PsyFi Search

Wednesday 8 December 2010

Technical Analysis on Display

Patterns in the Stars

The idea that technical analysis of stocks – the extraction of useful trading data from various charts mapping share prices and other data – is actually anything more than a hangover from pre-computer days is generally scoffed at by those who spend their time number crunching. From these lofty heights chartists are generally regarded with the same distain as scientists reserve for astrologers.

Which may be justified, or not, but there’s one area where the human brain can still outperform microprocessors: the unconscious extraction of patterns from visual information. Is it, in fact, possible that technical analysts are simply using the brain’s unique pattern recognition capabilities to outwit the supercomputers?

Picture Books and Representation

Chartists generally have a bad reputation amongst those people who think of themselves as number crunching analysts. After all, how can you really extract the true value of a corporation – or even its future price movements – from a pictorial representation of data? It’s like suggesting that a child’s picture book is equivalent to War and Peace.

However, it isn’t entirely clear why this is so. After all, if a chart is simply representing the same data as used by the fundamental analysts in a different way why should it not contain equally valid information? Moreover, if it does, and computers are poor at such analysis then maybe computer based analysis of the same data can’t – literally – see the information.

This idea of representation mattering to how we interpret data has a long history, but often surprises people who, living in the present, can conceive of no other ways of envisioning the world. Yet it’s crucial to the understanding of the universe we now think we possess.

As we saw In The Beginning Were The Accountants the development of our system of number representation was critical to advances in finance as well as science and mathematics, engineering and live poker websites. Along the way came the need to invent zero and that, eventually, led to joint-stock corporations, quantum mechanics, Bernie Madoff and very, very cold temperatures.

Roman Representation

To see why this mattered consider the use of the Roman numerals. I is the same as 1 and X the same as 10, so why should these different representations matter? Well, can you imagine trying to do even the simplest mathematical operation in Roman numerals? Try XXIII times VII.

Got it?

If you bothered to figure out the answer you converted each number into a decimal representation, performed the calculation, and then converted back again. Sadly for them, Romans didn’t have that option – although given they were a bit busy conquering the known world and inventing the bikini they didn’t regard this as a huge loss. Even if they’d wanted to, however, the Romans would have struggled to make any advances in mathematics or science as we know it. The Arabic breakthrough of representing numbers the way that we commonly know them was critical to the development of our knowledge of the way the world works.

Presentation's Important

Now if you think about it, that’s a pretty odd state of affairs. It’s not enough to have numbers, you also need the form of the numbers to be of a certain type. If you get the representation wrong then you can’t understand the universe – at least not the way we do. Of course, this leaves us with the sneaking suspicion that there may be yet more ways of representing numbers that would allow us to understand things differently. After all, if 1, 2, 3 is better than I, II, III then maybe something’s better than 1, 2, 3.

So we can probably agree that the way we represent data is critical to the way we can analyse it and the way we end up understanding it. This idea has been a long-term preoccupation of Edward R. Tufte, who has long been concerned with information design and how badly presented information can simply overwhelm the message with superfluous data. He’s particularly scathing about the way that PowerPoint is used, often inappropriately, and some of his best work is reserved for NASA:
“How is it that elaborate architecture of thought always fits exactly on one slide? The rigid slide-by-slide hierarchies, indifferent to content, slice and dice the evidence into arbitrary compartments, producing an anti-narrative with choppy continuity. Medieval in its preoccupation with hierarchical distinctions, the PowerPoint format signals every bullet’s status in 4 or 5 different simultaneous ways.”
Effective Display

If this was some school project this might be disappointing, but in fact it’s a review of the risk to the Columbia space shuttle caused by damage on take-off undertaken while the Columbia was still orbiting the Earth. The result of this analysis was that NASA decided no further investigation was needed. The Columbia was destroyed on re-entry.

Tufte’s main point is that it’s not just enough to present data, to make best use of it you need to present it effectively. Effective presenting means making use of various chart based methods of representing data and it certainly seems that, done properly, our visual processing capabilities are much enhanced.

The evidence that we’re better at processing a visual representation of information is, widespread. As Dansereau and Simpson explain, in A Picture is Worth a Thousand Words: The Case for Graphic Representations:
“It appears that humans are hardwired with neural processes for organizing visual displays according to gestalt principles such as similarity, proximity, continuation, symmetry, closure, and common boundary. These processes allow efficient recognition of important complex patterns that are difficult to comprehend via spoken and written language”.
Extracting Data from the Noise

Perhaps unsurprisingly given that visual information is easily – if not always accurately – processed while language and number representations are mediated by learning it seems that we have some ability to analyse financial information presented in chart form. In Is It Real or Is It Randomised?: A Financial Turing Test the researchers showed that with some training participants were able to distinguish between charts of genuine market returns and those which were simply made up.

