The Art of Good Guesswork
I stumbled onto Michael Mauboussin’s writings with his white paper, which researched the moats of companies. He has written several of the books about decision making. My favorite was The Success Equation. He states the purpose of the book is to help people to think clearly about skill and luck and to provide some analytical methods to untangle the two. I tried to pull out and summarize the main points.
Understand where you are on the luck skill continuum.
How much of the result you saw is do to luck and how much is do to skill. Certain activities like investing have a high degree of luck (especially in the short term). For an activity like chess where there is almost no luck, then focusing on outcomes can be a good way to get feedback on how well you are doing. If the activity involves a lot of luck focusing on the process rather than the outcomes makes a lot of sense.
What makes a good decision making process?
Mauboussin recommends developing a good process by focusing on the analytical methods, psychological factors, and organizational structure. An example of this in investing would be as follows:
- Analytical Method: Value based investing approach which has a long track record of good performance over the long run.
- Psychological Factors: Can you gut it out and hold onto stocks instead of selling? Adding in rules that prevent you from selling more than X% of your portfolio in a given year.
- Organizational Structure: If some one else is investing your money, you need a way to align your interests to tackle the Principal/Agent problem. Instead of the typical 2% of assets and 20% of profit approach, can you do 25% of profits above 6% (compensation structure from the Buffett Limited Partnership)?
Compare results against a null model.
One way to determine validity of a result is to figure out the probability that it happened by chance. If you know the distribution of outcomes, where does this one fall? Its important though to remember that a 95th percentile event still occurs 1 out of 20 times due to pure chance. In activities where skill and luck are present, its a good idea to attribute some of the outcome to each.
Make use of counterfactuals
Another useful way to deal with activities involving luck is to look at the other paths that could have happened. This will keep you thinking about outcomes in a probabilistic way, and it prevents hindsight bias from attributing more certainty to the outcome than is warranted.
A great experiment that highlights this is the “Music Lab” experiment. In this example, experimenters loaded PC’s with different songs. When they had people listen and rate the songs in isolation, the results where fairly consistent. Songs in the middle, tended to finish in the middle. When they networked the feedback so that you could see what other people thought about the songs, the results were inconsistent. Some times “middle” songs would finish in the top spot if they got a couple good ratings early. Other times, one of the top songs would finish in the top spot. By running the experiment multiple times, the experimenters highlighted the probabilistic nature of the outcome.
Make reversion to the mean work for you.
If you had a really bad result, remind yourself that the next time you do this activity, you are likely to do better. If you had a really good result, remind yourself that the next time you might not do quite as good, just due to chance. This can lift your spirits and help keep you humble.
Take the outside view.
When ever you are trying to estimate or predict, it can be very useful to ask what the average and best case scenario are. This can help combat many of the biases that we have. One example of this is the planning fallacy which says that people almost always under estimate how long things take. By understanding how long the task typically takes or what's the shortest time that it has ever been done in, you can help move your prediction back in line with reality.
Develop Useful Statistics
There are two key factors when evaluating a statistic. How consistent is it and how predictive is it. A straight forward example of this is baseball. The team that scores the most runs wins the game. There is a decent correlation between a batter’s hitting percentage and how many runs they generate. It takes a decent # of at bats to see the true batting average take place. A more consistent statistic is strike out percentage, but it doesn’t have as strong of a connection to runs generated. An even better statistic is on base percentage. It corelates better to runs generated than strike out percentage, and it is more consistent than batting average.
Conclusion:
Michael Mauboussin has a lot of great ideas for how to make decisions, and how to think about problems under uncertain conditions. His books have definitely had an influence on how I think about my investing, and how I evaluate my portfolio's performance.
If you enjoyed this story, feel free to applaud and follow me to be notified of more stories like this.
