PsyFi Search

Tuesday 1 March 2016

The Chart Illusion

Timing Dragons

Although, logically, I’ve always felt that the idea of investing by charts should be something on the map between “Dragons” and “Free Lunches” I’ve never been so sure of this as to be outright skeptical. And I know a few smart people who insist that they are at least helpful in timing investing decisions.

Of course, the idea of “timing” anything in investment is fairly ludicrous anyway, but some recent research suggests the whole charting concept does actually help in prediction. Unfortunately what it predicts is a bunch of irrational self-fulfilling behaviour.


Logically chart based investing can’t work. The argument can be run one of two ways. Firstly, let’s assume that charts do provide valuable investing information, ignoring all those tricky questions about what this information might be. Well, charts are freely available so if there is some informational advantage associated with them you’d expect more and more people to pick up on this and, inevitably, it becomes a crowded trade.

The second point is that, in a market dominated by hugely well funded corporations the idea that they’ll have missed something as obvious as chart based signalling is too ludicrous to be even worth considering. If such signals exist the gigantic computing engines of the securities industry will be trading them to death within nanoseconds of their birth.


And yet, and yet ... There’s a faint chance that there might actually be some people who can discern information in charts that even supercomputers can’t extract. The human brain is an amazing pattern recognition engine that, as yet, not even the best IT can outperform (see: Google Charts You Can Trust and Technical Analysis on Display).

For example, weather forecasting has improved out of all recognition in the past thirty years. It may not feel like it when you’re caught in a freak sunny spell in London in the middle of August, but the gradual introduction of supercomputers and better and better algorithms have significantly improved performance. Yet weather forecasts are still finally analysed by human beings because despite all of the improvements in the technology a good pattern recognizing human can improve the final forecast accuracy for rainfall by about 25% (well, according to Nate Silver in The Signal and the Noise). Which is pretty amazing, to be honest.


Now the people who can do this are not your average mortals. They’ve very well trained and they’re selected from a much bigger pool of people because they’re simply better at detecting the anomalies in weather patterns than others. But the point is that there are a few people who are better at this than you might expect – which at least keeps the door open to the possibility that while most stock chartists are the modern equivalent of a medieval alchemist, there might be the odd real chemist in their midst.

Even if this was the case, though, it would still suggest that for most people using charts for investment purposes is simply a very complicated way of generating random numbers. Which leaves open the issue of why, if it doesn’t work, do so many people keep on doing it? The answer, of course, is psychology.


In signalling parlance markets are very noisy places – it’s difficult to detect the informational signal because of all the noise. In fact Fischer Black long ago predicted that a lot of people would mistake the noise for the signal and start trading on it and, indeed, it seems that there are a lot of noise traders in the marketplace (see: Idiot Noise Traders). It’s a form of illusory pattern matching, seeing patterns that simply don’t exist.

Jennifer Bender, C.L. Osler and David Simon wondered whether the obstinate refusal of charting to die a decent death might be because adherents were confusing noise and signal. In Noise Trading and Illusory Correlations in U.S. Equity Markets they looked at whether a head-and-shoulders chart pattern was actually predictive of anything at all.

More Noise

A head-and-shoulders pattern is probably the most commonly used chart – it’s simply a sequence of three price peaks, with the middle one as the highest. It’s supposed to predict the start of a downtrend in the stock or market. However, as the paper indicates:
“These tests consistently indicate that head-and-shoulders patterns do not profitably predict directional price moves in U.S. equities.”
In fact what they do find is that a head-and-shoulders pattern is associated with increased trading and reduced bid-ask spreads:
“Trading volume is over 60 percent higher than normal when traders would normally enter positions based on head-and-shoulders patterns. We next show that spreads narrow contemporaneously with this excess trading. The narrowing amounts to nine percent of average spreads”
This looks an awful lot like noise trading. Essentially there isn’t any real information about stock or market movements associated with the pattern – but there is real information about what traders will do when faced with it. They’ll start trading among themselves and moving the stock or the market. And, as markets are self-fulfilling prophesies, they may even create the market movements they’re expecting.


Bender, Osler and Simon speculate that the underlying causes of this behavior are a few of our old favorite biases: overconfidence, memory bias and wishful thinking, mixed in with a dose of illusory pattern recognition. In the end people are convincing themselves that they’re right and ignoring examples that indicate that they’re often wrong.

Now, if this trading is moving stocks with no real underlying information it stands to reason that sometimes those movements are just plain stupid, so in fact the best way of trading on chart patterns might be to look for them and then bet against the movements that make no sense in fundamental terms. It’s simply weighting the odds in your favor. Although I have no doubt that there are people out there doing that already.

It’s All About You

The idea that charts can be self-fulfilling and that adherents are simply trading against noise in the system is compelling. But, of course, because people are predictably noise trading there is some short-term information being generated – but it’s not about stocks or markets, it’s about traders.

I certainly don’t dismiss the possibility that there may be a few people out there who can discern some signal in the noise, but there’s no way that any of the numerous well-known chart based systems are going to work for most of us. After all, if they did everyone would be using them. And then, of course, they’d stop working.

1 comment:

  1. Sorry but the main argument about everybody using it is the same for fundamental analysis. Go and have a look at the variance of company valuations (very small), which means everybody does quite the same, and then check all the papers that show how terrible are these predictions. And finally have a look at how active investing is doing against passive strategies. At the end you are going to conclude the same for any financial tool. Very few people know how to use it and they are special.