I would highly recommend "The Success Equation" for anyone looking to better understand how luck affects our lives. Thankfully, Mauboussin's new book also provides a host of mental tools that can be used to cope with the fact that luck - not just our ability or skill - determines a large portion of our fate. In fact, it is for that reason that "The Success Equation" goes well beyond one of my other favorite books on the topic, "Fooled by Randomness" by Nassim Taleb. Taleb's excellent book rightly highlighted the human tendency to overestimate causality in certain environments, such as the business world. I credit "Fooled by Randomness" for making me more cautious in my assessment of the perceived skill (or lack of skill) of many individuals or groups (companies, teams, etc.), but I regret that it did not provide many tools (aside from Monte Carlo) or mental models for figuring out the relative contributions of luck and skill to a particular outcome. Mauboussin's book does just this, and for that reason alone it should be a mainstay on the bookshelf of anyone who works in a field where luck plays a central role ("which is most of us," as one of the blurbs says).
Now, Mauboussin doesn't argue that one can make an exact science out of this attempt to disentangle the relative contributions of luck and skill, but he does provide some tools for properly framing the discussion and for at least making the attempt to do so. The author says his goal is to help readers make better decisions than those completely in the dark. I agree with him that this is an important exercise. At least in the business world, it seems like on a daily basis one encounters muddled thinking when it comes to discussing the success or failure of businesses and individuals. And that's despite excellent work from researchers such as Jerker Denrell and Phil Rosenzweig, who dismantled much of the pseudoscience endemic to popular explanations of business success.
Now, I'll just share a few of my favorite takeaways:
* Which information to focus on when making statistical predictions: Mauboussin applies Kahneman and Tversky's ideas of base rates (prior information) and specific evidence to skill and luck. In realms where skill is dominant, use specific information (you know Usain Bolt is the fastest guy so you predict he wins the race). In realms where luck dominates, use base rates (just because a mutual fund manager beat the market on a one-year basis doesn't necessarily tell you she will outperform next year).
* Let's relax with the aphorisms: many people react to the notion of luck with comments like "you make your own luck," etc. These are really just word games. If you work hard and put yourself in a greater position to succeed, that's arguably not luck but rather skill (e.g., "skill" by improving your odds).
* Process versus outcome: when luck exerts a large influence, it is better to focus on the individual's process, not the short-term results. If skill exerts a large influence, you're likely to see a strong link between cause and effect, process and outcome. Similarly, the greater the influence of luck, the larger the sample size you need to make any sensible conclusions. Can we please make knowledge of that fact a requirement for high school graduation?!
* Paradox of skill: Defensive readers take note (at first I was a little offended by all the luck talk!). As skill improves, the variance of skill gets smaller, which paradoxically increases the influence of luck on outcomes. Mauboussin offers examples of this from business, investing, and sports, but I particularly liked the example he borrowed from Gould regarding Ted Williams' .406 season.
* Methods for parsing the relative contribution of skill and luck: there's definitely some art involved here but I think the tools here should be more widely used in various fields. The methods presented include a subjective analysis (using a few key questions), simulations, and true score theory.
* Social influence: the success of many songs, movies, and books is contingent to some extent on social interactions, which means that one person's propensity to purchase an item is dependent on external factors (e.g., his social group or the popular media). This positive feedback amplifies success for the luckiest and vice-versa. Here, the author nicely ties in Watts' fascinating work on social influence (Everything Is Obvious) to the skill/luck discussion.
* Identifying and applying useful statistics: use a statistic that is both persistent - you can expect something to recur -- and predictive - it leads to the objective you seek. The test for persistence relates to the skill/luck discussion - skill is the driving force for statistics that are more persistent, or reliable, and vice-versa. In terms of prediction, we can simplistically think of field goal percentage in basketball, whereby a better percentage translates into your goal of scoring more points on offense. The author presents some interesting findings on a variety of statistics in different fields: in baseball, on-base percentage is more useful than batting average; in business, the two most popular stats, earnings per share and sales, provide mixed usefulness - earnings per share is less persistent than sales but it is more predictive of a corporation's goal, to deliver returns to shareholders; and in investing, the author was unable to proclaim "useful" some statistics commonly relied upon to choose future outperformers, such as past performance or expense ratio. He does however find promising a statistic known as active share.
* Managing luck: includes techniques for diluting the skill of an opponent if you are disadvantaged, and the recommendation to utilize small controlled experiments to better understand perceived randomness. And since we are in the realm of randomness, the author is careful to cite Taleb's crucial advice to know our limits of what can be known, particularly surrounding events with incomputable probabilities and massive consequences.
* Improving guesswork when skill and luck are involved: includes a long list of recommendations, including "always consider a null hypothesis," "assess sample size, significance, and swans," "make use of counterfactuals," and "develop useful statistics."