Coaches and portfolio managers both experience high-pressure situations where they have to decide whether to play it safe, or take the risk.
In the waning days of summer, American Football is the talk of the sporting world. I say American not to confuse our international readers! Football grips Americans as Soccer does us Europeans, and fans both casual and rabid gossip about their team’s prospects. Sport franchises are a great analogy for asset managers and can actually provide valuable lessons on organization, competitive edge, and risk. In the NFL, spending power for talent is equally distributed across teams in order to foster parity and a more exciting viewing experience. No team can leverage their owners’ war chest to hoard better talent. Decisions to overpay (or underpay) haunt teams for years to come. With such rules, success is largely influenced by cohesion, the quality and nature of preparation, managerial decisions, and a dash of luck. Unsurprisingly, we see certain organizations consistently succeeding over rolling time periods and others consistently underperforming. Analytics can help us unpack why.
Decision Making & Loss Aversion
One area where casual fans can directly feel the impact of decision making in games is what head coaches decide to do on nail-biting “fourth down” situations. The classic situation is on a so-called “fourth and inches,” the tentative coach decides to punt, and the fans collectively groan. This recurring moment has captured the attention of the sports analytics community, stemming from seminal research on the efficacy of decision making during these high-pressure situations. The first such paper was by David Romer from UC Berkeley in 2006, which found that coaches systematically fail to attempt short-yardage conversions on fourth down. Coaches instead elect to punt/kick when the odds are indeed more favorable for an aggressive attempt.
This foundational work was furthered by researchers in collaboration with the New York Times, which gave birth to the 4th Down Bot, a real-time analytical tool which provides success expectancy via Twitter so fans can compare their team’s decision with the implied odds. This is the closest thing to Moneyball we see in the NFL. Existing research in the field points to loss aversion as the culprit for passivity. Loss aversion was discovered by behavioral economists Daniel Kahneman and Amos Tversky as a behavioral tendency for humans to misjudge risk/rewards, a topic dear to investors hearts. Another paper on fourth down behavior summarized their research succinctly, quoting a former NFL coach who said, “Losing hurts twice as bad as winning feels good.”
But knowing the reason for passivity and actually measuring the cost of passivity are two separate issues. A recent paper in the Journal of Sports Analytics by Professors Derrick Yam and Michael Lopez of Brown University and Skidmore College focus on providing an accurate approximation of that cost. By using advanced statistical techniques to more adequately measure the ‘slippage’ of risk aversion, they were able to approximate a .4 game per season underperformance on average across NFL teams. For context, that is roughly the difference between a mediocre season (approximately, 7-9) and a borderline playoff team, a staggering cost.
These researchers analyzed over 13 seasons of fourth down plays across all NFL teams. They measured key attributes about the situations such as the quality of the teams (i.e., their win/loss totals as of that date), the score of the game, time left in the game, position on the field, and even temperature/wind characteristics (which may influence the probability of success for kicking, running, or passing). This dataset allowed them to construct a baseline understanding of each situation’s key variables at hand. Unsurprisingly, the decision to go for it (the riskier choice) was most correlated to teams that were trailing in the game and thus needed to catch-up. Notice the strong peak below (in green) for games where the point differential is tight. This shows how in higher stakes the default action is passivity!
Once the researchers collated all of the relevant variables, they used a Random Forest algorithm to assess the importance of these variables in deciding whether or not the choice teams made was indeed favorable, leveraging the aforementioned NYTimes 4th Down Bot’s probabilistic assessment. What they found was fascinating.
Bimodal Probabilities & Loss Aversion
After performing all this rigorous statistical analysis, their findings can be summarized by the chart below. The decision not to go for it is conservative and has its modal peak at a neutral value. This value represents the amount that the decision improved the team’s probability of winning (the green distribution). That is to say, the default decision of kicking/punting rarely improves (or degrades) the team’s probability of winning, and thus is the safe choice.
This contrasts with the decision to go for it (orange), which has a meaningful impact on win probability. This impact is less straightforward, as it has both negative (in situations where the attempts fail) and positive modes. In short, the distribution is bimodal. However, you can clearly see that the positive distribution is denser (higher peak, longer positive tail), meaning it occurs more often. Over time, aggression adds more than the failures detract, thus the positive expectancy of going for it.
Bimodal distributions are scary for humans to interpret in the heat of the moment. Fear of failing is what usually dominates our attention in these situations. With the option to stay in the middle, it’s unsurprising that NFL coaches predominantly choose to kick instead of going for it in these high probability situations.
The analogy to our field of focus (asset management) would be a portfolio manager who chooses to be more conservative in their management of risk after they get ahead of the market in the second half of the year. We’ve all heard the quip about the portfolio manager who is beating the market in October, decides to go to cash and go on vacation. While it’s rare for an allocator to receive a Bahamas-marked postcard from their manager in November, we do see this behavior at the margin. This is especially the case for incentive-based fee arrangements, where the portfolio manager is further incentivized to protect their carry.
How Adopting a Culture of Probabilities leads to Organizational Excellence
Besides demonstrating the cost of risk aversion when probabilities are on one’s side, the study was able to make some longitudinal findings around excellence in a highly competitive arena. As we mentioned before, the NFL is designed to promote parity, and yet we see certain organizations demonstrating consistent long-term success. Perhaps one key insight into this comes from the paper’s cross-sectional analysis of fourth down trends by each team. In doing so, they were able to benchmark if teams were underperforming or outperforming through their fourth-down decision making. Unsurprisingly, the two teams with the best existing track record of fourth-down decisions were New England and New Orleans, two teams that consistently have made it to the playoffs, and in the case of New England, that have consistently won the Super Bowl. The two teams that were most deficient during this time period were Cleveland and Jacksonville, two of the worst franchises in the league.
Key Takeaways for Portfolio Managers
Portfolio managers also operate in a highly competitive arena. Talent is scarce and comes at a highly competitive cost. Finding an edge most often comes by implementing refinements to one’s process. Being able to analyze the deficiency in decision making under stress (such as on a fourth down play for an NFL coach) is key to finding high ROI edge. Just as in this study, it starts with data, progresses to tools, and is synthesized through analysis. Successful organizations in the NFL engage, invest, and ultimately embrace this. So should asset managers who want to build successful, enduring franchises.Published on September 5, 2019