Sunday, November 8, 2015

The ability to make Fermi-ized estimates - a skill for any trader


Physicist Enrico Fermi was renowned at being able to come up with good estimates to answer seemingly intractable problems. He did this through breaking down the problem into  more manageable subproblems and then went through a process of combination. By doing this, he was able to clarify and examine key assumptions along the way.  He was an expert on dimensionality analysis and approximation. He was able to look at problems and determine by the number of steps in the estimate the correct order of magnitude with some precision.

The intractable problem that requires a Fermi-ized estimate has now become a standard  interview question for many firms. For example, "How many cigar smokers are there in Denmark?"Boston?" Ok, in reality, most are not going to be faced with these questions, but the focus is on the logic of how you would arrive at an estimate. 

For many trading problems, the focus has to be on the process of getting an estimate for a forecast. The problems may not be intractable, but they may require some guessing that needs some bounds.

Take the simple case of estimating whether the US will get 2% inflation. Many suggest that if oil prices go up it would not take much for inflation to reach 2%. The forecast will be a function of how important are energy prices to the inflation index and then computing how large an increase in oil prices would have to be to get the index to 2%. If you have a guess on oil with a probability measure, you should translate that information into an inflation estimate it some confidence.

The forecast process can be broken into subproblems that can be reviewed to generate the end prediction. You may be wrong with the forecast, but the estimate will be based on clear assumption that can be explained to others.

No comments: