Jonathan Kinlay, Systematic Strategies
By Active Trader Staff
Although debate continues regarding both the immediate market impact and the longer-term implications of high-frequency trading, there’s no denying it has become a consistent source of profits on Wall Street.
Case in point: Systematic Strategies, a New York based high-frequency trading fund founded by British quant pro Jonathan Kinlay that, since its inception nearly four years ago, has been nothing if not consistent. From January 2007 through August of this year the fund has reported just four losing months, with the worst, November 2008, checking in at -3.89 percent (Figure 1). That downtick came directly on the heels of October’s 18.23-percent gain, the fund’s best month. Overall, the fund’s median monthly return through August 2010 is 2.43 percent, with a gross return since the beginning of 2007 of well over 200 percent.
Whether HFT as it exists today will still be around a year or five years from now is anyone’s guess, but for now it is a thriving business in an industry that has seen some of its traditional sources of revenue shrink over the past two years. Although he’s understandably tight-lipped regarding the specifics of his trading operation, Kinlay is straightforward about the nature of the high-frequency business and how it might evolve in the near future. It is, as he points out, a new industry, and one that is likely to undergo significant change as it matures.
Kinlay’s work in high frequency could be viewed as the natural evolution of a classic quant career trajectory: undergraduate degrees in math and computer science followed by a master’s degree in statistics, along with an MBA from the London Business School. He began his professional career in 1980s London in the foreign exchange area, working in risk and trading models for banks before moving on to help develop quantitative strategies for a European hedge fund.
In the 1990s he founded a research firm, Investment Analytics, which continues to perform consulting work for hedge funds in the area of investment research and strategy. (He also is the CEO of Algorithmic Execution, a broker-dealer specializing in solutions for high-frequency trading operations.) Kinlay also spent time in academia, teaching at Carnegie Mellon University from 1996 to 1999 when it was setting up a master’s program in computational finance.
Systematic Strategies had its origins in a hedge fund Kinlay started in 2003, Proteom Capital Management, which trades low-frequency strategies. Prior to Proteom, he had set up his first hedge fund, Caissa Capital, a volatility arbitrage fund that grew to about $400 million between 2001 and 2004. In 2003 Kinlay began operating Proteom, selling his interest in Caissa in 2004 to focus on his new fund full time. (He also had a brief stint as head of model development at Bear Stearns in 2007-2008.)
Proteom was based on gene-sequencing techniques derived from work done by his then-colleague Haftan Eckholdt, who was at one time head of neuroscience at Yeshiva University in New York.
“You’re essentially looking for patterns the same way you look for a gene sequence in a large, complex protein molecule,” Kinlay explains.
After an approximately 18-month process of “kicking the tires on [the idea] pretty hard,” Kinlay’s team came to the conclusion it was capable of “returning alpha.” The gene-sequencing techniques allowed them, in effect, to create a virtual trading reality.
“What we’ll do is literally create a huge number of algorithms — hundreds of millions of them — and then sample from that algorithm space to create what you might call a ‘virtual trading desk’ consisting of perhaps 50,000 virtual traders,” Kinlay says.
These traders, he explains, would be “self-selecting,” naturally organizing themselves into groups and strategies.
“You’d find algorithms that would do some very recognizable things: They’d be momentum traders, or mean-reversion traders, or trade tech stocks, or trade high-dividend stocks, or what have you,” Kinlay says. “Some were lower-frequency, some were higher frequency. Basically, what we had with the virtual trading desk was a replica of much of what is out there in the markets in terms of strategies.”
The process paid dividends on several levels. First, Kinlay says, it offered insight into how the players in the market might respond to different market developments — everything from macroeconomic news to stock-specific news. Also, it provided “alpha signals.”
“You could track where you were seeing evidence of increasing alpha and where you were seeing evidence of negative or decreasing alpha, and reallocate your portfolio accordingly,” he says.
This, Kinlay notes, mirrors the way many firms work: When allocating money to managers and traders, they tend to reward the successful ones with more capital and penalize the less successful ones by taking it away.
“This is kind of like that, except you’ve got a much larger trading desk — 50,000 different types of strategies — so you have a much broader area to play in, and you see many more signals,” he says.
The other key benefit of this intensive approach was it helped uncover patterns that were off the beaten path, so to speak.
“Many times the process uncovered identifiable patterns, such as momentum or mean-reversion tendencies, but other times the pattern the algorithm was trading had no clearly identifiable form — in other words, it was new,” Kinlay says. “By applying a great deal of computational power, you could identify new, recurring patterns in the underlying data.”
Kinlay, who once won the British under-21 chess championship and still plays chess when he has the time, says the game provides a good analogy.
“Back in the 1970s and 1980s chess programs were abysmal — you could beat them easily,” he says. “Now, of course, they’re stronger than the strongest human players in the world. It’s entirely possible for a computer program to look at a position and announce a forced checkmate for one side or the other within some large number of moves — 150 or 200, for example.”
Just as this level of pattern recognition is completely beyond the capability of even world champion chess players, so too are the market “micro-structure” patterns Systematic Strategies attempts to exploit beyond the reach of even the most gifted traders to identify by themselves.
“This is an example of how computers, just through raw computational power and largely based on pattern-recognition ideas, are able to identify patterns and solutions to problems that are beyond the power of the human mind to envision,” Kinlay says. “Part of the theory behind Proteom was that by applying similarly large-scale computer power we could identify new types of recurring patterns that human traders, typically using a single machine, or chart and paper (laughs), wouldn’t be able to identify.”
Kinlay began work in the high-frequency area around 2005. He started trading some of the resulting strategies within Proteom in 2007 and 2008, and made the decision at the end of 2008 to spin off Systematic Strategies as an independent high-frequency entity. The fund, which has $210 million under management, currently only trades the U.S. equity market, but Kinlay says other products are being planned. For the complete article, see the November 2010 issue of Active Trader magazine. Click here to subscribe.