Super Models

Jim Simons, the famous mathematician and hedge fund manager of Renaissance Technologies, has built a money management empire based on using mathematical models.  Renaissance manages an estimated $21 billion throughout its funds, and it’s $3.5 billion Medallion Fund has amassed estimated annual returns of 35% since March of 1988.  Models are a powerful tool in all forms of research.  Research shows that on average models outperform expert human analysis consistently.  A meta-analysis study done by Grove, Zald, Lebow, Snitz, and Nelson shows the power of this.  Meta-analysis is the statistical procedure for combining data from multiple studies.  When the treatment or affect is common amongst the studies the data can be used to show comparisons.  In the Grove, Zald, et al study comparing decision based models as compared to human based decisions using metrics in different fields that ranged from military training to university admissions showed that models equal or beat experts a whopping 94.12% of the time.  In the study experts only beat the models a mere 5.88% of the time with the models beating experts 46.32% of the time.  Models equaled the experts 47.79% of the time.  

What is even more amazing is an additional model versus expert study performed by Leli and Filskov experts where put up against models and the algorithm based models where accurate 83.3% of the time and the experts had a success rate 58.3% of the time, and inexperienced clinicians where accurate 62.5% of the time.  The study was then taken one step further where the algorithm based model was provided to the experts for use in their analysis.  They expected the experts armed with the model to outperform the models.  In human fashion the experts improved their accuracy up to 75%, and the inexperienced clinicians raised their scores to 66.5%, yet neither of the groups outperformed the model in it’s isolated form. All the experts had to do in the study was follow the models and they would equal the models performance.  Instead they under performed the models by adding, changing, or questioning the data.  Human behavior in it’s typical form slayed itself to create an outcome less than machines.  The human brain, with its empirical computing power combined with human emotions, takes simplicity and creates complexity.

What this and other studies prove is that humans will always struggle making investment decisions or choices.  Human emotions and thoughts consistently question oneself to the point of inefficiency.  These types of studies back up the reasoning behind making sure an investor takes a long term perspective on their investing, and focuses on using value for decisions as a framework.  Using these concepts provides two important elements that prevent human behaviors from getting in their own way.  One, long term investing keeps the investor away from their choices, so they aren’t constantly questioning or second guessing themselves.  Two, when choosing an investment its value is a black and white element.  Something is either under, at, or overvalued at the time of analysis, so a rule based decision process can be built around it in order to make decisions.

The last comment I will make on this subject is that today computers dominate the investment market.  They move enormous amounts of data and money around fast.  Institutions and hedge funds thrive in this environment.  As an individual investor you want to stay away from any type of element that could expose you to this.  If you are buying businesses or investing for the long term your only exposure to these elements is when you buy something and when you sell something.  Other than that if you want to be involved take the approach of supporting businesses you believe in, and monitor them by using or visiting them, and use that as your analysis over time.  Forget the market.