Thursday, May 11, 2006
New Statistics I Like to See
"Moneyball" (that's code for "a new generation of statistical analysis as described in Michael Lewis's baseball book Moneyball)is coming to the NBA in a big way. Dean Oliver has been making waves in Seattle. Dan Rosenbaum has a job in the NBA now. Daryl Morey is going from advising the Celtics and teaching basketball statistics at MIT to running the Houston Rockets. 82games.com is regularly featured on SI.com. SI has also described the whole trend. The list goes on and on.
People are suspicious of Moneyball in all sports. All these geeks with their charts and their computers coming and telling the old-timers what works. As if!
What's so great about statistics anyway? Statistics don't put the ball in the hoop at the end of the game.
But I love it. I'm no expert in it, as I have proven again and again on TrueHoop, but it makes perfect sense to me that this is an area where basketball can develop for the better.
We All Use Basketball Statistics Already, and Most of them Suck
Here's why I say that: even the oldest of old-timers are relying on statistics already. Points and rebounds are so central to the game it's sometimes hard to remember that they're statistics. But they are, and they give you a certain limited view of what happens in a game. But if a kid scores 50 points, or gets 25 rebounds in college, you better believe that'll increase the chances one of those scouts will make the trip to go see him play. (Dajuan Wagner put himself on the basketball map with statistics from one game: his 100 points in a high school game.)
We are using statistics to measure basketball quality, and we're doing a clumsy job of it. High scorers are still high earners. But the object of the game is winning, not scoring a lot (right, Starbury?). Getting players who get a lot of points and rebounds does little to build you a winning basketball team. Can you imagine getting all the highest scorers and rebounders on one team? Hello failed Team USA of the last few years. The fact is that there are lots of chores that are important to basketball that are not measured by points and rebounds. (Like defense for instance.) You only have to look at the Detroit Pistons to know that. Guys like Tayshaun Prince win a lot more games than guys like Stephon Marbury but our old-school basketball stats don't have a clue as to why.
One of the big points of Moneyball is that on base percentage, long the forgotten step-child of baseball stats, correlates with wins a lot more than the long-trusted ERA or batting averages. And errors do a terrible job of measuring a player's defensive ability, because players with great range on defense can almost get a ton of balls--and they make a lot of plays. Slow fielders with little range can only get balls hit directly at them. But the former player typically gets more errors, because so many balls come within his range, and he's trying for everything.
It is Early, and That's the Fun Part
It's a messy process of transition. I suspect we are a 20 or 30 year period at the end of which there will be new standards for basketball stats. People like Dean Oliver, Dan Rosenbaum, Kevin Pelton, John Hollinger, and several others are essentially experimenting at this point, and some of what they come up with will really prove, over time, to contribute to a better understanding of basketball and more wins. And we're fairly close. (The seeds of a lot of them are outlined on the "New Visitor's " page of 82games.com.) In the interim, the crusty old-timers will seethe, and the nerds will be learning the ropes. At some point, the two groups will see eye to eye all over the league, and we'll finally be using statistics in a smarter way.
But for now? We're in that transition period, and there's opportunity for teams that get it. There are also opportunities for some very interesting conversations and ideas. I have ideas for new stats I'd like to see all the time.
A Time for Ideas
For instance, I talked to Dan Rosenbaum a few years ago for some article. We got to chatting about how old-school stats don't measure so many things that are important. I trotted out my theory that one of the important ones to track, if you could, would be how "sticky" a player is. By that I mean, if a ball passes near him anywhere on the court--a loose ball, a pass, a rebound, a dribbler, anything--what are the chances that player comes up with the ball? Playing basketball through the years I have noticed that if there's a player on the court with incredibly sticky hands, his or her team wins more than their fair share of the time. Not only do they come up with loose balls, steals, rebounds, etc., but just as importantly, if there's a guy like that on the floor, the offense naturally stays away. It's a way to own a whole bunch of territory, and force opponents to operate in a limited space. It's game-changing.
What's so exciting to me is that while that's something that was long considered impossible to measure, in this age of digital technology, there really might soon be a good way to measure and rank the "stickiest" players in the league. If I were choosing players for my team, I'd like to know that. Dan seemed interested enough. Maybe someone will check it out for us.
Another statistic that occured to me as I was playing last night: I would love to know what percentage of possessions a given team has a scorer wide open in prime scoring position who does not get the ball. For instance, the other day I saw a point guard make his way through three defenders and then make a tough layup. It was a great feat of skill and athleticism. But the whole time his power forward was standing right there under the hoop, all alone, with a nice passing lane. The point guard took a shot that was probably 40% likely to go in, while passing up a 99% likely to succeed play. In our old stats, the PG was perfect. In this one, he'd have demonstrated a minor flaw that over time could lead to many fewer points.
This would be so handy! For example, I'm sure the Pacers would like to know this off-season whether or not they should be giving Anthony Johnson the ball and telling him to go be an All-Star next year. There's no easy way to know. Was his performance in the playoffs proof that he can be a winner as a high-scoring point guard? One thing that would be good to know is... in that game when he scored 40, did he have guys wide open in scoring position who didn't get the ball five times? 15 times? That number would help you to know whether or not he's making the best decisions for the team. You need to know if the Pacers are better off with him trying to be a 25 points and 5 assists guy, or a 9 points and 11 assists guy, and this could help you figure that out. (Or indeed, perhaps you'd decide instead to give his roster spot to Dajuan Wagner.)
It's Going to Be Good
Be patient, ye skeptics. This will be good. It might not be perfect now, but soon this new breed of analysis will absolutely change the game for the better. Maybe there will be fewer wasted draft picks, more value on team-oriented players, and new wrinkles in coaching strategies. I'm looking forward to