On Being Wrong
For decades, Peter Lynch’s Fidelity Magellan fund dramatically outperformed the market, averaging 29% versus 9.5% for the S&P500 between 1977 and 1990. But when Fidelity studied the performance of investors in Lynch’s fund, the firm found the average investor had actually lost money.
Investors, it seems, have a curious ability to mistime the market, selling at the bottom and buying at the top. One recent study found the gap between equity market investors and the returns of the S&P500 was 4.2% per annum over the past 20 years. Those that trade more frequently, cycling out of the stocks on which they were “wrong” and into the newer “hotter” stocks on which they have a chance to be “right,” do consistently worse. On average stocks sold do 3.3% better than purchased stocks.
Error intolerance leads investors to limit their returns by chasing an unattainable “perfect” strategy. But research suggests that the perfect is the enemy of the good. For example, one study pitted Yale students against rats in a test of predictive ability. Two food dispensers were positioned at either end of a T-shaped maze, with food dispensed food 60% from one dispenser and 40% from the other. The rats quickly learned to choose the end of the maze from whence the food emerged 60% of the time. The Yale students' performance was significantly worse, at 52% accuracy, because they tried to find patterns in the random distribution and didn't stick to the higher probability food dispensary.
Investors who strive to over-control, like the Yale students, are doomed to failure, because they have committed themselves to a strategy (performance chasing) that is empirically proven to limit returns. And much research has shown that even perfect strategies experience large drawdowns.
Indeed, our friends at Alpha Architect designed a “perfect portfolio.” They “looked ahead” five years from portfolio formation and built portfolios of large cap stocks poised to do the best. The performance was amazing, but they also found something surprising, higher volatility and similar drawdowns compared with the market as a whole.
Figure 1: Five-year performance of “God’s Portfolio”
Source: Alpha Architects
Their conclusion: “even GOD HIMSELF would get fired multiple times over. The performance on the perfect hedge fund would get crushed many times over by the passive index.”
Academic research reveals that people will even abandon good processes in favor of bad processes if they see a good process fail. As we have pointed out before, simple algorithms consistently outperform humans at forecasting. But when academics ran a psychology study that gave observers the ability to bet on either an algorithm or a person, the observers preferred the human even after seeing the algorithm perform better.
The study authors write that people have greater intolerance for error in algorithms, and people are more likely to forgive people for making mistakes than to forgive algorithms. Participants, having seen the algorithm err had to be much more confident in its predictive ability to choose the model at the same rate as the control group.
Figure 2: % of Participants Choosing Statistical Model
Based on a figure from Dietvorst, Simmons and Massey
Research suggests that people are more likely to flee from volatile strategies and more likely to abandon effective algorithms than ineffective experts. When an expert sells us a low volatility strategy, they are really selling lower returns in exchange for greater piece of mind and less of a chance they will get fired. When it comes to getting fired, an expert can dodge the axe by explaining how 95% of their analysis was nothing short of clairvoyant, but the last, unexpected 5% ruined the prediction. Algorithms however have no such dexterity; every incorrect prediction calls into question their entire process and accordingly their value.
Perhaps this is because of unreasonable expectations. We expect our fund manager to perform every year, and algorithms to be flawless. But the world works probabilistically. We’re looking not for the 100% right outcome — this is likely to be the result of over-fitting or pure chance — but for the process that yields the right answer >50% of the time. We want the rat, not the Yale student.
This requires a mindset shift, away from hoping for perfect forecasts from algorithms, away from preferring steady, smooth returns to volatile and dynamic ones, and toward strategies that harness probabilistic insights to generate significant returns in a volatile world. This shift necessitates a reversal in how we perceive volatility and errors.
Nassim Taleb argues that investors should relish errors. “What is antifragile loves randomness and uncertainty, which also means — crucially — a love of errors,” he wrote. As long as those errors are calculated as part of a probabilistic mental model, the investor who is most comfortable with mistakes, like the rat, will be most successful, or as he describes it “antifragile.” Investors who seek consistent high returns in a stochastic market will achieve neither and be left explaining how they were mostly right.