Jeff Bezos used to work at DE Shaw.
Computer science + Value investing =
Very interesting read. Also, what took quant funds so long
The small team at start-up Havelock London has an ambitious mission: to mesh computer science with the Sage of Omaha’s “value investing” principles.
Havelock was founded last year by the former chief investment officer of Winton Capital Management — one of the biggest quantitative hedge funds in the industry with $23bn in assets. But the start-up is ploughing a different furrow from most other algorithmic investors.
Instead of trading in and out of stocks at a moment’s notice, or trying to ride hot market themes like Winton, Havelock’s chief executive Matthew Beddall wants to build a system more akin to a computer-powered private equity firm, going deep into a small number of companies.
Paraphrasing Mr Buffett’s mentor, the famed value investor Benjamin Graham, he argues that quants typically try to make money out of the market’s short-term “voting machine”. Havelock, on the other hand is attempting to profit from the market’s longer-term “weighing machine”.
“Traditionally, quants are a mile wide and an inch deep. We try to be a mile deep and an inch wide,” says Mr Beddall, who spent 17 years at Winton, latterly as chief investment officer. “With the rise of quantitative investing, the market’s attention span has shortened and shortened. We want to build better models to value businesses.”
Quants have been in the ascendant over the past decade, with algo-powered hedge funds like Renaissance Technologies, DE Shaw, Two Sigma and Bridgewater enjoying inflows as much of the rest of the industry struggles. Half of last year’s top 20 hedge funds, ranked by the fees paid to managers, were primarily quantitative — and several of the remainder use at least some strategies based on algorithms.
Even traditional mutual fund groups like Fidelity, T Rowe Price and Capital Group are now spending huge sums on technology and technologists in an attempt to enhance the abilities of their portfolio managers.
The results are not always good, with the solid performance mostly concentrated in the biggest funds. Only 11 per cent of US equity quant funds have managed to beat their benchmarks this year, according to Bank of America.
Some analysts say that with quants increasingly mining real-time feeds of alternative data — such as credit card sales, app downloads, satellite images, social media chatter and mobile phone geolocation — the entire investment industry is speeding up. The catch is that the profitability of many of these signals tends to decay rapidly, forcing funds into a never-ending hunt for new ones.
However, that may have opened up richer opportunities for investors with a longer-term horizon, according to Savita Subramanian, head of US equity and quantitative strategy at BofA in New York. She estimates that valuations explain nearly 90 per cent of the S&P 500’s returns over a 10-year horizon — better than any other factor the bank’s analysts pored over.
Havelock’s six employees currently track 38 companies, and are adding just one company a month, they say, to ensure analytical rigour. The investment group builds models to value these businesses using a combination of human judgment and algorithmic analysis, and once the model is constructed Havelock trades largely on autopilot. The £14m fund was launched in August 2018 and has returned 5 per cent since then, roughly double the gain of the MSCI World index over that period.
The fund manager is satisfied with the initial performance, but admits that Havelock’s value-oriented approach has been a headwind in a market that still prefers faster-growing but pricey companies. Value investors prefer cheaper stocks, either measured by their earnings or their assets versus their share price. “Buying expensive stocks doesn’t make much sense at this point, but it’s the most expensive things that have done the best over the past 12 months,” said Mr Beddall.
Combining the systematic investing approach of quants with longer-term, deep-research value investing style is tricky. Quants say it is much easier to use statistical sciences to predict near-term performance of a security — whether over a day or a millisecond — than over the next year, given the vast amount of factors that influence a company’s performance.
However, Mr Beddall remains optimistic that combining these two approaches is feasible. “It’s not impossible, it’s just a little harder,” he said. “If you look at what Buffett’s done, he looks at what companies might be worth, rather than what they’re going to do in the next quarter.”