Katonah Bucks Trend of AI Underperformance
Katonah Capital appears to be one of the few artificial-intelligence fund operations to profit from last month’s volatility spike.
The $220 million Katonah Capital Partners fund gained 3.5% in February. That compares to a 3.3% decline for the Eurekahedge AI Hedge Fund Index, with nine of 15 component funds reporting as of Feb. 20. February was shaping up to be the worst month in the history of the index, which dates to January 2011.
Quantitative hedge funds as a whole performed poorly in February, with the HFRI Equity Hedge Quantitative Directional Index slipping 2.3% and the HFRI Macro Systematic Diversified Index dropping 6.25%. Those relying largely on artificial intelligence — still a tiny subset of the quant universe — could have been expected to outperform their peers, since the technology is designed to respond to changing market dynamics in real time and automatically implement fresh strategies.
Sources said a big problem with artificial intelligence, or machine-learning, funds isn’t the algorithms they employ but the data they process. In most cases, they have been executing trades based on market trends since the 2008 financial crisis — and thus hadn’t before been tested on stressed market environments. Financial data from earlier periods generally isn’t as detailed, the sources said.
“In the end, data sets are limited in many artificial-intelligence approaches,” said an executive at a quant-fund operation. “If you look at the very large managers, they have artificial-intelligence departments and they add [artificial intelligence] to their core programs. But they are still relatively small in comparison to their overall exposure in classical programs such as trend-following.”
As for Katonah, the firm attributed its February results in part to a program that takes a dynamic, forward-looking approach to setting the fund’s leverage levels, rather than relying on “backward-looking” estimates. As a result, the fund wasn’t overly exposed when the stock market turned suddenly volatile in early February amid heightened fears about inflation.
“Although the market pundits will tell you that [the volatility] was unprecedented, our view is that herd behavior has fattened the tails of these market distributions, and our process picks this up quantitatively,” Katonah wrote. “Prudence based on measured risk paid off.”
Another manager employing artificial intelligence, Castle Ridge Asset Management, lost 1% in February, but was up an impressive 14% year-to-date in mid-March, thanks in part to a 7.3% gain in January. Cerebellum Capital’s Cerebellum Machine Learning Fund lost a relatively modest 0.3% in February, leaving it up 3.5% for the year.
Katonah launched in 2016. Year-to-date, the fund was up 5.3% at the end of February, following a 4.7% gain last year. The vehicle is negatively correlated to most indexes, including the S&P 500 Index, Barclay BTOP50 Index and Barclays Capital Macro Index.
The Katonah, N.Y., firm is led by Vivin Oberoi, formerly of Telluride Asset Management and Bridgewater Associates; Al Vinjamur, formerly of SAC Capital; and Alex Loureiro, formerly of Cantor Fitzgerald.