Lynne Kiesling
I’ve recommended Jonah Lehrer’s The Frontal Cortex blog before, and if you haven’t checked it out, here are two more reasons to do so. His most recent post discusses Bill Belichick’s decision to go for the first down from 4th and 2 in Sunday night’s Patriots game, and ties it to David Gordon’s research on whether or not NFL coaches follow the optimal 4th down strategy:
… it illustrates the difficulty of making rational decisions, even when the evidence supports the call.
I’ve blogged about the research of UC Berkeley economist David Romer before, but his basic thesis, based on an exhaustive statistical analysis of 4th down scenarios, is that NFL coaches are irrationally risk-averse. They punt the ball way too frequently and kick far too many field goals.
Belichick was an econ major, and has expressed a familiarity with Romer’s research.
Lehrer then goes on to discuss this risk-aversion research, with links to other analyses of Belichick’s decision. One of the fascinating aspects of the 4th down decision that Lehrer highlights is that Belichick was statistically correct to go for it, but it’s emotionally difficult for coaches to make that call (and for fans to endure it). The probability part is also interesting — even with a higher probability of making a field goal, this research shows that going for the 1st down on 4th down increases the probability of winning.
On Tuesday Lehrer also remarked on the research of my Kellogg colleague Jennifer Brown, who does some of the most interesting work I’ve seen in a long time. In her new working paper, “Quitters Never Win: The (Adverse) Incentive Effects of Competing with Superstars“, Jen finds that golfers in PGA tournaments perform more poorly when competing against Tiger Woods, especially when Woods is playing well. She and Lehrer have different hypotheses for this result, as Lehrer notes:
Brown argues that this phenomenon is caused when “competitors scale back their effort in events where they believe Woods will surely win.” After all, why waste energy and angst on an impossible contest?
That hypothesis is certainly possible, but I’d argue that the superstar effect has more to do with “paralysis by analysis” than with decreased motivation. I’d bet that playing with Tiger Woods makes golfers extra self-conscious, and that such self-consciousness leads to choking and decreased performance. The problem, then, isn’t that golfers aren’t trying hard enough when playing against Tiger – it’s that they’re trying too hard.
I wonder if there’s a way to test these two hypotheses? I think given her data that it might be difficult; testing such a hypothesis may require biometric data like heart rate, sweating, etc. I frankly am more inclined toward Lehrer’s hypothesis, based on my reading of neuropsychology and my non-Tiger-Woods-like experience of athletic competition; the “trying too hard” fits with my experience of athlete psychology. But I’d really like to see if there’s a way to discriminate between the two.
Most professional athletes are highly competitive people who hate to lose. I would tend to agree with Lehrer.
I think you could test Lehrer’s hypothesis by looking at the performance of Tiger’s playing partners compared to what they and/or others do when not playing with Tiger.
If he really is intimidating, then I think he would have the greatest impact when his opponents get a first-hand look at him.
I suspect that detailed examination of golfer’s shots would allow discrimination between “scale back efforts” and “trying too hard.” That is to say, golfer’s likely play holes differently under the two hypotheses, so seeing how they actually play when up against Tiger could reveal which explanation works better.
Simpler to examine, and possibly revealing a difference, I would guess that “scale back” implies scores that are higher on average but likely with lower variance, while “trying too hard” produces scores that are higher on average but with a higher variance. (Trying “hard enough, but not too hard” is the ideal, which gives a desirable below average score and a typical variance.)
So, what happens to the variance of scores of all of the non-Tigers with and without Tiger in the tournament?
… a brief pause while I look at Brown’s paper …
She says: “Results in Table 10 suggest that the presence of the superstar does not lead to increased variance in players’ scores.” and “Overall, this hole-level analysis provides little evidence that Woods’s presence induces players riskier strategies that result in higher scores and the observed superstar effect.”
Case closed on the “trying too hard” explanation?
Dan, I interpret Table 9 of the paper, the “intimidation” analysis, as getting at what you have in mind, and the data don’t suggest any “intimidation factor” relative to playing with others. But that analysis may not be what you had in mind …
Mike, I saw that result, and I agree that it’s suggestive and consistent with the “scale back effort” strategy. To rephrase my question to incorporate this discussion: in the “trying too hard” hypothesis, is there a way that such a strategy would manifest itself other than in score variance? How might golfers try too hard that might differ from “riskier” strategies?
The cool thing about Jen’s dataset is that she can control for within- and between-tournament variation, as well as cross-player variation, and that she can do the hole-by-hole analysis in the “risky strategies” analysis. This yields an estimate of within-round variance by player, so we can compare across players with and without Tiger.
I interpret Lehrer’s “paralysis by analysis” as possibly manifesting itself in some other way than through riskier shots on a hole-by-hole basis, which is what Jen tested for in Table 10’s estimates. Here’s what I have in mind: professional athletes train for precision and consistency, so even if they have a “paralysis by analysis”, they are still so consistent and precise because of their skill and training that they are still performing in a very high range, but that being self-conscious and overthinking things at the margin still results in lower within-round average scores by hole, even though it doesn’t result in higher within-round variance.
So I’m still searching for a way to tease the psychology result out of the data!
Hey, this is a fun discussion, we should do stuff like this more often …