Post of the Day
March 20, 2000

Board Name:
Echelon Corporation

Posts selected for this feature rarely stand alone. They are usually a part of an ongoing thread, and are out of context when presented here. The material should be read in that light.

Subject:  On weak form efficiency
Author:  howardroark

Against my better judgment, I'm going to weigh in here. Warning: Incredibly long winded diatribe below. Do not read while operating heavy machinery.

To start, I might point out any acrimony between the believers and non-believers is as absurd as that between long holders and short sellers. No uttered opinion will effect different results. If the other camp is more right than yours, their message is invaluable. If they are less right, well, then it's because of them that you get to live on the right side of Lake Wobegon.

At first glance, it seems silly to dismiss technical analysis outright. Why not use all available information to inform your decision? After all, it is intuitive that things like MACD (moving average convergence/divergence), OBV (on balance volume), various filters rules, various support and resistance iterations and so on, all might represent snapshots of the dynamic complex adaptive market psychology. Many of these rely on fairly uncontroversial group psychology tenets: people move in herds, sentiment has intermediate term momentum, information is processed unevenly, etc. And even those technical rules that aren't as immediately compelling could conceivably be shortcuts that mimic the sort of emergent properties that complexity theorists study across macro populations.

The random walker would argue, though, that merely being initially compelling or intuitively sensible doesn't get the job done. Louis Bachelier, who explained the Random Walk theory before Malkiel was born, pointed out that observable patterns would be washed quickly away in a fair, liquid market of people doing what they could to make money, leaving "[t]he mathematical expectation of the speculator at zero." In other words, any recognizable pattern that might show up in the market that could be exploited by observing past price and other public information would be quickly eliminated by investors. And Bachelier supported his assertions by showing that stock prices did not follow any obvious pattern, though of course his results weren't anywhere close to the final word.

So what? What's wrong with continuously looking for technical signals that work? One problem might come from the multiple hypothesis phenomena. Not unlike the notorious coin-flip example, if you sort through the same set of data with enough trading signals, many of them are bound to look like they work, even though they don't. A few years ago, a guy named David Leinweber went through a United Nations CD-Rom and found that the absolute best predictor of the S&P index was the production of butter in Bangladesh. And it lasted over a ten year period with stronger correlations than any technical rule I've ever seen tested. It could have made you rich. So merely finding patterns - be they people with unusual success in TA trading or actual technical strategies - can definitely be misleading.

One famous academic study from the Journal of Finance (1992) did find those strong patterns (though not quite Bangladeshi in magnitude), after analyzing 26 different TA rules. The "BLL" study found that some rules did seem to outperform over time (90 years); specifically, the 50 day moving average and a support and resistance based model. More recently, a bunch of authors (download Acrobat PDF file here (long)) checked that study for the coin-flipping problem (the possibility that the 26 rules chosen were data-mined...that they represented conscious and unconscious datamining of past traders out of the infinite possibilities), using a universe of just under 8,000 different TA rules to see if the trends were stronger than the predicted outliers in a random distribution.

It turns out they were, meaning that it looked like the successful trading rules seemed unlikely to be due to chance (note that a 5-day moving average was actually the superior indicator using 100 years of data from the DJIA). While that part is not settled, they did notice two other things. One, that using the best performing strategy from the previous ten year period would mean that you would not outperform. Second, that the original BLL strategies (50 day MA, etc.) did not work out of sample (in the ten years after the study). Other technical anomalies have looked strong or not strong depending on who you believe, including intermediate term (previous 6 month) momentum for lower decile stocks, the January effect, and various pseudo fundamental ideas like book/market and price/sales.

Assuming anyone is still reading, this is all just to say that common sense can be both compelling and deceptive. Every prediction we make about the future is based on observations of the past. The ones we can rely on produce out of sample success in the future. But even those are subject to anecdotal imperfections. Fundamental analysis, similarly scorned by most EMTers, also relies on past observations to predict future results. But in a different way than technical analysis does, because it merely assumes that there is some inefficiency in the market, and uses the history of capitalist industry to do its data-mining, rather than the history of the ticker tape.

I think that anyone using TA should at least be cognizant of a few things. First, only the winners are likely to speak out. If the Fool requested users to submit their IQs, the average wouldn't likely be a good representative sample of all of fooldom, as it would be biased with the proud and the lying. Second, there are many, many intelligent people with ridiculous resources using things like fractals, chaos theory and neural networks to pick patterns out of this monster - and of course eliminate them. Even though the market will probably never be "perfectly efficient," even in the weak form (and if it is, we'll probably never know), it's a tough field to be competing against, and transactions and tax costs can be killer, so I would at least keep a healthy skepticism about the whole thing.

http://papers.ssrn.com/paper.taf?ABSTRACT_ID=15108

Go To This Post |  This Board |  Post a Reply

Liked this post?
Read more posts by this author.

More Recommended Posts
Get past Posts of the Day in the Archives