My ESPN the Magazine showed up in the mail the other day, and it’s their analytics issue. Obviously you have to walk something of a line in doing something like this; it has to be appealing and understandable to general fans while still respecting the stats work that went into it as well as what’s going on now. I thought I’d give a rundown of some of the highs, lows, and questions that cropped up as I read through.
Early on there’s an article on how the three-pointer has created a drop in the mid-range game in college basketball. It’s portrayed as a bad thing, with quotes from Rick Majerus about how to bring it back and a picture of a tombstone. But why is it bad? There’s no mention of why this is a smart thing for teams to do. There’s mention of how the three-point distance has changed over years, but that isn’t really the issue. If you move back a foot, the probability of making a shot drops maybe a percentage point or two. But the benefit goes up by the extra point for making a three, and that’s true regardless of where the line is. Unless you move the line back so far that three pointers are outrageous, ‘mid-range’ will always be worse than taking a three pointer; even then, the three would be better but simply impractical, like a half-court shot. Unless there’s a separate skill involved with the mid-range game that you don’t get from the three-point game, this doesn’t strike me as a big loss for the game.
The main story in the analytics section proper is about Brandon McCarthy, a pitcher who reinvented his pitching style to fit more with sabremetric principles and stay in the league. McCarthy’s primary pitch was getting him shelled in the big leagues, so he learned a new pitch or two and turned his game around. Baseball isn’t really my thing, so I had more questions than anything else. If you took any pitcher and gave him a new pitch, would he get dramatically better? It seems like simply changing your pitching pattern and having more options would be a benefit as long as you could throw the pitch well enough. How stable are pitchers in general? Is this kind of turnaround a rarity, or are pitchers fairly variable? Maybe McCarthy is just an extreme example and happened to also make changes to his game. But I thought it was a good article in general.
There’s an article about a bookie that isn’t much about analytics, but obviously is numbers-related. I thought the interesting part was that the guy in the article, who works completely alone and only has about 60 customers, makes six figures a year but still lives a pretty middle-class lifestyle. First, if that’s what one guy makes with a limited clientele, imagine what Vegas or a betting website pulls in! Second, it’s a good illustration of the main constraint of sports betting. The margins are thin enough that to make a reasonable amount of money you have to bet a lot of money, which means you have to have a lot of money on hand, but a lot of it is going to be locked up in bets at any given time. So even if you have a lot of money around, you need to live reasonably because you could need that money suddenly. On the betting side, there’s also the issue of simply being able to make bets. They mention that one of the dangers to the bookie is a “sharp” who has an analytical or informational advantage. He had a guy win a bunch of obscure bets and just stopped his account.
There’s an article on Kevin Garnett that really serves to talk about the growth of analytics in basketball. Despite being on a cruddy Timberwolves team for years, numbers showed that Garnett was one of, if not the, best players in the league. What I thought was more interesting was a box in the article rating various sports on how involved analytics is in each sport. Baseball is at one end and boxing at the other, which seems reasonable to me. Basketball is just below baseball. What surprised me is that mixed martial arts is smack in the middle. I know that numbers are tracked for MMA, like strikes landed per minute or takedowns per round, but is that something that fighters or trainers pay attention to? College football is down at the bad end, just above boxing and tennis, even though there are plenty of football stats around. This is because the coaches presumably aren’t paying attention. Is MMA really that much better? Or golf for that matter?
The Oklahoma City Thunder also got a good-sized article as a young team put together by Sam Presti, a GM with an analytical view. There are a few interesting points. One section is on how they want Kevin Durant to be more of a playmaker, and so his turnovers are increasing. Neither Durant nor his coach (Scott Brooks) seem concerned because those turnovers come from ‘good intentions’, as in he’s trying to make opportunities for him teammates. But there’s no mention of if this is a good approach. Is a ‘good’ turnover any different from a bad turnover? Do players who make good turnovers initially go on to reduce their turnovers and increase their assists, like the Thunder hope Durant will do?
