Being a student of Mathematics, the topic of Black-Box Quant Models has always intrigued me. I do not have the programming acumen to develop the models myself but would love to see what goes into them. I admit that they are black-box for a reason, and I'm sure very sophisticated and complex. That being said, I still am somewhat skeptical of whether they are simply self-fulfilling in nature and quite possibly dangerous.
Before writing this story, I tried to scan the web for articles on this topic and some ideas that I had about it. One such article was posted on Information Arbitrage back in November. It looks at whether these models and trading systems are a revolution or a marketing gimmick. He writes:
"Maybe, but I'm cynical. No knock on Ray (he is clearly one of the most brilliant thinkers of our generation), but I think there are enough brilliant minds working in enough related areas with enough access to capital to make any demonstrable advantage fleeting at best. Call me a cynic, but after 20 years kicking around the markets and with a sense of history it takes a lot more than a few good years to convince me that a new paradigm is upon us."
I could not agree with him more. I mentioned earlier that I was a student of Math. Let me provide some background to illustrate where I am coming from. During my stint at UNC-Chapel Hill, I had the pleasure of enrolling in Mathematical Decision Sciences (MDS) Program. This program was put together a year or two before and was more or less a hodgepodge of Calculus, Statistics, Probability, and Operations Research. It wasn't really weaved together as a program yet, but I had no problem with that as I prefer to fill in the blanks myself. After being a top Math Student in high school, I had a lot of trouble in upper level Calculus, because of my lack of interest in Physics (or possibly my enormous interest in Busch Light and Carolina girls but correlation does not imply causation. Right?). Anyways, finance-minded as I was, I had a tough time applying it to many of my interests in that area because I have always thought of Finance and Economics as Social Sciences, and incapable of being explained by Applied Mathematics. My thoughts were if there was some jerk-off sitting at the other end of the funnel losing money because of the rate at which I calculated fluids were flowing, he would stick his thumb over the other end and change the rate. Although this guy would possibly lose some money in the short run, he would eventually realize that he could change the size of the hole thereby making himself profitable. It is this consciousness involved in social sciences which creates the bubbles and busts that George Soros writes about in the Alchemy of Finance. This is at the heart of the Theory of Reflexivity, a dominant force in his work.
I do not believe at this time we are sophisticated enough to develop micro-level models that deconstruct the market enough to continuously capture alpha. The only physics equation that explains their success, in my humble opinion (as I am very aware of my ignorance) is Newton's Laws of motion, F=ma. Momentum is dictated by Mass and Velocity. These hedge funds are massive – with great amounts of assets and access to leverage. They also have enormous velocity as they go after arbitrage opportunities quickly, before they disappear. Therefore they exert great price momentum and therefore validate their models. The reason why they have been successful is at these levels is that success grows much like an asset. If a fund experiences this effect and profits it brings a few times, it becomes ever more massive and maintains its expected profit – more or less.
Instead, I believe these models are more of a justification for fees than anything else. If I am a millionaire, I would feel much safer in the knowledge that I had a Nobel-winning Nuclear Physicist from MIT with a cutting edge neural-network model managing my money, rather than a 24 year old blogger with an Excel spreadsheet and a library card. Not only that but I'm willing to pay that fund 2 and 20 or more for it. Then when the fund blows up because of some "Fat-Tail" event attributable to a scientific anomaly (yeah, almost as anomalous as the last one) they can use good 'ol Dr. Feelgood and his years of research to explain just how rare an event this was. They may lose the money but maintain their stature. Medical doctors may get sued if a patient has a freak reaction to some treatment and dies, but they usually stay in business afterwards and command just as much respect among their peers and perhaps some added sympathy for having that terrible "luck".
Assuming the models are useful and alpha generating – then how dangerous are they? Drawing further from Ehrenberg:
"Without question, quantitative trading approaches - carrying names such as "black box trading," "algorithmic trading" and "statistical arbitrage" - are all the rage. Lumped in with these mysterious-sounding approaches are high-IQ terms like "pattern matching," "genetic algorithms" and "neural networks." At the essence of these strategies are two distinct features: (1) humans aren't involved in the decision-making process; and (2) models are designed to either "learn" like humans or to detect non-intuitive relationships among a sea of data that can't be readily seen by humans. Basically, creating models and approaches that are, ultimately, better than humans because they can act faster, trade more cheaply, make decisions dispassionately, process more information and see things humans simply can't due to the limits of our ceberal cortex….
