Print
Close Window
From: www.itworld.com
The next Wall Street tycoon will be a computer program
June 4, 2008 —
Novelist Matthew Klein concocted a flamboyant fictional character in his book, Con Ed, who claimed that his dotcom company had the technology to accurately predict the stock market. Wall Street observer Matthewsuch things were already happening in real life; and then dotcom entrepreneur Matthew Klein founded Collective2.com and made it into reality. In a behind-the-scenes look at how today's traders operate, we see that much of the decision-making really is made by computers -- and that Wall Street will never be the same again
We always see images in the movies of frantic traders shouting and waving pieces of paper, is it really still like that?
I don't think there have been pieces of paper or frantic traders since the Reagan administration. The popular image of human beings doing most of the trading and speculating on Wall Street disappeared in reality long ago. Today the biggest volume is traded by computers and software. Almost every investment bank has a proprietary trading arm that uses algorithmic trading methods to make decisions in an automated fashion. And while there are still human beings that make trading decisions -- let's call them "discretionary traders" -- these human beings ultimately rely on computer programs to make their trades happen. They rely on sophisticated computer programs to execute their trades in a very clever way, to send them down to the electronic markets, and to disguise their intentions so that their competitors don't know what they're up to. To summarize, computers do almost everything for traders nowadays.
When did computer automated trading first start to appear?
You began to see computerized trading with the rise of the minicomputers in the early to mid 80s. But computerized trading first burst onto the public consciousness after the market meltdown of 1987. You may recall that was when the stock market plunged almost 30 percent in just a few minutes. Congress held hearings to investigate what happened, because as you know, Congress is always interested in putting blame on someone that's not Congress. And what they concluded was that computerized trading was to blame for the meltdown, because all the computers and the software engaged in the same strategy simultaneously, so that everyone kept selling and selling and selling, which made the market drop precipitously. So that was the first public evidence that computers were trading large amounts of volume. But clearly that had been going on for several years before, in a more discreet fashion.
So was it as Congress said, were computers responsible for that crash?
While that's the popular conception, I don't believe that's what actually happened. There have been academic papers analyzing the crash of 1987 and they come to a similar conclusion. If you look at the actual historical events, markets melted down simultaneously across the world during a 24-hour period. The Asian markets melted down first, hours before the American market, and there was no computerized trading in the Asian market at that time. Computerized trading was an American innovation that hadn't yet been put to use in Asia. So there's a lot of evidence that the crash had much more to do with macroeconomic factors and overpriced stock markets worldwide than computers. But computers are a nice villain because nobody really takes the fall when a computer is responsible.
But regardless of the crash, and Congress placing the blame, for better or worse on computers, is automated trading still prevalent today?
Yes, and what's really interesting about automated trading and where it's gotten to at this point is, like all technologies, it has become more democratized. What used to be something that required a team of PhDs and expensive minicomputers in air conditioned rooms now has filtered down to the masses. And now there are people who trade in their homes using automated trading strategies that are more complex than some of the strategies that the investment banks were using 27 years ago. So it's a case where technology has improved exponentially and has become exponentially less expensive, so that now even traders in their pajamas can take advantage of automated trading techniques and strategies.
So instead of traders in the pit of the NYSE frantically waving pieces of paper, there are now traders in their pajamas frantically waving at their laptops in their living rooms.
That's right. I think the most frantic thing they do is go and get themselves another cup of coffee when things aren't going well.
How has technology changed the nature of investing over the past five years?
Clearly technology has become less expensive and more widely disseminated. That answers the obvious question, but I think there's a much more interesting question that we could talk about, which is: what is the implication of having millions of people using automated algorithmic trading strategies? I think the answer is that it becomes a lot easier for people to become misguided. I'm not saying that the PhDs and the rocket scientists are always right, and that they know how to use complicated technology while the Average Joe does not. We see a lot of evidence that they can make horrendous blunders. Long Term Capital Management, which, as you might recall, was a hedge fund that made a trillion-dollar incorrect bet, is another famous example of the smartest guys in the room misjudging events and losing a lot of money based on computer models that proved to be just plain wrong. The point is that now everyone is capable of making horrendous blunders, and while not everyone has billions of dollars at risk, they have a lot of their personal wealth at risk. The benefit of automated trading obviously is that it allows you to make very quick decisions using mathematics and science. The downside of it is that very often even the people that develop these formulas and algorithms don't really understand the risks involved. In that way, risk becomes much more hidden from the trader. It's hidden behind formulas and hidden behind computer screens. So it's easy to place trades, but it's hard to understand the risks of the trades you're placing, and that could be a dangerous combination.
What gave you the idea first to create an automated trading platform?
The idea came because I noticed that algorithmic trading had become highly democratized, in the sense that everyone had access to trading algorithms and trading systems. It had suddenly become perfectly acceptable for the average Joe Trader to use these trading methods. In fact, people started marketing automated trading systems to average retail traders. But there was no third-party objective platform that monitored these systems and allowed them to be judged and evaluated. So that's what caused me to begin building Collective2.com. Collective2 is, first and foremost, an objective auditor and evaluator of the risks and rewards of algorithmic trading systems. Later it evolved to become an actual trading platform where those algorithms and formulas could be executed on traders' behalf.
