes. [lacking definition here a bit]

- Beta is defined as the percentage change between the lowest price in a time period and the highest price in that same time period [note that he doesn't make it clear whether or not that is the same 99-week period mentioned above]

- Alpha is the ratio of RS to Beta.

He then goes onto something else and doesn’t complete the thought – that is, he doesn’t show how he would actually calculate alpha in his model. Later, during the comparison of different RS techniques, he states that alpha is simple the intercept of the general market and the instrument being tested. Ok, not very helpful.

Later he also shows that the alpha technique doesn’t work as well as others, so maybe that’s why he doesn’t get a lot more specific. He found that the Ratio of moving averages worked best. I haven’t been able to validate his finding in my own work – meaning I have not found that to be a good technique.

Oh, and don’t bother with the software he mentions in the book – really not ready for prime-time.

on January 29, 2009 at 5:01 pm |John DyerThank you for your complete, well thoughtout reply. You cleared up my some of my flustration with Carr’s book. Thought I must be missing something regarding his coverage of alpha!

The not ready for prime-time software company he mentioned has folded. Thanks again!

on January 29, 2009 at 5:21 pm |dskillsNot a problem John – thanks for your comments.

on April 11, 2009 at 1:33 pm |Michael CarrAlpha can be caluculated with a 5-column spreadsheet in Excel.

Col 1: Date

Col 2: Stock price

Col 3: S&P 500 price

Col 4: % change in stock

Col 5: % change in S&P 500

Graph the % change columns as a scatterplot, fit a trendline (linear regression) and display the formula as y=mx+b. The ‘b’ component is alpha.

Pierce’s alpha is provided as an example. Please contact me via email for his paper if interested.

I belive alpha/standard deviation works best and use that. For individual traders, I think that other techniques are better since they are easier to understand (ROC and ratios).

Monocle went out of business – I had no affiliation with them and found the software useful. TechniFilter can be programmed to do the same things.

on April 11, 2009 at 2:15 pm |dskillsMichael – thanks for your comments. I’d love to see the Pierce’s alpha example – I’ll reach out to you via email.

Questions about the above:

1. % change in stock/S&P500 – do you mean daily?

2. I couldn’t see how you could do it in Technifilter – any pointers on that?

3. I don’t know how to “display the formula as y-mx+b” in Excel – I’ll have to see if I can figure it out.

Thanks again!

on April 11, 2009 at 11:28 pm |Michael CarrPierce’s paper is at https://www.mta.org/eweb/docs/Issues/48%20-%201997%20Summer.pdf

It starts on page 11 of the pdf file.

on April 11, 2009 at 11:33 pm |Michael CarrIn Excel, create the chart from columns 4 & 5 in the above example. You need to regress % change; regressing prices directly will usually show a very high correlation due to some autocorrelation that exists bewteen the series.

Create the chart as scatter plot. Then click on a data point and you should see an option to “add trendline” select that and in the type tab, select linear regression. Under the options tab, select ‘display equation on chart’

The slope intercept form of finding alpha and beta is drawn directly from Sharpe’s original work.

on April 11, 2009 at 11:34 pm |Michael CarrWill find the TechniFilter reference by Monday.

on April 13, 2009 at 2:20 pm |Michael CarrIn the TechniFilter Program Guide, Chapter 3, Session 4 provides an example of how to do RS testing.

on May 23, 2009 at 9:54 pm |trader847I have read Michael Carr’s new book and have only been able to use the moving average ratio as an RS tool. I was happy to see this discussion because I wanted to use the alpha strategy in Tradestation but was unable to figure out the calculation. Since I’m doing a linear regression on the price changes of the stock and S&P 500, do I use daily price changes or weekly and how many data points do I need. I wanted to do the regression in Tradestation and not Excel and form an array of price changes for each stock and the index so that I could run the calculation on a universe of stocks. What is the exact calculation for alpha? And can I just select the high alpha stocks; what need is there to divide by the standard deviation of the alpha’s for the universe of stocks. Thank you – Frank

on March 2, 2011 at 3:34 pm |Rick BajacksonI have been using 6 month RS on ETFs, fairly successfully. I was just wondering if you are still trading based on RS.

on March 2, 2011 at 3:36 pm |dskillsYeah I’m still using it – it’s been great. I’m now posting over at http://www.etfprophet.com by the way – not so much here.