The "ensemble" approach to finding highest-conviction stock picks from top-class managers, and then finding those stocks with the strongest consensus agreement, is an approach that has been known for years. What's new, however, is using AI tech to sift through databases of managers' choices.
One of those wealth management debates that never quite goes away is whether chasing “Alpha” – market-beating returns – is worth the cost if one believes that markets are broadly efficient over time. (And that, of course, is the big “if.”)
When financial markets have risen with few interruptions for more than a decade, as has been the case since the end of the 2008 financial wreck, it’s easy to see why riding an index higher via a low-cost entity such as an exchange traded fund is appealing. Even so, the hope of being in the top end of investment performers never disappears. And harnessing technology such as artificial intelligence to achieve good results is particularly appealing.
A term that is worth an airing is the “ensemble method.” Ensemble Active Management technology captures the highest conviction stock picks from a range of top-performing fund managers, and then identifies the stocks with the strongest consensus agreement.
The EAM method is not new, but what is new is how AI tech can be used to extract data from tens of thousands of managers’ stock picks to build what a particular consensus view is at any one point. A few decades ago, that was impossibly laborious to do.
Alexey Panchekha, president and co-founder of Turing Technology (a firm based in Wilmington, Delaware), recently talked to this news service about how its model is changing the investment landscape.
Turing has developed the “Hercules System,” a trademarked machine learning platform that can replicate the holdings and portfolio weights of almost 2,000 mutual funds in real time. This allows it to drive EAM technology.
Explaining how the EAM approach works, Panchekha set out the process: assemble a multi-manager platform; extract the “predictive engine” from each fund based on the daily fund holdings, and apply ensemble methods mathematics.
“The extracted, underlying predictive engines are processed through an ensemble methods algorithm, which is then used to build an EAM Portfolio. This final step, the application of ensemble methods to the underlying mutual funds (i.e., predictive engines), creates a new forecasting ‘engine’ that is more accurate than the underlying approaches,” Panchekha said.
Turing works with clients such as wealth managers, family offices, private banks, and others investment firms.
Panchekha said the EAM approach can help deal with some of the biases investors demonstrate as shown in behavioral finance; the approach creates an added layer of discipline. The ensemble approach is not about removing or trying to extinguish bias – everyone has biases – the aim is trying to find those which over time appear to deliver results, consistently.
The scale of what can be done with EAM approaches is vast.
“There are over a hundred million unique and custom EAM portfolios that can be built,” Panchekha said.
The outcomes of the EAM process can be used to feed into exchange-traded funds.
“Every investment portfolio manager is obligated to make ‘bets’ versus the market. Otherwise, they are an index manager. Those ‘bets’ reflect areas that [the] manager believes will translate into added value (i.e., alpha). You can refer to the manager’s bets as their bias. The problem with single manager biases is that every set of bets/biases will work sometimes, but never at all times,” Panchekha said. He gave the example of a belief that technology companies will outperform so the manager overweights tech. When tech does well, the manager does well but when tech lags the manager is obligated to underperform.
“Ensemble methods applied to a multi-expert set of predictors (or a multi-fund platform) are explicitly designed to look across each individual bias, and determine [whether] most of the biases are in agreement. This mathematically translates to a higher likelihood of overall success. Therefore, no one bias can disrupt the accuracy of the whole,” he said.
While some of the terminology appears novel, the idea seems basic
enough: finding where there is strongest agreement on the
investment positions that have done well, and using these
insights to deliver results consistently, thus avoiding the
individual biases and foibles of specific investment managers.
Maybe the term “wisdom of crowds” applies here.