The next section is on Nick Collison, who is largely unknown but is a plus/minus superstar by any flavor (straight, adjusted, or regularized). They quote Scott Brooks as saying that he likes plus/minus and is really happy with Collison. But Collison is the back-up power forward on the team, playing behind a couple different guys and never averaging more than 26 minutes a game since the team moved to OKC. If the coach and GM are so up on analytics and are aware that they say Collison is one of the best players in the league, why is he on the bench for almost half the game? Even if he was a Manu Ginobili type, asked to keep things going when a starter sits down, he should still get over 30 minutes a game. It’s very odd.
The next section, a defense of Russell Westbrook, is practically an illustration of how not to address a question with statistics. Concern 1: he has a bad attitude. Is there a number that gets at this? The author points out that Westbrook assists on over a third of Durant’s shots, but isn’t that his job? I thought the issue was more that Westbrook looks for his shot too much, at the expense of Durant’s shots. Maybe Westbrook should have more assists? Concern 2: Westbrook is not a point guard. I think this is related to what I was just mentioning. The first counterpoint is that the Thunder have a strong offensive efficiency in the last season and a half. What does that have to do with Westbrook’s point guard-ness? The second stat is more on-point; Westbrook is 7th in the league in assist percentage in the same time. But is that a measure of point guard-ness, or just having the ball a lot? Westbrook is just ahead of Derrick Rose, who’s often noted for shooting a lot and needing to be less of a point guard to make up for his teammates. The top of that list certainly has a lot of point guards on it, but it seems like a real point guard index would include shots taken somehow. Concern 3: Westbrook is not a winner. I’ve never heard this one personally, and it appears he has won a fair amount. Concern 4: Westbrook shoots too often. This is the worst one, in my opinion. They note that Westbrook has taken many more free throws than the next highest point guard on the list. Isn’t that an indication of taking a lot of shots, or at least having the ball a lot? He also leads point guards in offensive rebound percentage. What does this have to do with his shooting? Unless you want to argue that he turns those rebounds into quick put-backs, which they don’t. Westbrook is tied for second in the league in usage % in the same period this article discusses, which seems awfully high when you play with one of the most efficient volume shooters in the game.
The last Thunder section was on the trade of Jeff Green for Kendrick Perkins; it says that Presti was teary when they moved Green, “a cornerstone of his rebuilding process”. It’s nice to hear about a GM being emotionally involved in his team, but I don’t see why losing Green would be such a blow, especially to an analytical mind. For his career, Green has been consistently below-average on PER (so he doesn’t ‘look’ like a good player), floated around average on Win Shares, has been very below-average on RAPM, and the same for Wins Produced. In short, there’s no statistical sign that Green has ever done much to help any team he’s played for. Perkins doesn’t set the world on fire statistically either, but he was better than Green and above average on some measures in the few years before the trade. That seems like a no-brainer.
This is getting a bit long, so I’ll just note a couple other things that jumped out at me. Apparently the Dodgers have a guy who tries to determine if a player is likely to get injured in the future. That seems interesting to me; I wonder how wide-spread this is across sports. The new coach of the U. Miami basketball team has a drill where players scrimmage with a fixed number of balls, with a ball being taken away after a turnover. The spend the rest of the practice running. I like that drill. The section header “crazy crap analytics can prove” is dangerous at best. How often does analytics ‘prove’ anything? It certainly doesn’t prove that Andrew Luck will not turn into a star for the Colts. Basically, Luck played worse against good teams last year than against bad teams. Really? Show me someone who didn’t, and I’ll show you a small sample size. I never would have guessed there was a specific industry for insuring sports contests (like half-court shots for a car), and I would love to hear how they figure out the numbers on those things.
In general, I like that ESPN is bringing analytics to the public domain. But I wish there would be a bit more of a focus on good analytics. The basketball articles are littered with PER and APM, and the numbers aren’t always deployed in the best way. It’s frustrating to see that there are coaches aware of some of these numbers yet they still act counter to them. At the same time it’s tough to blame them when the popular numbers they see aren’t the best ones (or even good ones necessarily). ESPN could go a long way in advancing stats knowledge for coaches and fans if they tried to focus on some better measures or some general education. Their Total Quarterback Rating is a step in that direction, but you know what I just realized? It doesn’t show up in the magazine at all.