Saturday, June 21, 2014

The 140 page ad

kw: book reviews, advertisements, stock speculating, technical products

If you have a stock broker or money manager taking care of your investments, ask this question:
"If you have two clients that seem identical, in age, sex, life goals, income level and rick tolerance, do you make the same recommendations to both?"
Any answer but, "Of course!" is cause for suspicion. Here is why. Brokers tend to hedge their bets. Suppose in an extreme case the broker makes the same recommendation to every client, and is not just wrong, but expensively wrong. There go most of the clients! The business flounders. So of course, a smart agent will tailor investments to the characteristics of clients. But if a few clients are really very, very similar, they are still likely to get differing advice.

One of the oldest "advice" scams is an either-or tree. This works best if the advisor is very good at advertising, or can afford a top-notch advertising team. We'll analyze the action in exact powers of two for simplicity:

  • Tout yourself as an expert advisor and offer to answer as many as three yes/no questions at no cost.
  • The questions must be asked one at a time, perhaps by mail or e-mail. Answers will be delivered similarly.
  • If you really do have an area of expertise, you may be able to give knowledgeable answers. Otherwise, flip a coin.
  • Suppose you started with 1,024 people willing to ask you the first three questions. If you use coin flips (I know, your thumb will get tired), you will be right about half the time. If you are truly expert you may be right more than this, but we'll stick with half for now.
  • For 128 folks, you've been right three times. For 128, you've been wrong three times. You'll certainly lose those latter 128.
  • For the rest, you've been right twice for 448 and right only once for 448. You might lose all these as future, paying clients, but maybe not. Anyway, you have at least 128 people who are quite impressed with your "expertise". (Actual numbers will vary, for statistical reasons, but using exact math makes the analysis simpler.)
  • Let's assume all the 128 happy clients are willing to start paying for answers. You introduce a sliding scale, with enough flexibility that you can charge more later on, pretty much at your "discretion". But of course what you tell them is that you charge more if the question needs further "research".
  • Three more questions later, you have 16 clients for whom your score is perfect. You may have been able to keep some of the others as well, by arguing that a wrong answer here or there is to be expected.
  • But those 16 are now your bread and butter. They will pay a lot more for questions requiring more "research" (it takes you an extra day or two to flip the coin), and be more tolerant of finding you wrong something like half the time!
Of course, if you are truly an expert, and are wrong substantially less than half the time, you'll do even better than someone relying on coin flips. The danger is believing too much in your own infallibility. Successful advisors must be quite dispassionate. That is why the ones who rake in the big bucks are total psychopaths such as Bernie Madoff, who added a pyramid scheme to his coin-flipping.

Now for the phrase that pays: Technical Analysis for stock picking is a way of hiding the coin flip amidst double-talk and jargon. The fancier the computer screen on which the "analysis" is presented, the more one can charge for the "advice".

I've been in and around the US stock market for more than 50 years. I've analyzed things nine ways from Sunday. Much technical analysis assumes a normal (Gaussian) distribution of daily moves. It is easy to plot a couple years' data for any stock you like, and see that large moves are more common than the Gaussian distribution would predict. Some have claimed that the actual distribution is a Cauchy (AKA Pareto) distribution, most notably Nick Taleb in The Black Swan. This distribution is favored by the "you can't pick it" crowd because, while it has a central tendency, it has no mean and no moments, and extra-large moves are possible at any time, that can wipe out days' worth of gains in a stroke. But the actual distribution is not quite so extreme.

For those with statistical expertise, this will be a meaningful explanation: The Cauchy distribution is a two-tailed analog of the Scale-Free distribution (log-log straight line) so beloved of chaos theorists. The real distribution is the complex square root of a Lognormal distribution. It has no name yet, and I haven't thought of a good one. Don't anybody name it after me! Anyway, it has fat tails, just not as fat as the Cauchy. For this reason, it just might be very slightly predictable, but less so than if market moves actually had a Gaussian distribution.

It isn't hard to do a sequence analysis. Download daily closing prices from your site of choice (I like into an Excel sheet. Be careful to pick a time frame of at least a year, in which the first and last closing prices are very nearly the same.

Calculate the daily moves (a simple Excel formula subtracting today from yesterday and so forth), then copy (paste values) them to the next column with a 1-day shift. Plot them in a scatter plot, or calculate a correlation coefficient. I just did this with the closing prices for McDonald's, 3/4/2013 to 3/4/2014. The chart is immediately below, and the correlation coefficient between the two columns is -0.055, or -5.5%. It requires correlations greater than 50% or less than -50% for statistical methods to make you any money. That's your simple proof that technical analysis cannot work!

The first and last closing prices were 95.07 and 95.02. However, during this time, the stock returned a 3.3% dividend. So, you could beat your head against the wall trying to make money on daily trades when a stock is going nowhere, or just hold it for a year and collect the dividend.

Into this arena I find a new book, Advanced Charting Techniques for High Probability Trading, by Joseph R. Hooper, Aaron R. Zalewski and Edwin L. Watanabe. They claim to have licked the barriers to successful technical trading, AKA timing the market. I read about a quarter into the book, and gradually realized that they were long on claims ("Many clients earn 20% or more monthly") but short on meaningful specifics.

Oh, there are plenty of specifics, but they all relate to using a software product to which you subscribe for $80/month (I got this from their web site). Now, these fellows have two sets of methods. For about a decade they have promoted and taught an option technique called "covered calls and LEAPS". They claim it earns high returns in both up and down markets. I reckon the trick is finding speculators who will buy the options on terms favorable to you. Perhaps Barnum was right, that "There is a sucker born every minute", and you just need to find these suckers.

This book adds loosely-described charting techniques that are supposed to enhance the method. I am not sure if using them means you pay more for your subscription. I couldn't find out without revealing more about myself than I was willing to. However, given my deep suspicion of all charting techniques, no matter how fancy their names, I can't give any credence. Let them say all they want about creating millionaires; how many of their clients are non-millionaires? or even non-gainers?

It became clear as I read that this was mainly a confidence-building exercise, and that the book is an advertisement for the products. Maybe this really is the cat's meow. I have yet to be convinced. But I suspect the authors make a lot more from their products than from using their own methods.

No comments: