Daniel Kahneman speaks out on NBA fans, Golden State Warriors, and Klay Thompson

The following interview with Nobel Prize winner Daniel Kahneman may have taken place.

I recently was given an opportunity to sit down with Dr. Daniel Kahneman who won the Nobel Prize in Economic Science back in 2002. Kahneman is a pyschologist by training and spent his career working on the psychology of human decision-making. What follows is the transcript of our session.

****************

EZ: Dr. Kahneman, it's so great to meet you. I recently read your book Thinking Fast & Slow and it really opened my eyes to a lot of the flaws inherent in human decision-making. When I heard from an unnamed NBA league source that you were beginning to work on the psychology of NBA fans and analysts, I knew I had to reach out to you and get your opinions on some things that have been bothering me. Before we get to that, can you just talk a little bit about how you came to be a NBA fan and, specifically, why you root for the Golden State Warriors. Is it true you once met Jessica Alba?

DK: Evan, first let me say it is a pleasure to meet with you. As you know, I follow you on twitter, and ever since Hoopdata went down, nbawowy.com has been my main source for getting advanced stats when players are on or off the court. Thanks so much for making that data available to "stat geeks" like myself.

One thing I must get off my chest, though. Why are you so hard on the young Harrison Barnes? He seems like such a mench [Ed. note: "mench" is a yiddish word meaning essentially "good guy"]. Take it easy on him. You know, he's just 21.

So you want to know how I became an NBA fan? Well, as you probably know, basketball is very popular in Israel. Although I have been in America for many years, starting at Berkeley back in the mid-1980's, I have followed Israeli players who have come over to the states. You might remember Nadav Henefeld who played for UConn Huskies many years ago, and now you have some recent players here, including Gal Mekel who is playing for the Mavericks and Omri Casspi who has moved around a bit, but seems to be helping the Rockets off the bench.

I became a Warriors fan while at Berkeley and was a huge fan of Run TMC. It's great to see the team having some success now after so many years of terrible ownership. Even though I've been at Princeton for many years now, I still follow the Warriors, because let's face it, the alternative was not so great living in New Jersey. As for Ms. Alba, I think that's a story better left for another time in a forum, perhaps, a bit more private than this one.

EZ: For people who haven't read any of your work, what would be the "take home" message you would give them? As they say here in the Valley, what's your "elevator speech"?

DK: Basically, myself and my late colleague Amos Tversky [Ed. note: Tversky passed away in 1996, but would have won the Nobel Prize along with Kahneman, which is not awarded posthumously] theorized that human decision making can be modeled by two "Systems" in the brain, one which makes quick intuitive judgements based largely on heuristics ("tricks"), and the other which makes more rational decisions, but needs more time to operate. As I say in the book, you may think this slower more rational System 2 is "the hero" (much like Kobe Bryant at the end of close games), but it turns out that in the face of tough decisions, humans are often lazy and revert to the easier heuristics of their more intuitive System 1 side of the brain.

For example, humans are quite risk averse, and often make decisions that would be counter-intuitive if given more thought (even I make these same mistakes from time to time, and I literally wrote the book on it!). To give one example, let's say that you are given two choices:

A) 90% chance to make $100 and a 10% chance to make $0.

B) 100% chance to make $25.

Our research showed that most people will go with the "safe" $25, even though the expected value of choice A is much higher than that. It's actually quite interesting to think about how risk aversion applies to NBA analysis (especially among fans), and is one of the things I am now carefully studying.

Of course it's more complicated than this, so you'll have to buy my book on Amazon to get the full details.

EZ: Would you mind giving some examples of the kind of work you're focusing on regarding the NBA?

DK: Sure, no problem. Let's talk about one specific problem that I have spent considerable time studying, and that is how fans use the word "inconsistent" to describe certain players.

EZ: Interesting. Please go on.

DK: Well, let's take Klay Thompson. This kid is a prolific and productive shooter. He's Jerry West's "golden boy". And yet I come across many fans on twitter and other forums around the web (when I have time for such things) who continually criticize Klay for being "inconsistent". It got me to thinking about what that means, why fans might think he is inconsistent, and what it means in terms of psychology. It turns out, interestingly enough, that Klay Thompson is a perfect case study for my research. It's as if my life's work in human decision-making has led me to study the relationship of Warriors fans and this young "flame thrower".

EZ: I'm all ears, Dr. K. I'm a huge fan of Klay Thompson, and I've felt for a while now that some fans tend to judge him unfairly, and that it's possibly due more to psychology than actual productivity. I would be very interested to hear your assessment.

DK: It turns out that NBA fans, like any other form of human beings, exhibit common heuristics and biases in their decisions. Here is a specific example. We recently interviewed a panel of NBA "experts" from many different sources, including, for example, Grantland and ESPN. We split the panel into two groups (A & B) and asked the following questions, respectively:

• Group A: Klay Thompson is shooting 40% from 3-pt range this season. He has had 15 games in the past two seasons where he made 5 or more 3pt attempts. On a scale of 1-10 how do you rate Klay Thompson's consistency (1 being least consistent, 10 being most consistent)?

Group B: Klay Thompson is shooting 40% from 3-pt range this season. He has had 20 games in the past two seasons where he made 0 3pt attempts. On a scale of 1-10 how do you rate Klay Thompson's consistency (1 being least consistent, 10 being most consistent)?

You can probably guess the responses, right? It turns out that Group A rated Klay Thompson 7.5, as opposed to Group B, which gave him a 4.5 rating. This is a clear example of anchoring effect (look it up in my book!). People tend to be quite easily pushed in a certain numerical direction with a little help.

EZ: So I hear quite a lot of Warriors fans lobby this criticism at Klay, the notion that he is extremely inconsistent. Do you have anything to say about that?

DK: Well, I am not a statistician, per se, but the notion that fans have about "inconsistency" seems to quite often be dealt with quite lazily by System 1 (typically based on a recent outcome, say, the last game that was played), because it is not an easy issue to tackle for most people in a rigorous objective fashion (something that System 2 is better at, but again, often lazy).

Like you, Evan, I have heard some fans complain that Klay is inconsistent, so I actually consulted with some statistics colleagues of mine and they helped me do a bit of investigation into the data. First, we developed a fairly simple model of Klay's 3-point attempts per game since he became a starter a couple seasons ago, about 150 games of sample data. For a first pass, we assume that the distribution of 3-point attempts in each game is modeled by a Poisson distribution, which is quite often used to model the occurrence of discrete phenomena in sports (think touchdowns in football or goals scored in soccer). Klay has averaged about 6 attempts per game as a starter, so we used this figure as our "lambda" (the governing parameter for a Poisson distribution). You can see the distribution of attempts in a simulation here:

Now compare that distribution with the actual or historical distribution of 3pta in the sample for Klay shown below:

Now, the model is by no means a perfect replication of reality, but indeed, when we look at Klay's actual historical attempts per game, it looks almost more "consistent" than the model. The distribution is roughly what one might expect. Most of the time Klay is within a couple shots of his average, sometimes quite a bit less, and every so often, quite a bit more. Note that the model, while not quite predicting games with 15 attempts as really happened, does predict a few games of 13+ attempts. Not bad for a first pass.

So with the attempts model taken care of, we then turned our attention to 3pt shots made and did the same analysis. Only this time we modeled each game with a binomial distribution, given the attempts from the initial poisson distribution. In fact, in case your readers might be interested in replicating such a model, here is the very short amount of code needed to do that in R:

require(plyr)
vals <- function(num_games,avg_3pta,eff) {
  attempts <- rpois(num_games,avg_3pta)
  makes <- ldply(attempts,function(x) c(x,rbinom(1,x,eff)))
  colnames(makes)<-c("fga","fgm")
  makes
}

Again, here are the results from the model of field goals made:

And compare those to the reality:

Once again we see that while the model is not perfectly replicating the reality, it does seem to represent the broad strokes. Perhaps, if you were an art lover, you might call this a Monet model.

EZ: Haha. Are you implying Monet Ball is the next big thing in analytics? Just kidding. Please continue Professor! This is fascinating stuff.

DK: Indeed, how do the kids say it? I see what you did there!

So, to cut to the chase, the idea here is that purely by creating a model of randomness, we can replicate quite well "inconsistency" in a player's 3pt shot making ability. It turns out that when you have only a few samples of something in each game, in this case 3pt shots, you can't expect a player to truly be "consistent". Imagine what that might mean even. Would one expect, for example, a shooter to make 3 out of 8 shots behind the arc in every game? That would be remarkably, if not supernaturally consistent. If those are your expectations as a fan, you are sure to be disappointed.

The other interesting thing to note about this phenomenon --- and here I mean the fan reaction --- is that fans typically do not complain about "inconsistency" unless a player has just had a bad game.

You don't hear fans after Klay goes 7 for 8 complaining that he is too inconsistent!

EZ: Of course! That makes perfect sense. Fans want all of the upside of probability with none of the downside, and their System 2 side of the brain can't quite comprehend all the implications of the randomness involved. Do I have that correct?

DK: Yes, Evan! It basically boils down to the fact that fans do not cope very well with statistics that are any more complicated than say, how many points a player scores in a game or how many rebounds he pulls down. Statisical notions like "consistency" are often lazily swept to the System 1 side where an easier heuristic like the result of the last game is substituted for the more complex reality. It's really quite fascinating.

EZ: So what would you suggest fans do about all this? Just give up?

DK: Well, some of these notions are not trivial, and we can't expect everyone to understand Poisson distributions, etc. But hopefully there are some fans (perhaps, like yourself) who can once in a while try to give some perspective and just make them think twice about what they really mean when they use nebulous terms, such as "inconsistency". I suspect if you actually asked such a person for data to backup their assertions, they might realize they spoke too soon. Indeed, they might find it interesting to have a longer discussion about how they could go about finding such data and what kind of analysis they could perform.

EZ: Well, Dr. K, we're about out of time. But it's really been a pleasure talking with you today. I hope you might come back some time, and we can talk about your work on hedonic pyschology.

DK: It's been my pleasure, Evan. Keep up the good work. Maybe we'll meet at the MIT Sloan Conference one of these days! Let's go Warriors!