Man Plus Machine

This weekend on Masters In Business Barry had on (for the second time) one of my favorite thinkers, Michael Mauboussin. Early on in the interview, while they were discussing quantitative models, Mauboussin brought up an interesting point- what happens when you combine the power of a machine with the power of the human brain?

Mauboussin says:

Kasparov loses to deep blue in 1997 and so machine beats man, fine. But shortly thereafter what emerges is something called free-style chess. That means you and I are playing a match but we can avail ourselves of whatever aides we want. So we can run computer programs, you can call your lifeline, your grand master buddy or whatever it is. And it turns out these free-style teams are better than the programs by themselves or of course any man or person. So man plus machine beats man or machine…..This to me is a very intriguing model to say are there cases, where most of the time you default to the quantitative approach or the program, but every now and then if you have a good feel for the game you can step in and do something a little bit different that adds value. Now that’s an open question, whether that’s true in investing but to me that’s a really intriguing model for thinking about where humans and quantitative techniques may emerge in the future.

The idea of combining man and model was tested a few years ago by Joel Greenblatt. Here’s Tobias Carlisle and Wesley Gray From Quantitative Value (emphasis mine):

In 2012, Greenblatt conducted a study into the performance of retail investors using the Magic Formula over the period May 1, 2009, to April 30, 2011. Greenblatt’s firm offers two choices for retail investors wishing to use the Magic Formula, a “self-managed” account, and a “professionally managed” account….What happened? The self-managed accounts, where clients could choose their own stocks from the preapproved list and then exercise discretion about the timing of the trades, slightly underperformed the market….The aggregated professionally managed accounts returned 84.1 percent after all expenses over the same two years, beating the self-managed accounts by almost 25 percent.

This itself is proof of nothing. It’s not surprising that retail investors can’t add value to a quantitative model. But Greenblatt then turned this around on himself to see if he, one of the most successful investors ever, could add value to his model. Here’s Carlisle and Gray again:

Even Greenblatt has said that he cannot outperform the Magic Formula. He ran an experiment in which the Magic Formula selected stocks for him and his business partner to buy, and then they both went through the list and removed those stocks they “knew” to be value traps. They lost to the Magic Formula. Even a stock in which Greenblatt had personal experience fooled him. The model selected a pharmaceutical stock, but he knew that its main product was about to go off-patent, so he overruled the model. “And the thing doubled in six months.

Again, this is just one example, using one model, over one short period of time. However, I’m skeptical we’ll ever find evidence that humans can interfere with an investing model and add value, over time. If you’ve wasted two hours watching Terminator Genisys, you’ll know exactly what I’m talking about.


Masters In Business

Quantitative Value