How do those trading systems that are being compared generate trade signals? Is that also based on computer automation and intelligence?
Part of what's valuable about the Collective2 platform - the "C2 platform" in our marketing lingo - is that it provides a degree of anonymity to participants. So I don't know what's going on behind the scenes for any given trading algorithm. People talk about how trading algorithms are "black boxes." And that is really the case at Collective2. People can use, market, and trade their black box algorithmic systems, and it's completely opaque -- not only to the customers of the trading systems, but even to the people at Collective2. We have no idea what generates those trading signals that come flowing through our platform. Is it a very sophisticated computer program? Or is it some guy in a long floppy hat examining bird entrails? Who knows?
When you were creating the Collective2 platform, what was the greatest challenge you had in putting it together?
There are three sets of challenges in technology businesses. The obvious one, and probably the one that's most interesting to me personally, is the technical challenge. Can you put together a fault-tolerant system that allows millions of dollars of transactions to be executed in a very timely way? Then there's the business challenge, which is to make this business feasible and profitable. But really the biggest challenge is the public perception challenge. When you're inventing a new category of product and service, can you teach people what it is? Can you make people recognize the need for it? So that was the biggest hump we needed to get over at Collective2, to show people the need for a neutral ground electronic trading platform, where different algorithms could be used and judged against one another in a very fair and objective way.
Tell me about the technology behind the platform, what makes it tick?
I'm a big believer in using cheap, nearly free, off-the-shelf stuff. So, from a technical point of view, we're big proponents of what people call the LAMP stack: Linux, Apache, Mod-Perl, and MySQL. As an entrepreneur trying to build a small, scrappy startup, free is really the best price of all. Once we have the infrastructure in place, then we tend to throw more hardware behind the project so that we have redundant everything, but the very core of the software is running on an open source, inexpensive infrastructure.
Are computers really better at making investment decisions than human beings?
No, if anything, they're worse. Let me put it this way. If there were an obvious way to make money in the markets in a risk-free way, we would all be billionaires and we would all be sunning ourselves on our private Caribbean islands. I don't know about you, but not many people that I associate with have their own island. So you have to assume that speculation is hard and most people fail at it. Computers can only make you faster and more efficient at what you normally do, and as I alluded to earlier, that's part of the problem with automating the process of speculation. In many cases you just get a lot better and faster at losing your money.
There are obviously some risks involved in turning over some of these investment decisions to a computer program. What are some of those risks?
There are two sets of risks. There's the obvious kind of risk, which is: can you design your computer program to follow your instructions accurately? Because, after all, computer programs are good at following instructions. If you tell your computer program, "Shut down all trading when I lose $X dollars in one day," then presumably -- if you're a competent programmer -- your software will do exactly what you say. But there's always a risk that you will program something incorrectly, or that you will introduce a bug that causes the program to do something other than what you intended. That's the first set of risks. It's easy to scoff at those, by the way, and to say, "Well, any programmer worth his salt will avoid producing buggy software," but of course the world is filled with examples where even simple programming tasks are done incorrectly. I always think about the example of the NASA Mars probe that was lost in space because the programmers neglected to convert from metric to English measurements, or vice-versa.
So, okay, making buggy software is one kind of risk, when you rely on computers to gamble your hard-earned money. But the really interesting type of risk is in a completely different category. That's the risk of the unknown risk. You may remember Donald Rumsfeld was criticized early in the Iraq War for speaking about "known unknowns and unknown unknowns." The press had a field day with that comment, but I think computer programmers understood what he meant. There are risks that we can anticipate, and then there are risks that are completely outside our limited field of experience. That's where real risk lies, in unforeseen events, in what people call six-sigma events, which are so outside the common human experience that you just don't even know how to anticipate them. It's all well and good to tell your computer program to stop your losses at X dollars, but what if you wake up one day and the market has completely vaporized and there is no market available on which to trade? Your computer can be very quick at shutting down trading, but by that time, it's too late. Computers and automation never solve the risk of these outlier events. The problem is that people who tend to use technical trading and algorithmic trading are very confident in their ability to control risk. But I think they're overly confident. It behooves all traders to keep in mind that something really terrible that you can't imagine can happen at any time. And you have to be ready for that.
Is it possible to imagine a type of a black box trading system where you put money in, you go out and play golf, and then you come back rich?
Certainly there are black boxes where you put money in and then you go out and play golf. But what happens when you come back -- whether you're rich or whether you're poor -- depends on what goes on inside that black box. It has been the fantasy of speculators and traders from the time of the ancient Greeks to just turn on a machine and then come back and count the money. If this fantasy appeals to you, and you can stomach the risks, then algorithmic trading is for you. Everything can happen automatically. If you think about it, financial speculation is one of the few human endeavors that can be conducted without actually lifting a finger; and that, I think is one of the reasons that it's so fascinating and attractive to people. Now, the question is whether these robots or computer programs or black boxes can make money. That depends obviously on the skill of the software designers.
The specter of increasing levels of automation and predictive technology in the investment world sounds complicated and expensive, is that shutting out the ordinary middle class investor?
