Rose vs Howard and the Inescapability of Narrative
In a very strange year where every candidate has something “wrong” with them, we’re still debating the MVP race with only a few games left. The race has narrowed around the star of the golden team (Derrick Rose) and the one MVP candidate from last year who can say he’s improved (Dwight Howard).
Opinions tend to be with the mainstream guys going for Rose and the heavy blogger thinker guys on the internet going for Howard. (And let’s note that the poll on the APBRmetrics board sides with LeBron.) I’m personally on record as being on the fence between Rose and Howard, so what gets me ornery enough to write a piece is someone being dead set that even sitting on the fence is wrong.
Enter John Hollinger. With him, it seems I’m guaranteed to find both compelling arguments, and some stuff I disagree with enough that I end up writing about it. (Unfortunate that I end up writing about the “bad” because I really do respect the guy, and I absolutely use his stats.) He’s now written Truth about the Derrick Rose Story, trying to take down Rose, and favoring Howard, so it’s on.
Turn Me On, Dead Man
Hollinger’s first two paragraphs make clear his thesis:
What bothers me so much about this year’s MVP coronation of “The Derrick Rose Story” is not so much that it’s a mistake — we’ve had bad award votes before and will have them again — but that it’s the same mistake, for the fifth time in 11 years.
This is an inherent risk in the MVP selection process. When you ask people whose life’s work is to seek out and tell great stories to vote on this award, we shouldn’t be surprised when they turn out to vote for the best story rather than the player who is most valuable.
Now, John actually means this quite literally. He goes on to talk about how perimeter players win the MVP because people like cheering for David over Goliath. I’m going to ignore that because I find that point distracting from the more powerful cousin to this argument which is that people seek out a simple causal narrative to explain large consequences, even where no simple cause exists.
Over at Back Picks, ElGee sums up the proposed Rose narrative succinctly in an article where he quite astutely berates MVP voters:
The Bulls are in first despite Carlos Boozer missing a lot of time. The Bulls are in first despite Joakim Noah missing a lot of time. The Bulls have no other noticeable stars. Derrick Rose is MVP-worthy!
I don’t think there’s any doubt that there is a tendency for people to infer a player’s value by working from the top down like this, and obviously it can be quite dangerous to do so. One ends up with the possibility of wildly re-calibrating a player’s value every time there’s an injury on his team – which basically means you can never really be certain of anything, and are particularly vulnerable when evaluating that hasn’t been together for that long. Have the Bulls been together long enough? Uh, well, if you thought along these lines, wouldn’t you really like to see how the Bulls do without Rose for a while?
Tyranny of Reason
Far better to work from the bottom up, right? Just spend time using all your energy to assessing every nuance of a player’s game and go from there -what could go wrong? Well, funny story – quite a lot actually.
Let’s go back to John. He gets into the idea that the proof of the “story” aspect of voting is in how certain players do far better in the MVP voting in some seasons than others. Taken as a general principle, I wouldn’t disagree with this. Take a look at this specific thought though:
Nash is an equally strong example. As everyone knows, he won the MVP
in 2005, sporting a player efficiency rating of 22.04 while joining with
a dominant power forward to lead a 62-win team. What few people realize
is that two years earlier, he had teamed up with a dominant power
forward to win 60 games and tie for the best record in the West; he had
a better PER that season (23.51) and played more minutes. For his
efforts he received one fifth-place vote.
The difference between those seasons, obviously, was that in 2004-05
Nash was a great story, because he had just joined a 29-win team that
surprisingly rose to first in the West. The 2002-03 Mavericks were
already good, so his performance there was deemed a minor event.
My lord, dripping with assumptions. Because Steve Nash didn’t improve in PER (the stat Hollinger made up), he couldn’t possibly have actually increased his value to his new club by leaps and bounds. No mention at all that Nash played a very different role in Phoenix in a very different offense than was in Dallas. No mention that Dallas saw no major drop off without Nash, but that Phoenix for the past 7 years has seen a glaring falloff without Nash in the game. Just a black & white mindset based on the set of statistics scorekeepers have tracked for decades.
And you know what’s funny about that? The problem with John’s thinking can be described in the exact same wording we used to describe the voter issue above: Subscribing to an overly simplistic narrative to explain consequences.
Now some of you may object to that. I skipped ahead a couple steps so let me go back. Let’s first ask ourselves: What is a narrative? Well there are multiple senses of the word, but when we use it to describe what happened, what we’re really doing is describing why or how something happened (as well as the other interrogative pronouns). That means we’re stating the causality of the situation.
When voters (or whoever) use team wins and injuries to allocate credit, they are assessing who caused the team’s success by indirect means.
Is Hollinger’s method truly any different? When it comes to telling stories or not telling stories, the only difference is that he’s possibly more nuanced in his methods. His formulas factor in rebounds, blocks, turnovers, etc – do they state definitively who the most valuable player is? Certainly not. They are merely a recording of some good and bad things that a player did during the game. By no means do they encompass every effect the player had on the game’s result, and by no means do they represent a statement of how much better the team did than they would have done without the player. All Hollinger is doing with his stats is trying a different approach allocate credit by using detailed numerical correlations to inform his conclusions.
Now, while I’ve been considering my MVP choice, I’ll level with you that the cases I’ve read for Howard have been more impressive in general than the cases I’ve read for Rose. Don’t think I’m dismissing all arguments as equal because they are all imperfect, or that I’m more against Hollinger’s side because this article is directed at him instead of Ric Bucher. The reality is that it’s just more clear what I object to about on the more analytical side – I’m not as sure where to even begin when people brush aside questions with an appeal to authority a la “the access I have, and have had for nearly 20 years, informs my opinion. You don’t have that access”. In short, I expect the men of reason to be quite reasonable, and chafe when I feel they aren’t.
Stand back! I’m going to try science!
Let me liken this to the world of science. There are a couple well known sayings scientists use:
1) Correlation does not imply causation
2) Theories can not be proven true, only false. The best you can ever say about a theory is that it is “not yet false”.
The root behind both of these is that all an experiment ever does is test correlation – and correlation on its own is useless. One can only use principles that actually cause things to happen, not simply tend to correlate with their occurrence. Scientists thus must supply causal narrative in the form of theories and hypotheses to have it be something that’s worthwhile, but in doing so they run the risk of running wild with something that was just coincidental.
It can be maddening, and mistakes are made – but on the whole, I’d say science has been pretty successful wouldn’t you?
The use of narrative to infer player value is inescapable
Back to basketball now: Folks, if science can’t work without narrative construction, basketball analysts aren’t going to be able to get away from it either. As with most types of analysis, the most beneficial way to be is to recognize that there’s no way to get this stuff perfect, and thus to be constantly seeking to refine your methods and reassess your conclusions.
This should involve statistics of several kinds because you are human being who can’t watch every minute of every game, and even if you could you aren’t mentally capable enough of processing that information with the right proportions.
This should involve watching games and talking to other (hopefully more knowledgeable) people who watch games to come to your conclusions.
This should involve going back and forth between the two with questions and hypotheses.
And last but not least, this should involve a general outlook that you may be wrong, and that that would be a good thing because you’d learn something.