Now the general view is that charts of stock prices contain no useful information, no matter what chart analysts suggest. These charts represent information about stocks in a very specific way and the suspicion remains that there is information encoded in them that can be detected by human pattern recognition. Indeed, as Andrew Lo, Harry Mamaysky and Jiang Wang suggest:
“Technical analysis is primarily visual, whereas quantitative finance is primarily algebraic and numerical … technical analysis has survived through the years, perhaps because its visual mode of analysis is more conducive to human cognition, and perhaps because pattern recognition is one of the few repetitive activities for which computers do not have an absolute advantage (yet)”.
Testing Visual Analysis

The argument here is subtle – that the inability of computers to perform visual pattern recognition to the same level as humans means that it is – theoretically at least – possible for people to extract useful information from charts where computer based analysis fails. As the researchers point out there is some evidence that past prices can be used to predict future ones and the researchers argue that some people may be able to extract this information from a visual representation of historical prices.

To test this they created a computer simulation of a technical analyst – pointing out along the way that the capability to do this has only recently become widely available – and set about seeing if they could detect some of the more common technical patterns. What they found is vaguely positive, if equivocal:
“We find that certain patterns, when applied to many stocks over many time periods, do provide incremental information, especially for Nasdaq stocks. Although this does not necessarily imply that technical analysis can be used to generate “excess” trading profits, it does raise the possibility that technical analysis can add value to the investment process”.
Visualising the Prospects

Which amounts to a not proven, but not entirely unproven, verdict. In fact if you take the view that quantitative analysis can identify stocks that will fly or die then assuming you could effectively represent the same information visually then there’s no obvious reason that some version of technical analysis couldn’t yield the same results. Although the impossibility of identifying a decent management or resounding competitive advantage from either process is probably more the point.

Still, if there’s merit in the idea it would seem to make sense that the way that information is represented visually would make a difference to the ability of people to process it. As it happens Tufte's followers are already onto that one: and even if you don’t agree with his ideas his books are, unsurprisingly, beautiful and a joy to read.


Related articles: Technical Analysis, Killed By Popularity, Momentum Trading Madness, What's The Shape Of Your Recession?

6 comments:

  1. Sir,
    Just one comment. The Arabic numerals are generally credited to the Indians; from whom the Arabs seem to have borrowed the original series, along with the 'Zero'.

    Thanks,
    Mike

    ReplyDelete
  2. When a large group of people all react in a predictable manner there's an opportunity to undercut some of them.

    Fibonacci sequence, Harami Cross or whatever, as more people use it the better it works. A self fulfilling prophecy.

    ReplyDelete
  3. "If you tell a lie too often it becomes the truth".

    And so it seems could have happened with Arabs taking credit for zero (amongst other things)!

    ReplyDelete
  4. The Arabic numerals are generally credited to the Indians; from whom the Arabs seem to have borrowed the original series, along with the 'Zero'.

    I previously covered this in the companion piece to this article: In the Beginning Were the Accountants.

    And so it seems could have happened with Arabs taking credit for zero (amongst other things)!

    The problem was more to do with the Europeans who credited the Arabs from whom they received the new number system with the discovery, rather than Arabic mathematicians themselves. In fact, the trajectory of Islamic invention is fascinating: while Western European civilisation was clinging on by its fingertips Islamic scholars were making discoveries in mathematics, architecture, science and even economics. It seems to have been the gradual ossification of religious practices which closed off this flowering of knowledge, but in many ways the Arabian states were a lifeboat, preserving Greek knowledge for a millenium while Europe suffered the Dark Ages.

    ReplyDelete
  5. I don't personally believe that technical analysis works but I try not to be dogmatic on this point. The idea certainly makes sense. TA is rooted in the idea that human emotions play out in at least somewhat predictable patterns. If that's so (I have grave doubts but don't feel that I know for sure), why wouldn't this work?

    The people who make fun of TA are people who don't like the reality that emotions affect stock prices. They want to believe that stock prices are determined by hard economic realities. The very fact that TA bothers them so much shows that they don't have confidence in their own beliefs.

    Rob

    ReplyDelete
  6. Based on my several years of experience and speaking of stock indexes, TA is more or less predictive depending on your time frame, market context, news flow, and how important a level you are at (50 day MA, 200 day MA more important). In other words, your mileage WILL vary.

    What it is very useful for for a shorter term trader is seeing where price is relative to where it's been and making educated guesses whether it will continue a trend or looks like it has stalled and is headed in the opposite direction.

    Not very revelatory I know .... :)

    ReplyDelete