I think not. I think one of the great things about the financial industry is that it's a very competitive industry, and that it's a very hungry industry. It is hungry for capital, and it will take it from whoever has it. Technology is driving down the costs of automated trading and algorithmic trading, and with companies like Collective2, it is driving down the cost of assembling portfolios of algorithmic black boxes. Because costs are coming down, now it's even more efficient for smaller investors to participate in the marketplace. The overhead and the frictional costs go down, which means that even with a smaller amount of capital you can efficiently participate. So, to answer your question, I think it's exactly the opposite of what you propose. In the old days only the elites got to participate in this world, but nowadays almost anyone with a modest sized amount of risk capital can do so.
But even with a black box trading system, doesn't it still take some sort of wisdom? Or is it really something that somebody with absolutely no knowledge of the market could really do effectively?
We will always need rocket scientists and geniuses -- the guys that build the algorithms. Typically they're employed by investment banks and hedge funds. These skills are hard to find. You're always going to need very smart people to do this job. But the good news is that companies like Collective2 allow average people to take advantage of these peoples' efforts in this area. In other words, now those black boxes and algorithms can be plugged in and reassembled into portfolios for the average investor. In the old days you needed to engage the services of a hedge fund (and to pay hedge funds' outrageous fees) in order to go out there and figure out which black boxes to use and even to figure out the mechanisms for how to assemble them into a portfolio. And I think what companies like Collective2 do is to allow an ambitious average retail trader to go out and make his own choices about these black boxes. So you don't really need to know what's going on inside the box, you just need to know the box's risk profile and its expected returns, and then you can plug together a lot of these boxes and try to assemble a portfolio that works for your given risk profile.
What's to stop computers from completely taking over Wall Street?
They already have. To the extent that the actual dirty work is done by computers, computers already run Wall Street. I don't think that's a big secret. What's left for the human beings is to program the computers. One of the interesting points here is that you see an arms race effect in the field of automated trading. People develop strategies that become very effective for short periods of time, but as soon as that strategy is unleashed on the market, other market participants adapt to it and make it less efficient. It's fairly common knowledge that some of the most famous quant funds or algorithmic trading houses, like D.E. Shaw or Renaissance, develop strategies with very short half lives. Not on purpose, but that's just how it works. So to answer your question, I think there's always going to be a place for human beings to generate new strategies and new black boxes. I think ultimately human beings aren't going anywhere, because human beings are hungry and competitive and always want to do better than they did yesterday.
In five years, what will the state of technology in the investment world be like?
One of the long term trends that I feel qualified to comment on, anyway, is I think that you're beginning to see a breakdown between what used to be thought of as very discrete types of financial instruments. You're also beginning to see that retail investors are more comfortable investing outside the United States. Nowadays an investor sitting in a suburban house outside Indianapolis can invest in world markets very easily and can invest in stocks, futures, and over the counter foreign exchange. These were opportunities that were not available to people even ten years ago. I think that governmental regulation is really the only thing that keeps the opportunities from expanding even further. But the market forces driving this expansion are implacable. Even regulation won't stop it for long. As soon as the regulators figure out how to insure their own job security, they'll allow average people to trade exotic instruments and to trade in markets world-wide. People want to be able to invest all over the world and in multiple types of instruments.
You're a real Renaissance man. You've also written a novel about Silicon Valley crime and intrigue called Con Ed. Tell me about that.
I spent about ten years of my life in Silicon Valley starting high tech companies. But unfortunately I was a terrible businessperson. My companies wound up being real disasters -- disasters where investors lost money, and where people lost jobs. I'm not very proud of that period in my life -- after all, who is proud of their failures? But at least that experience led me to write a comic crime novel called Con Ed. Con Ed is about Silicon Valley during the first Internet boom. It's about a con man who pretends to start a high tech company that can predict the stock market. But of course it's really a very elaborate con to steal money from the Russian mob. I'm pleased to say that Con Ed got really nice reviews, and it has been translated into something like ten languages. I still get great fan mail from people around the world. That's probably the most thrilling part of writing a novel -- hearing from the people that like your book.
You obviously have a long personal history as a dotcom entrepreneur, does the book have a little bit of your personal observations and things you've seen, personal experiences?
Sure. Although I did deal with some very aggressive and steely venture capitalists I have to admit I never dealt with anyone in the Russian mob. But in any novel, the author's real life will inform events in the book, and I think I drew from some of the people I met, and some of the experiences I had in Silicon Valley, when I wrote Con Ed.
In closing, what advice would you have for an ordinary trader who wants to bring automation into his or her investing decisions?
The most important advice, and I can't stress this enough, is to go slow, and be very careful and very thoughtful. Assume that, despite what you think can happen, any amount of money you invest can be lost in an instant. That's why I strongly encourage people to be thoughtful and careful, much more careful than they would be if they were placing trades manually with a broker on a telephone, or on a Web site. Automation makes it so easy to place trades. But it also makes it easy to lose money. When you see a trading system or algorithm that seems too good to be true, it probably is. So just use very small amounts of money when you're starting out. And be very careful.
ITworld