Mr. Market Gets Lost in the Funhouse of Algorithmic Recursiveness

However, Mr. Market suffers from some rather incurable emotional problems; you see, he is very temperamental. When Mr. Market is overcome by boundless optimism or bottomless pessimism, he will quote you a price that seems to you a little short of silly.

--Benjamin Graham, The Intelligent Investor

In a volatile market week like this, it is remarkable how impressed journalists are by one day’s precipitous plunge and the next day’s soaring highs. You almost want to scream at the TV set: “What part of volatility don’t you get, you silly himbo?!” Until volatility subsides, one 500-point move up or down doesn’t mean absolutely anything.

One of the tenets of modern finance is that markets cannot be interpreted as having behaviors. Above all, you shouldn’t anthropomorphize them. Journalists are fond of saying the market “rose due to optimism regarding an economic recovery” or “swooned over fears about a slowdown in manufacturing.” It is equally dangerous to ascribe simplistic causal explanations for this or that market movement. However, 80% of financial journalism seems to be designed to commit exactly this sort of fallacy on a regular basis. “The market is concerned about this” or “the market doubts the finance minister’s commitment to that.” One almost pictures the market stroking its beard thoughtfully as it ponders the slings and arrows of outrageous fortune.

There is a famous example cited by Nassim Nicholas Taleb about a Bloomberg headline for the day in which Saddam Hussein was captured: 

One day in December 2003, when Saddam Hussein was captured, Bloomberg News flashed the following headline at 13:01: U.S. TREASURIES RISE; HUSSEIN CAPTURE MAY NOT CURB TERRORISM. 

Whenever there is a market move, the news media feel obligated to give the “reason.” Half an hour later, they had to issue a new headline. As these U.S. Treasury bonds fell in price (they fluctuate all day long, so there was nothing special about that), Bloomberg News had a new reason for the fall: Saddam’s capture (the same Saddam). At 13:31 they issued the next bulletin: U.S. TREASURIES FALL; HUSSEIN CAPTURE BOOSTS ALLURE OF RISK ASSETS. 

So it was the same capture (the cause) explaining one event and its exact opposite. Clearly, this can’t be; these two facts cannot be linked. (The Black Swan: The Impact of the Highly Improbable, p. 74)

This type of totally backward causal explanation is even more dubious in an era when approximately seven out of every ten trades are transacted by computers. That’s right, beeping and tooting machines buy and sell the stocks that are going to pay for your retirement. Sounds really reassuring, right? We might as well have a lottery. When you turn sixty-five, you turn over a poker card and, bingo, the computer says you get to stop working and you can devote your time to playing with your grandchildren. Turn over another card and, sorry, direct to the soylent green factory for you, gramps. This unease is especially acute in the face of events like the flash crash, when the stock market went insane for about an hour or so: the DJIA lost 700 points in about five minutes and Apple shares at one point traded for $100,000. An investigation later pointed the finger at one massive trade introduced using these sophisticated algorithms.

John Maynard Keynes famously compared the stock market to a contest in a newspaper with dozens of portraits of human faces. The exercise consists in selecting the six prettiest faces. However, a prize will be awarded to the person who selects the most popular faces. Immediately you have a very complex, recursive exercise. You won’t select the faces that you think are the prettiest, but rather the six faces that you suspect most people will consider most attractive. With that simple step, you enter a level of second-order complexity and recursiveness that can loop and loop forever. Because you can take it to the nth level and never stop if you start strategizing. For example, you can suspect that most people will also select not their favorite faces but the faces they think most people will select. So then you take it to the next level and you select the six faces that most people will think other people will consider attractive. That is already a third or fourth order calculation. And so on and so on.

Keynes’s suggestion was that the stock market works in the same way. Generally, investors don’t buy the stocks they think are the best. They buy the stocks that they think others will find attractive. And often they just buy stocks that they think others will view as popular in the eyes of the generalized mass of investors. The question of whether the company in question is good or bad sort of tends to get lost in the maze of first-to-second-to-third order recursiveness.

However, that sort of complexity has a limit for a human brain. A lot of comedy is based on this difficulty. Plot lines get more and more complex until the characters forget who is fooling whom. Computers, on the contrary, can play that game for as many levels as you program them. And they seem to be doing exactly that at the moment, with a vengeance.

The result is a universe of competing lines of code, each of them trying to outsmart and one-up the other. “We often discuss it in terms of The Hunt for Red October, like submarine warfare,” says Dan Mathisson, head of Advanced Execution Services at Credit Suisse. “There are predatory traders out there that are constantly probing in the dark, trying to detect the presence of a big submarine coming through. And the job of the algorithmic trader is to make that submarine as stealth as possible.” 

Meanwhile, these algorithms tend to see the market from a machine’s point of view, which can be very different from a human’s. Rather than focus on the behavior of individual stocks, for instance, many prop-trading algorithms look at the market as a vast weather system, with trends and movements that can be predicted and capitalized upon. These patterns may not be visible to humans, but computers, with their ability to analyze massive amounts of data at lightning speed, can sense them. (Algorithms Take Control of Wall Street)

As the financial world becomes more and more dependent on this sort of program, it is also becoming more fearful of the flaws of these algorithms. It is a recurrent trope of the past thirty years. In the 1980s, computerized “portfolio insurance” led to the 1987 crash. When signals of declines were triggered, these programs were trained to sell stocks in order to minimize losses. The problem was that everyone was using the same program and everybody sold at the same time. This in turn fed a more and more extreme decline, triggering stronger and stronger sell signals until you had a disastrous one-day crash. The quantitative approach is also behind the chaos of Long Term Capital Management. This hedge fund’s models famously couldn’t conceive that the spread between essentially equal securities could widen beyond a certain point. But given a corrupt Russian government, a half-drunk president and a chaotic default that is exactly what can happen: what the models say is impossible.

So tonight when you turn on the TV set and hear that “it was a wild ride on Wall Street today, as stocks first rose and then fell on fears” of this or that, allow yourself a little chuckle.

That is, unless you are invested in this market.

Miguel Llorens is a freelance financial translator based in Madrid who works from Spanish into English. He is specialized in equity research, economics, accounting, and investment strategy. He has worked as a translator for Goldman Sachs, the US Government's Open Source Center, several small-and-medium-sized brokerages, asset management institutions based in Spain, and H.B.O. International. To contact him, visit his website and write to the address listed there. You can also join his LinkedIn network or follow him on Twitter.