Why John Hollinger Is to the NBA Stats Revolution as Micheal Bay Is to MoviesApril 21, 2010
On June 1st, I will have worked for the NBA for one year.
It all began with hearing a podcast with Bill Simmons and Houston Rockets General Manager, Daryl Morey, the gist of which was about how the Rockets were using advanced metrics to evaluate basketball and how teams hid their stats to keep a competitive edge.
I'd read Moneyball when it came out, I'd read Bill James Baseball Prospectus, and the idea of expanding those ideas to basketball fascinated me.
Basketball was my favorite sport, while simultaneously being the thing I've sucked at more than anything I've ever tried. And here was a chance to really get into it in a way that could only grow.
First off, I read Dean Oliver's, Basketball on Paper. The main point of this book was to introduce and analyze the idea efficiency rankings.
Offensive Efficiency, Defensive Efficiency, Rebounding Rates, True Shooting Percentage, Effective Field Goal Percentage.
For example, the Allen Iverson debate. Is he good for his team? Well, yes and no. But not for the reasons you've thought.
Points are great, but points are a function of your team and your usage. As your usage increases, your efficiency falls.
During his heyday, Iverson was up there close to scoring leaders, which was amazing for a point guard. But when you look at his efficiency and his shooting percentages, he hogs the ball.
Iverson was one of the only players to be in the 30 percentile of usage on a winning team. Now you could argue that he needed to, that his surrounding cast couldn't carry the load. But when you look at the trends in his teammates performances, they were under utilized, and this was because Iverson was a poor point guard.
The book is great, you should read it. But there's a reason why the Denver Nuggets don't let Oliver write anymore books.
He doesn't look into these things anymore—these are numbers the teams looked at 10 years ago—so they're a terribly incomplete analysis.
When the New York Times ran an article about Shane Battier, we all got a peek at what the statistics teams were looking at.
Folders given to players before games filled with shot preferences, shooting zones broken up by efficiency, how they fluctuate based on defense, or the effects of making a player go right or left. None of these are open to the public. And we know these aren't all of the stats.
Basketball is an infinitely more dynamic and complex game than baseball. It's really nice to have a box score, it gives a decent summary of what happened in that game, but in terms of predicting what will happen, in describing why a team wins?
Did Stephon Marbury's numbers on a bad team really translate into a winning team? How about Eddy Curry's 15 rebound games on a Bulls team that squeaked into the playoffs in an incredibly weak East?
There are close to 100 possessions a game, give or take. There are always going to be so many points scored, so many rebounds grabbed. These are all functions of the larger team, and even then are incomplete in describing the team.
They tell you how a team does offensively and sometimes defensively, but they never tell you how or why. They don't let you predict or model what will happen in the future in the most accurate way.
So I need to address John Hollinger before I describe how basketball is actually looked at. Hollinger's Player Efficiency Ranking (PER) is the 'Transformers 2' of stats. It's analysis stripped of the complexities so that it can be easily absorbed without any threat of someone not being able to understand it.
PER takes things like points, shooting, minutes, assists, and rebounds and basically runs a giant regression for how they predict each other as well as wins. They predict how well a player fills up the box score.
Guess what? That means very little. It's the equivalent of saying, "Hey we added hits and walks, what would happen if we added stolen bases and fielding percentage too, then subtracted errors and strike outs?"
Do you get a number that might sum up what a player did during that game or stretch of games? Kinda, sorta. Is it in any way predictive of future performance? No. Not in any way.
Not statistically, not to the anti-stats guys (Golden State, Minnesota, and the Knicks have no stats specialists). It's an easily digestible number that ultimately means nothing. Hollinger making PER is no better than sports broadcasts still showing RBI.
You want to really look at advanced stats? Look at five-man units. Look at shot zones. Look at shot selection. Look at the effects of different five-man units on shot selection.
On how defenses contest shots or force players into open shot in a less efficient zone.
On blocked shots under the rim, a serious predictor in loss of athleticism in forwards and one day will be known as a 'Ron Artest' (well, sadly so many other things will earn that title first).
Say you want to take a player from a shitty team and add him to a great team. You want to know how the Sacramento Kings' Kevin Martin will be able to do at Houston when he's not taking all of the shots in a terrible offense. Well why not look at shooting percentage when coming off screens, shots that result from a pick and roll?
Hollinger ignores defensive stats for the most part. When the Oklahoma City "Zombies" played the Lakers last night, they needed to keep Nedard Kristic farther away from Andrew Bynum so that he could help on Pau Gasol if needed.
That means that Russell Westbrook needed to stick with Bynum when he drives into the paint. He didn't, defense was out of position, pass back to Gasol when everyone is out of position leads to the basket.
Look at the pick and roll. Best example to look at is with the Portland Trailblazers. When LaMarcus Aldridge and Andre Miller are in, Aldridge does a great job of seeing the pick being set and stepping out into the path of the pick, giving Miller enough time to come around and defend it.
Or, if the guard is still quick, it allows Aldridge to block the passing lane to the big man in the paint while Miller rotates to the taller man. Does this open the option of throwing it over Miller? Yeah, but efficiency sure does drop with that pass over when Batum doesn't step up as well and Jerryd Bayless has to guard the big man.
Teams keep track of these kind of things. They do because they're accurate indicators of what actually goes into winning. Sure you can look at points scored and allowed, but that doesn't tell you how a team wins.
Hollinger has a ton of interns and workers at ESPN to record this stuff for him. Hell, I've been doing it for a year and there are tons of countless forums with people suggesting new things, new ideas. Those people want to record these things and if you give them any incentive to, they will.
But Hollinger doesn't do those things. He doesn't spend all $160 million of the 'Transformers 2' budget making his own 3—D camera or investing in solid acting, good script writers, and picking out the most innovative effects.
He takes the most visible, easy to understand things he knows will work and he slaps an ESPN insider code in the url (easily bypassed by the way) to keep people from reading it.
In Moneyball, there's a quote from Bill James about how keeping stats from the fans is the equivalent of locking the doors to the game. John Hollinger means nothing to the basketball stats revolution because he's expanded no more than Micheal Bay has expanded the discussion on film as an art form.