Happy New Years! A common tradition around this time is to do some self reflection and promise to better yourself. And that leads us to the theme of today’s post. In fact, lucky you, I have two! The first is an oldie but goodie: “What wins games?” and the second is “What to do when you’re wrong?”
What Wins Games? Starring Mark Cuban
At the start of the season we got into a disagreement with Mark Cuban about Chris Kaman. Our argument was that the most important part of winning games was getting good players and that traditionally Kaman has not been good.
The response from Cuban essentially boiled down to good coaching and good systems. In essence, some players in the right system could perform very well. And Chris Kaman and O.J. Mayo started out playing very well in the Mavericks’ system. In fact, some might have called it unsustainably well. And we’ve seen just that. Kaman has regressed to the mean and the Mavericks are in a free fall. And the nail in the coffin is none other than the Mavericks’ coach himself. Via the great Kelly Dwyer at Yahoo Sports:
I [have] to be inventive and find ways. I don’t have a better answer than that. The last week, I’ve had to literally scream in the face of two guys in practices and shoot arounds to get the point across. And I will continue to do that.
If I have to start suspending guys for not doing things they’re supposed to be doing on the court, I’ll do it. And Mark and I will get into it about that. But somehow, things have got to change and it can’t just be about that it’s a tough schedule. It just can’t.
Cuban argues that his coaches can make the players he is paying more productive. In other words, his coaches have the answers. And now we see from the head coach of the Mavericks that he really doesn’t have the answers.
Well, we could go back to the WoW Journal answer. You want to win in the NBA? Pay for productive players! When you hire unproductive players… well, you get to lose a lot. It really isn’t all that complicated.
Yes coaching, lineups, etc. matter. But these matter a heck of a lot less than getting good players. It’s no surprise that a traditionally terribly run franchise in New York has grabbed the best two players off the Mavericks and done well. It’s also no surprise that a traditionally well run franchise in Dallas is doing poorly after losing its top three players (Dirk Nowitzki to injury and age). Good players can do well in poor systems. However, a system, no matter how good, can not really force poor players into being good ones.
Now as fun as it is to point out when people are wrong, it’s hardly a unique trait. The simple fact is that anyone working on anything complicated will run into being wrong sooner or later. This brings me to my next point.
Changing Your Mind
Before the season started, we graded the Andre Drummond draft pick by the Pistons as poor. And since then, we’ve changed our tune a bit. In fact, the Wages of Wins, Pistons by the Numbers and the NBA Geek have all chanted that Drummond needs more minutes and is currently the Pistons best player! And this topic has come up in the comment section on the NBA Geek. Namely, shouldn’t we be more “humble” about being wrong? Patrick had a great reply to that:
This hasn’t got anything to do with humility. We were wrong, and we’re fine with that. We’ve been saying “we were obviously wrong” about Durmmond all season long. If we wanted everyone to forget that we were wrong about Drummond, we wouldn’t be bringing him up so much, would we? We thought Drummond was too risky. I still think, given the info at the time, that this was the correct analysis. But the information has changed. Detroit got lucky and the risk paid off.
So, we’ve changed our minds and admitted we were wrong in the face of new evidence. We’re all pretty Bayesian here. When you get new information, you re-evaluate. 600 minutes of professional ball (+ a bunch of pre-season minutes, which incidentally correlate well with regular season performance) is a lot of new evidence. Drummond will hardly be the last player we were wrong about (see Michael Beasley for an example of how we were wrong in the opposite way).
Detroit, however, is doing the opposite. They decided Drummond was “raw” and “needs lots of work” at the beginning of the season, and are and sticking stubbornly to that belief. despite all the emerging evidence that this is not true (or that it doesn’t matter — even if he’s raw, he’s still producing more wins than the “polished” guys on the roster).
Patrick uses the word wrong liberally to start. As we freely admit with rookies, they’re hard to predict. Even using a much better model than “conventional wisdom”, we’ve noticed we’ll miss more often than hit on predicting rookies (Arturo’s at about 45%) The point in grading a decision is to look at the information at the time and judge if the decision was wise.
But Patrick brings up the next point well. Even if you make a good (or bad) decision, you should update when new information becomes available. (Which is a point that Nate Silver, stats current golden boy, brings up over and over in his new book)
And I’ll be honest, this path is not always the easiest. As a writer on a blog that’s changed its mind on several things since I started, I can assure you that this often leaves you open to criticism. However, the goal of a basketball team is to win games and make money. And surprisingly, it turns out that winning games is what convinces fans to pay. (And the goal of doing statistical analysis is to gain useful information from the stats, not make everybody happy) And if that’s the goal, then criticism should not be what guides your actions (unless it is good criticism that can be applied or tested. See how easy it is to jump all over the place?)
The easy way to end this post would be to say “See, we were right!” Except, I can assure you, we’ll be wrong in the future about something. The bigger point I want people to take away from this is how to assess information. If your team is doing well, and it seems like your coach is the cause, well test that. And if the same coach doing poorly, while his old players do well on new teams isn’t enough? Well, I don’t know what to tell you. And the other key point is if you gain new information, even if it isn’t pleasant, make sure to update your model with it. Proving that you’re right 100% of the time is not a great strategy, especially in fields with uncertainty and moving targets.