The result was nonlinear decision making processes more akin to how a brain operates. So-called "neural networks" and "genetic algorithms" have become common in higher-level computer science. Neural networks permit computers to create new rules and automatically change underlying assumptions by experimenting with thousands of random sequences and processes. Genetic algorithms encourage software to "evolve" by letting different rules compete, and combining the most successful outcomes.
Wall Street has rushed to mimic the techniques. Because arbitrage opportunities disappear so quickly now, neural networks have emerged that can consider thousands of scenarios at once. It is unlikely, for instance, that Microsoft will begin selling ice-cream or I.B.M. will declare bankruptcy, but a nonlinear system can consider such possibilities, and thousands of others, without overtaxing computers that must be ready to react in milliseconds."
Two comments on this: First, error and bias is automatically introduced into these models from the start. Neural Networks are great and a major area for growth in market research for the years to come, but putting them to work assumes that we know a great deal about well – brains. I contend that we know very little. Pfizer and Eli Lilley are still making boat loads off of selling crazy pills and my ex-girlfriend is still nuts no matter how many times Oprah tries to set her straight. The fact is it is difficult for us to understand the brain because we are the brain. Second, he mentions that the models are "non-linear" in nature. If you know anything about linear regression analysis is that bias in your model can be very hard to detect and eliminate. When taken to a non-linear level, with different orders and inflection points, bias is even harder to detect and errors may been greater due to magnified due to exponential nonlinearities. This produces the ultimate Neural Network – a computer model that is somewhat biased, prone to error – all along learning the wrong things. Last time I checked this could be construed as a human – so congratulations, millions of dollars were invested to simulate the guy in the E-Trade commercial.
Finally, I want to explore the idea of black box investing, artificial intelligence, and our financial system. Let us assume that these models actually can "outsmart" the market. We all know what a tough time Kasparov had with Deep Blue and he is a chess genius. By handing over power and money to machines left unchecked we may be setting ourselves up for an ugly scenario. What if they are THAT good? What if they can learn? A mild result would be a lack of interest in the financial market system from the ordinary investor. If expected returns go negative – much like a casino, you will see much less investment and much more hording of money. There will always be the gamblers but those with the best machine will always win – I believe they are called the house at the Borgata. The markets will then be decided by who has the best model and last time I checked $9.95 a trade doesn't cover sophisticated Neural Network Models. Worse, what is stopping the self learning models from collaborating and colluding - accumulating wealth at will and tearing down the seams of our financial system? Now I understand that this is on the verge of Science Fiction and I write it mostly in jest, but the last time I checked James Simons earned almost $1.7 billion last year (more than double the $670 million in 2004) and the price of my bagel and coffee in the morning went up from $1.50 to $2.75. Since they are unchecked, practically deified for the money they bring in – all along programmed to simply profit, not sit back and reason or have a conscious, this is not out of the realm of possibility.
I am not advocating for regulation in fact I think that would be terrible. I am simply pointing out some of the potential problems I see with great sums of money relying on these opaque models. Perhaps they are attributable to some of the "peculiarities" that have occurred in the markets the past few years. But, if there really is a "Rise of the Machines", I'm not all that worried because last time I checked - the Governator is still alive.
β
Note: Once again, as I have mentioned before, I did not create this site to be right or even rational (Although we all want to be both). I happen to believe that I am quite the opposite most times on both counts. I write this to introduce new ideas that I believe are not being examined in order to create a more comprehensive financial discussion. I welcome any comments and criticism with open arms as they make my site much more valuable.
Interesting comments on your blog so far. My own background as an undergrad is similar to yours, but I am a bit older :)
I too am interested in the correlation between media dynamics and the overall market. I think there is a lot of information regarding what you label as Active and Passive consumers that is both opaque and transparent. I think this is driving a revolution in what sorts of information is available to use, how we use and consume it, and what it means in the market place of finance, of citizens, and of nations.
Posted by: Barry Caplan | May 03, 2007 at 06:25 PM