Use access key #2 to skip to page content.

Alpha, and statistics in investing

Recs

13

August 22, 2012 – Comments (11)

I am currently studying for CFA, and right now I am studying the statistics part. While I find statsitics to be a relatively easy math, I don't see its usefulness in investing.

 I don't think you can use past data and analyze it statistically to anticipate the likelihood of an event occuring again with stocks.  Stocks are chaotic and constantly changing, unlike observing something in nature.

 

Also, I think the concept of alpha is ridiculous.  If a bluechip company has historically moved up or down only 1% per year, but it is in a dying industry...compare that to a company which got slammed because of the recession, and is down 90% from its recent high, but will survive the downturn.  So the company that is already down 90% is riskier, just because it made a bigger move?  Who cares what the "risk-adjusted return" is.  The only thing that matters is the ACTUAL return.  Volatility does not matter as long as you don't buy and sell at the wrong times, and if you are doing so, you should not be managing money as a career in the first place!

 

Discussion is welcomed. 

11 Comments – Post Your Own

#1) On August 22, 2012 at 5:07 PM, Valyooo (99.33) wrote:

Which is why I think the Sharpe ratio is absurd.  Oh you earned more money, but you COULDVE lost more, because one time that sector lost more.

Report this comment
#2) On August 22, 2012 at 5:10 PM, athebob (87.39) wrote:

I agree, the math needed for investing can all done on a 4 function calculator with the exception of compounding(which you can always do with iterations).

Report this comment
#3) On August 22, 2012 at 5:21 PM, trackjakeambrose (59.87) wrote:

   The only comparable thing to repetitive nature in stocks may be the psychological commonalities in investors which of course the investors ultimately control the share price.The way they repeat how they handle and react to certain elements. I know though that isnt the main point youre really hitting on. The main thing that does is simply create opportunities for the more educated investors to get out at overly high prices and buy in at the right entry points or better.  I agree with what you are ultimately hitting on though. You definetly cant look on the past to try to simply think the same cycle is going to repeat itself. The economy is ever changing that effects things, the evolution of many things, the business changes, different people may be running it......so much i could mention things all day. I also like what you mention at the end, lol, as some one myself who wants to be a money manager, i see people choosing the wrong entry and exit points as one of the biggest most costly mistakes many investors make.

Report this comment
#4) On August 22, 2012 at 5:37 PM, Mega (99.95) wrote:

I think alpha and sharpe aren't useful for evaluating individual securities, they are only useful for evaluating asset managers.

Alpha looks like this (see how it rises somewhat invariant to the market swings):

Report this comment
#5) On August 22, 2012 at 5:40 PM, Valyooo (99.33) wrote:

@megashort,

 

Why are they good for evaluating asset managers?  I would think beta would be better there...to see if a manager is good, or just uses leverage. 

Report this comment
#6) On August 22, 2012 at 6:01 PM, Mega (99.95) wrote:

Well alpha is a return measure, beta is a "risk" measure.  Both sides are important.  So I think risk-adjusted return (Sharpe, Sortino, etc.) is the best single number.

Even if you believe that historic volatility is a poor measure of risk in individual securities, I think it can be more useful when evaluating asset managers. If they just aim for returns without regard to volatility, they probably won't last long managing other people's money.

Report this comment
#7) On August 22, 2012 at 7:28 PM, TheDumbMoney2 (98.37) wrote:

Valyooo, you should get on twitter and follow some of the people I follow, some of the traders like @TraderFlorida, asset managers like @TheReformedBroker, and fundamental analysts like @EddyElfenbein, tech people like Leigh Drogan.  You would learn a lot, and you would really enjoy their tweets and worldview.

Report this comment
#8) On August 22, 2012 at 8:15 PM, somrh (88.31) wrote:

One critical aspect is that the academics who support these ideas (FWIW, CAPM was critiqued and abandoned as early as Fama and French 1992) is the belief that markets are "efficient". Your dying blue chip and the stock that already got slammed 90% have prices which (according to their theory) adequately reflect the present value of the market's best estimate of future cash flows discounted to the present (using an appropriate discount rate depending upon the "risk" of the asset). 

The idea of a "risk adjusted return" is a nice idea in principle but not so much in practice. In principle, returns should be an increasing function of risk. If I can get the same return on a relatively risk free investment versus one that is much riskier, I would much prefer that the low risk investment. I'll only go for the more riskier one if I think returns can be better.

In practice, it assumes that we can adequately quantify "risk" and then plug that into an equation which spits out returns.

I concur with you that standard deviation or beta may not adequately encompass "risk". Frankly, I'm skeptical that risk is easily quantifiable. Some risks (I prefer to think there are many kinds of "risks") can be but others cannot. I think Taleb's quadrants offer a good insight into that. See here:

The Fourth Quadrant: A Map of the Limits of Statistics

It seems to me that the peope who are less skeptical of these models seem to be implicitly assuming that all risk is like the risk involved in casino games (which can be modeled quite nicely).  

Report this comment
#9) On August 22, 2012 at 10:09 PM, awallejr (85.46) wrote:

And then there is fibonacci.

Report this comment
#10) On August 23, 2012 at 12:08 PM, ikkyu2 (99.27) wrote:

I came to the same conclusions when I got my master's in statistics, Valyooo.   The statistical methods most often used to analyze stock movements or rate people who manage stocks contain underlying assumptions that simply aren't true in the case of either stock price movements or stock managers' actions.

Report this comment
#11) On August 24, 2012 at 12:35 PM, Melaschasm (55.93) wrote:

Past results do not guarantee future returns, but history tends to repeat.

Statistically measuring the stock market as a whole provides many useful insights.  For example, there are one or two major market crashes that coincide with a recession every decade.  Thus if we are 5+ years into a bull market, it is time to start getting more defensive with your stock investing.

If we look at mutual fund performance, one of the best predictors of future results is the 10 year rate of return.  One of the worst indicators is the return during the most recent quarter.  It is so bad that I would be tempted to short  any mutual fund that has a quarter twice as good as their previous year average.

Statistics tend to work best when there are a very large number of data points.  When analyzing  specific company, data is more limited than for large markets.  We can still evaluate some trends with certain companies, but we should always consider qualitative analysis in addition to quantitative analysis.  For example certain companies tend to be less volatile than others.  This generally means the company's stock price will go up slower than the market as a whole during bull markets, and fall less than the market during bear markets.  This is great quantitative data that provides vital investing information, but it also needs to be considered with other analysis because sometimes a legitimate market shift changes the long term trend for a company, or even an entire industry.  For example, Blockbuster was disrupted by Netflix, ending their normal historical fluctuations.

Report this comment

Featured Broker Partners


Advertisement