Black box trading : excellent article

Sous le titre "Black Box Trading: Panacea or Promotion?", un excellent article de Roger qui ne va pas réconciler les discrétionnaires et les systématiques...:))

"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.

An interesting article in Saturday's New York Times raises an array of interesting issues and uses a new hedge fund started by one of the brainiacs of all-time, Ray Kurzweil, as the vehicle for exploring this fascinating topic. For context, it should be noted that rocket-scientist types running hedge funds is not new: figures such as David Shaw (DE Shaw) and Jim Simons (Renaissance Technologies) have been using higher-order math and computer science to extract value from market data for the last three decades. These skills are now being more broadly applied to news feeds, government filings and other data pools where entities can be extracted, sentiment gleaned and metadata created and analyzed. Other hedge funds as well as both buy- and sell-side firms are using similar technologies and approches in their businesses.

The computerization of trading and investment is a logical and noble pursuit. However, attempts such as these are not without their pitfalls for a variety of reasons. Statistical arbitrage strategies have become progressively less and less attractive as more capital has flowed into the area. Where a super-smart quantitative manager could once design a high frequency, quasi-market making strategy that was both very profitable, required little capital and entailed a small degree of market risk, they now need to extend signal horizons and seek to generate returns by doing what everybody else does when they reach for return - take on more risk and accept greater variation in returns. Further, these high frequency strategies are often not very scalable, a real hinderance for a manager that wants to grow and leverage their brand into a multi-billion dollar operation. Returns in the various statistical arbitrage strategies display asymptotic profiles, where early alpha generation is eventually squeezed to zero as more brains and assets focus on the strategy in question. Managers innovate, enjoy attractive returns for a period after which they need to move on and develop the next set of algorithms. It may appear to a layperson that a black box trading strategy would be great - few PMs with huge egos, relatively modest investment in programmers and hardware to build a scalable platform and a nimble, easy-to-adjust model to adapt to changing market conditions. This, my friends, is simply not the case.

Consider the two black box managers with the most successful long-term records and asset growth - the aforementioned DE Shaw and Renaissance. They both have armies of PhD.s of all stripes - computer scientists, mathematicians, physicists, biologists, chemists, linguists, etc. This is not exactly the cheap and scalable infrastructure many have in mind. It takes a lot of money, relentless and effective recruiting and a culture to support the degree of innovation required to succeed. I think about it as the "cycle of the 4 M's:"

1. Man, who develops the
2. Model, which is operated by the
3. Machine, which executes the Model in the
4. Market, which generates returns, results in feedback interpreted by Man, who modifies the Model, etc.

It is usually not the cute, campy story of a smart technician with his trusty computer building a successful and scalable hedge fund. Few have done it well, and it remains to be seen whether Ray Kurzweil and his lot will be able to make it into the pantheon of black box gurus like David and Jim. Do these new entrants have brains? Yes, and often in spades. But success ultimately requires A LOT more than brains, like:

1. Managerial skill
2. Risk management skill
3. Recruiting skill, and
4. Business-building skill, to name a few.

So let's turn to the NYT story for their take on things. Some interesting excerpts from the NYT story are as follows.

But in recent years, as algorithms and traditional quantitative techniques have multiplied, their successes have slowed.

“Now it’s an arms race,” said Andrew Lo, director of the Massachusetts Institute of Technology’s Laboratory for Financial Engineering. “Everyone is building more sophisticated algorithms, and the more competition exists, the smaller the profits.”

So investment firms have increasingly begun exploring mathematics’ furthest edges and turning to people like Mr. Kurzweil, who became an expert in pattern recognition building a reading machine for the blind.

********************

So Mr. Kurzweil and others took a different tack: instead of creating sequential rules to instruct a computer to read, they thought, why not create thousands of random rules and let the computer figure out what works?

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.

********************

“The downside with these systems is their black box-ness,” Mr. Williams said. “Traders have intuitive senses of how the world works. But with these systems you pour in a bunch of numbers, and something comes out the other end, and it’s not always intuitive or clear why the black box latched onto certain data or relationships.”

********************

“Right now, everyone basically has access to the same data,” said John Bates, a Progress Software executive. “To get an edge, we want to give investors the ability to immediately turn news into numbers. We want to automate what before required human analysis.”

But as these new techniques proliferate, some worry that promotion is outpacing reality. These techniques may be better for marketing than stock picking.

“Investment firms fall over themselves advertising their latest, most esoteric systems,” said Mr. Lo of M.I.T., who was asked by a $20 billion pension fund to design a neural network. He declined after discovering the investors had no real idea how such networks work.

“There are some pretty substantial misconceptions about what these things can and cannot do,” he said. “As with any black box, if you don’t know why it works, you won’t realize when it’s stopped working. Even a broken watch is right twice a day.”

So what are some of the key themes? There is an

1. Arms race, being led by the development of
2. Better models, though
3. Machines lack human intuitition, but can benefit from
4. Digitization of data, which reduces the need for humans, but
5. Is the hype around black box strategies outstripping the reality?

The arms race has been going on for decades. This is nothing new. It is simply the nature of the arms race that has changed. The last leg of the race was largely played with hardware and platforms, with FTP, execution costs being driven towards zero, competitive, low-latency platforms, real-time architectures fueled by incomprehensible processing power and "smart" trade execution systems. But the arms race is changing, with the next leg being driven by highly intelligent software and models. Programs that can take in feeds across different formats, analyze (and possibly create) the metadata at lightning speed, look for statistical and linguistic relationships among elements in the data set, and "learn" from history through enhanced algorithms leading to better performance. Ok, I get it. But it still doesn't answer the question of whether or not these new entrants will have the stuff to generate consistent, sustainable performance across a progressively larger asset base without killing returns and/or blowing up.

I wonder if this new-found emphasis on black box trading will, over time, drive alpha back towards the fundamental bottoms-up strategies. And I'm really not sure if neural networks, genetic algorithms and other ultra high-IQ approaches really change the calculus of how the markets and investor behavior works. Capital tends to flow from the "cold" strategies to the "hot" strategies, which naturally causes hot strategies to become cold and vice versa. Anyone remember convertible arbitrage? What was a darling in 2000-03 was a dog in 2004-05 and a darling once again in 2006. This is an inexorable game of "asset allocation tennis" that has taken place (and likely will continue) for time immemorial. So does it really matter if ever-more sophisticated tools and techniques are used? Or is Ray Kurzweil's knowledge base and its application to the markets so differentiated that he will enjoy a competitive advantage for a material amount of time that would enable him to build a true hedge fund firm with a lasting legacy? 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. Only time will tell.