Will Moneyball Finally Crack the Game of Active Management?
Billy Beane seems to think so.
Long-suffering active managers have taken a page from Billy Beane, the legendary Major League Baseball talent scout for the Oakland A’s who found extraordinary success using data analytics to build a stellar roster on a shoestring budget.
What’s more, it appears to be working. Using data analytics techniques made famous by Beane — the subject of Michael Lewis’s seminal biography of him, Moneyball — active stock pickers are beating their benchmarks by wide margins. That’s according to data collected by technology research and development firm Turing Technology Associates, which tracked results from recent live portfolios using so-called ensemble methods, the name for the data analytics techniques.
Now, asset managers and institutional investors are looking to offer products — including some soon-to-be-launched exchange-traded funds — and implement strategies.
“Asset management is a dinosaur. It’s the last industry that hasn’t embraced ensemble methods yet,” said Alexey Panchekha, a PhD in math and physics who co-founded Turing. Beane, now executive vice president of the A’s, sits on Turing’s board of advisers.
Now that results are coming in, Panchekha thinks it will kick off a race. He may be right: One chief investment officer for a large pension fund said his investment office is considering the approach with $400 million in equities and is now talking to external managers about participating in a pilot program.
Thirty-four portfolios using the techniques have beaten their benchmarks significantly in the 51 weeks on average that they have been up and running. Seventy one percent of the portfolios, which represent active managers in five stock categories, outperformed their corresponding peer group from the inception date of each through October 31st, according to Turing Technology Associates. (Using artificial intelligence, Turing determines managers’ positions and best stock picks from public data.)
Turing then evaluated 16 of the 34 portfolios that had a track record of 12 months or more against each of their Morningstar categories. In that case, 12 out of 16 of the portfolios are in the top quartile of their corresponding asset class, which includes both passive and active funds, as of the end of October. In addition, two portfolios were in the second quartile, and nine out of 12 were in the top two percent. Although the track record is short, there’s still less than 1 in a 12.5 billion chance that the strategies could randomly end up in the top 2 percent of results.
While back tests have shown that ensemble active management significantly improves the performance of active managers, this is the first test of live results. To build the 34 portfolios, separate managers and investors each picked their top slate of fundamental stock pickers and then used technology and data from Turing. The final portfolio was based on those stocks that had the highest consensus agreement among the underlying managers.
Panchekha recalled a recent conversation with Beane, who said that what changed baseball was the A’s shattering of the 100-year-old record for consecutive wins.
“That was something that wasn’t supposed to happen. That’s a direct parallel for us, you’re not supposed to launch 16 portfolios and a year later have nine in the top two percent,” said Panchekha.
The logic of using data analytics in asset management is straightforward. Human portfolio managers — who have a limited number of good ideas — have a better chance of outperforming benchmarks when they run concentrated portfolios. But their efforts to manage risks by diversifying into hundreds of stocks dilute their returns.
Ensemble methods, which are old news in other industries, help preserve active managers’ excess returns by providing a different approach to diversification altogether. They use multiple experts — active managers — for diversification.
John Sabre, chairman and CEO of Mount Yale Advisor Services, which advises wealthy families and provides institutional research and advice to multi-family offices and others, is now offering separately managed accounts in six styles using the techniques and expects to offer an ETF early in 2021.
Sabre said the firm met executives from Turing about a year ago — and, after seeing the research, set out to replicate the results using its own process for identifying managers. Mount Yale looks to use the highest-conviction stock picks of about 12 managers in each portfolio, which then hold about 20 positions.
In March, Mount Yale launched a large-cap growth SMA. “There’s obviously been quite a bit of volatility this year, but our portfolio is beating the index by 1000 basis points,” said Sabre. “And that’s the S&P 500, the most indexed portion of the market.”
Michael Ervolini, CEO of Cabot Research, which uses behavioral finance and data analysis to determine a portfolio manager’s specific skills, said the strategy seems promising, but the industry still needs to deal with the mistakes that it makes with active management. The premise of high-conviction ideas outperforming others shouldn’t be accepted at face value, for example. Managers often hold on to their best ideas too long because of behavioral biases like the endowment effect. “You’ll need a finer sieve to sort that out,” he said.
And ensemble methods won’t solve every problem. For one, ensemble methods can’t make mediocre managers great. “You still need managers who are good stock selectors. You can’t just throw darts. There has to be a research piece underneath it,” said Sabre.
It also uses multiple experts to provide diversification. That means large fund companies that use the techniques won’t get the same benefit because all their managers often sit in the same meetings and use the same research.
But Sabre is still excited about the results. “This is a game changer,” he said. “It tips the scale back to active management and stock selection.”