Hi Guys,
A few days ago an MFO member raised a question about the relative impact of computer trading over manual investment decisions and order placement. In my perhaps overly rapid response, I did not find that development too worrisome since the deployment of computers should decrease error rates and improve efficiency.
I do not fear the HAL scenario in the “2001: A Space Odyssey” movie because man is in charge of the computer programming. We control the machines which merely serve as a useful tool.
Significantly, the machines are coded by experts who determine the independent parameters to be considered in any decision matrix, the option tree to be explored, and define the final rules of engagement. While composing my post I was thinking of the IBM Deep Blue supercomputer pitted against chess champion Garry Kasparov battle won by the damn machine in the late 1990s.
Later, I coupled these thoughts with the almost universal finding that experts are not superior forecasters over us common folk. This is not new stuff even in the investment world. Way back in 1932 Alfred Crowles asked the question “Can Stock Market Forecasters Forecast?” His answer was firmly grounded in the negative after examining multiple data sets. Here is a Link to his study that I have posted earlier:
http://cowles.econ.yale.edu/archive/reprints/forecasters33.pdfThe observation that experts were no better at predicting the future than an ad hoc group of reasonably informed amateurs was reinforced by studies made and reported by Paul Meehl in the early 1950s. More recently Phil Tetlock has picked up the torch. Their combined findings over divergent subject matter and extensive timeframes have been persistent.
Meehl postulated that experts would make more errors in data interpretations than a mechanical device crafted for the same decision purpose. Tetlock’s findings reinforce Meehl’s earlier work. Here is a Link to a New Yorker magazine article that summarizes some critical discoveries in this arena:
http://www.newyorker.com/archive/2005/12/05/051205crbo_books1From this “Everybody’s an Expert” piece written by Louis Menand I have extracted a few pertinent observations.
“Knowing a little might make someone a more reliable forecaster, but Tetlock found that knowing a lot can actually make a person less reliable.”
And quoting from the Tetlock book: "Expert Political Judgment” is just one of more than a hundred studies that have pitted experts against statistical or actuarial formulas, and in almost all of those studies the people either do no better than the formulas or do worse.”
And finally, “The odds tend to be with the obvious.” That’s nothing more than a restatement of Occam’s Razor.
I recommend that you access the Menand article.
Paul Meehl has passed away, but Phil Tetlock has not only continued the march, but has accelerated the pace and scope. For the last few years he has been under contract to the federal IARPA unit to more fully explore and exploit his earlier research. His ultimate goal is to uncover ways to improve forecasting capability.
At the midpoint of this current project, he has identified a small cluster of talented(?) forecasters. With targeted training and specialized group formation and procedures, he believes he has achieved some marginal enhancements to forecasting scores. By the way, his team uses the Brier probabilistic scoring method to assess forecasting accuracy. Here is a reference to a recent article that reports on this subject:
http://www.nytimes.com/2013/03/22/opinion/brooks-forecasting-fox.html?_r=0This article addresses some of the tension and distinctions between Hedgehogs and Foxes. Read the article to learn why.
Perhaps within our lifetimes we will gain a little more success to more accurately forecast market returns.
I find this topic fascinating. I hope you do too.
Best Regards.
Comments