The Wall Street Journal published an interesting profile of a small hedge fund which employs an “artificial intelligence” strategy seeking to replace human judgment with the wisdom of computer algorithms.  Spencer Greenberg, founder of Rebellion Research, is the 27 year old son of well known value investor Glenn Greenberg.  The hedge fund has also attracted the personal funds of Jean-Marie Eveillard who is also well known for employing value investing strategies.  One interesting aspect of Mr. Greenberg’s fund is that holding periods average four months.  This is hardly a long term strategy but a marked contrast to the minutes or seconds that positions in high frequency strategies are often held.  A video interview of the managers of the fund appears below.

For RSS Feed subscribers, please click on this link for the video.

Taking human judgment and emotion out of the investing process could help some investors who lack the judgment to make intelligent decisions on their own, but the same is also true of index funds.  Of course, funds like Mr. Greenberg’s aim to beat the overall market rather than just match it. But can computers really substitute for the wisdom of skilled value investors applying years or decades of experience, some of which involves business analysis factors that cannot be quantified?

Whether the fund succeeds over longer periods of time with larger amounts under management is an open question.  It is possible that some algorithms will succeed by exploiting market inefficiencies that are undetected by other participants.

The fund does not claim to have a “value” strategy but does employ some criteria that are supposed to consider valuation factors.  Obviously, the majority of value investors would never employ a quantitative strategy of this kind but it will be interesting to watch the performance of the fund given the support it has received from well known value investors.

Disclosures:  None

Artificial Intelligence Attempts to Replace Human Judgment

6 thoughts on “Artificial Intelligence Attempts to Replace Human Judgment

  • July 14, 2010 at 4:25 pm

    “But can computers really substitute for the wisdom of skilled value investors applying years or decades of experience, some of which involves business analysis factors that cannot be quantified?”

    This question used to be asked about chess-playing programs. When they reached “grandmaster” level, people stopped wondering — the answer was:

    . . . Yes.

    Actually, the question is badly put. The _right_ question is something like:

    “Can computer programs use their capabilities of massive data analysis, free of emotion, to equal or exceed the performance of skilled (human) value investors?”

    Give it 5 years . . .


  • July 14, 2010 at 4:38 pm

    It’s a possibility but I’m in the skeptic camp given how the investment landscape has been littered with failed quantitative strategies many times in the past. But as I wrote there is a possibility that someone comes up with a proprietary way to exploit some market inefficiency. We can be sure that those who come up with strategies that really work won’t publish the details b/c as soon as others copy it, there goes the alpha …

  • July 14, 2010 at 5:23 pm

    Interesting topic.

    “But can computers really substitute for the wisdom of skilled value investors applying years or decades of experience, some of which involves business analysis factors that cannot be quantified?”

    Does Magic Formula investing qualify as a computer-driven investment process? If so, a potential powerful example of automated application of the accumulated wisdom to which you refer.

  • July 14, 2010 at 6:27 pm

    Interesting point on MF. The difference from my perspective is that the MF is transparent in terms of criteria and makes logical sense whereas many of these computer driven strategies are opaque at best, at least to those who didn’t write the code. Additionally, I view MF more as an idea sourcing strategy than something to follow mechanically, although according to Greenblatt’s website, the mechanical approach has far outperformed the market in the past.

  • July 15, 2010 at 6:42 am

    This is inevitable.
    Technology makes it easier to value stocks and to find mis-valued stocks.

    Simple stock-screens have already eliminated the ultra-cheap Graham / early-Buffett style bargains. It is virtually impossible to find a stock that fits all (not just a few!) of Graham’s suggested requirements.

    Many of the small cap stocks I invest in are of the ‘obviously cheap’ variety. It takes a lot of leg-work, but after I’ve looked at 100 cheap-looking stocks or so, usually 1 or 2 will just jump out as irrationally and obviously cheap.

    I’m not even sure that I’m especially good at it, but I’ve done well through sheer determination to only accept obvious bargains. As Buffett has said “I like to shoot fish in a barrel. But I like to do it after the water has run out”. Well, we’re going to get to the point where computers can recognize both fish and empty barrels, and I predict we’ll see the more obvious bargains start to disappear.

  • July 15, 2010 at 7:23 am

    I guess one question to ask is why algorithms have not *already* advanced to that point if in fact it is an inevitable outcome. Computers have existed for many decades, and have been very cheap and accessible for 20+ years. The computer science associated with this type of code is not particularly advanced, at least not compared with many other fields. Data has existed on markets and individual securities for decades, although XBRL and other data standards are perhaps making it easier to crunch numbers.

    My skepticism is driven by asking: Why now? What catalyst has made it possible for computer scientists to come up with code now that they couldn’t have written 20 years ago? Obviously the ability to have a breakthrough even in a steady state of technology is possible, but it seems doubtful to me. I think opportunities will always exist for human beings to identify securities that other humans (and the code humans write) will overlook. One major source of optimism is the fact that four months is considered “long term” in quant strategies. For those looking at 1 to 3 year horizons, there are opportunities that may be overlooked by computers because there is value but no near term catalyst that the code can pick up on.

Comments are closed.


Forgot Password?

Join Us

Password Reset
Please enter your e-mail address. You will receive a new password via e-mail.