Classes start this week at Cal-State Bakersfield. So I should have spent my day working on preparing either The Economics of Religion or Intermediate Macroeconomics (the two courses I teach this quarter). But I would rather comment on three items I saw on three of my favorite blogs.
These comments will be presented in three separate entries, with this entry focusing on a post from JC Bradbury — of Sabernomics and The Baseball Economist.
Last week Bradbury made the following observation:
“Triple Crown Stats on TV: Why are these the only stats listed for hitters when I watch a baseball game? Is it too much to include OBP and SLG? They contain more important information than the typical statistics. Even more annoying is when announcers refer to hitters as being good and bad based just on AVG. Just last night, I heard Joe Simpson refer to Brayan Pena- a .300 hitter in the minor leagues whose AVG ≈ OBP ≈ SLG-as a good hitter with poor defense. It’s inexcusable. You don’t have to be a stat-head to get this either.”
Let me try and offer an answer to Bradbury’s questions.
The statistically inclined sports fan is familiar with OPS, DIPS, and other sabermetric measures in baseball. In basketball we have NBA Efficiency, PERs, adjusted plus-minus, Win Score, and Wins Produced. For football there are the metrics of Football Outsider as well as QB Score. None of these, including the many measures I did not mention, are routinely mentioned in the media. Why?
Let’s divide sports fans into two groups: Those that are statistically inclined and the casual fan that is not aware – or does not care – about “advanced” (or even less advanced) statistical metrics.
Which group is bigger? My guess is that the number of casual fans far out-numbers the stat-heads. And given that the media is trying to reach the largest audience possible, we are not likely to see broadcasts of games reference any of the measures – and this includes Win Score and QB Score — we see around the Internet.
The disparity in population sizes, I think, explains why we still see baseball players evaluated in terms of batting average. And basketball players evaluated with points scored per game. It’s these simple stats the casual fan is most likely to know about and understand.
The lone exception to the focus on simple measures is the NFL’s QB Rating. As noted in The Wages of Wins, this is the most complicated measure commonly cited by the media. I think this exception exists because a) this measure was created by the NFL (specifically Don Smith, an executive with the Pro Football Hall of Fame, created this in 1971), b) it’s relatively old (again, it goes back more than 30 years), c) and there historically has not been any other metric that people have used to describe a quarterback’s performance. Consequently the NFL Quarterback Rating was introduced in a vacuum (a vacuum that led to its creation in the first place).
Perhaps this suggests that in time more complicated measures will be used in other sports. My sense, though, is that the circumstances that allowed the QB Rating to be adopted will not be repeated in baseball or basketball. And consequently, those few fans who hope that “better” measures will someday be the norm will probably continue to be frustrated.
– DJ
Here are the other two entries in this three-part series:
Darren Rovell Shows us that Michigan is a Winner
Stephen Dubner Notes the Power of Disgusting Advertising
Wages of Wins Journal 
Spergler
September 11, 2007
One key thing about the success of the quarterback rating is that it’s been given a binary interpretation. Most casual fans think of it as 100=good. I’m a pretty serious baseball fan, and I still don’t have a real good intuitive feel for what a good OPS is, for example.
dberri
September 11, 2007
Spergler,
Excellent point. Fans do need to know what “good” and “bad” is with a number, and that is not clear with OPS (and many of the other stats I noted).
Jason
September 11, 2007
I’ve noticed that at least in SF, the scoreboard stats for each player now reports on-base percentage for hitters and batting average for batters faced for pitchers.
My guess is that it’s mostly historic as to why certain stats are still reported over all others. It’s true that batting average and points per game are closer to intuitive, but my own suspicions are that most people aren’t really thinking much about how these numbers are generated either. It’s the context of knowing what “good” values are that dominates the landscape.
In this sense, it’s much like the problems of introducing the metric system in the US. I use it daily in my work, calculating picograms present and microliters necessary for reactions, but someone’s height in meters or centimeters doesn’t immediately suggest anything to me, nor do I know how to dress when someone tells me that temps will be in the 20s. The intuitive nature is key.
It can be learned though. I’ve got a good idea of how far something is in kilometers, probably from being a runner where race distances are usually presented with the metric system: though I still prefer speeds miles per hour, it’s now a fast conversion for me, more like being conversational in a foreign language rather than using a dictionary to translate.
Matthew
September 11, 2007
I’m a “stathead”, but I don’t think any of the advanced basketball metrics have earned the privilege of being cited widely yet. If we have several metrics attempted to quantify the same thing (in the case of the ones you mention, the whole value of a player), and they come to different — and in some cases vastly different — conclusions, why should they be cited?
To me we’re now in a period of uncertainty on these metrics. We’re still working out the kinks. It’s like cosmology in the 1950s where you had steady state vs. the big bang. It’s best if you hold off on citing these stats in widely read publications until we have a better handle on which one is right. So far WP/48 is leading the pack due to its predictive power.
Owen
September 11, 2007
Matthew – I think Win Score would make a great in-game stat. And I think True Shooting Percentage, though not by itself an advanced basketball metric, would make a fantastic substitute for fg%, I would vote, if win score doesn’t make, then ts%, points scored, rebounds. I suppose one problem, given how often stats get shown when players are the foul line, is that it sort of overlaps with ft%. But adding TS% would be a big improvement over the status quo.
Nate
September 11, 2007
I actually think there’s some hope as far as baseball stats; I’ve noticed OBP alongside the triple-crown stats on some baseball broadcasts this year (WGN?)
As far as familiarizing fans with the significance of a given player’s OPS, they could color-code it: blue for poor (say, below .700), yellow for near-average, orange for good (maybe .800-900), and red above that.
dberri
September 11, 2007
The other night I heard Al Michaels, in discussing the QB Rating, note that 100 is considered “good.” It is interesting that he thought it was necessary to tell his audience, which consists of football fans, how to interpret QB Rating. Again, they are broadcasting to a general audience.
As for the basketball metrics, surprisingly PERs and NBA Efficiency (per minute) come to very much the same conclusions. Just thought I would throw that out there.
Owen
September 11, 2007
Hmm, very interesting post and links on True Hoop re PER. It’s the eighth item in the Tuesday Bullets.
dberri
September 11, 2007
Owen,
Saw the link on PERs. Interesting what the comments focused upon (and what they ignored).
Matthew
September 11, 2007
Did you see Hollinger’s PER projections for next year? Allen Iverson is projected to be the 10th best PER, for example. The metrics are quite different, PER values volume scoring a lot more than wins produced; wins produced values rebounding a bit more. I don’t see how the media could report one (or both) of these metrics without confusing people. Again, the steady state vs. big bang analogy works here.
I definitely agree about ts%, though. At the very least they should be reporting efg% if they only want to talk about shooting efficiency.
Owen
September 11, 2007
Having processed all those blog posts and comment threads, I have to say I am very disappointed with the level of discussion. Are we really arguing about whether per-minute stats work better than per game stats? Is that what pasts for astute, up to date basketball blogging? And the WOW hardly gets any play in this conversation, as if there wasn’t a stat system that rates Lamar Odom as being a better player than Earl Boykins.
dberri
September 11, 2007
Not sure how you get links to work in the comments, but this link was interesting.
http://prosportsdaily.com/forums/showthread.php?t=142980
Here was one of the comments (the post is about which productivity model people like). “The only one that is really worth mentionning is PER, the NBA efficiency is a joke and the wins produced is terribly flawed. PER is useless without backround information though, so it cant be looked at as a holy grail for NBA fans to judge players by.”
I would again make this observation: PER and NBA Efficiency really tell very similar stories. So it is odd that someone would refer to PER as a model that is “worth mentioning” while also stating that NBA Efficiency is “a joke.”
Is it possible to agree (given the point I made last year about inefficient scorers) that PER and NBA Efficiency both have a problem (actually the same problem) without commenting on the validity of anything else (Wins Produced, adjusted plus-minus, etc…)?
Matthew
September 13, 2007
I think your wp48 and adjusted plus/minus have the same problem, in terms of exposure. Neither are easily accessible. The Jason Chandler website is the only one I know of that even has wins produced stats at all (and it only has wp48), and that’s just for the current season.
Adjusted +/- is even more difficult to track down. If we’re thinking of the same thing, I’m think of Dan Rosenbaum’s work, which I don’t think has been updated in a couple of years.
But I do think these methods are both superior to the more widely talked about EFF, PER, and unadjusted +/-.