Michael Lewis' 2003 book "Moneyball" opened the eyes of many to the potential benefits of analyzing baseball by more than simple scouting reports and statistics such as batting average, HRs, RBI, ERA and wins for a pitcher. In the years since, advanced statistics have revolutionized the way MLB front offices value player traits and production, as followers of Oakland A's general manager Billy Beane have spread throughout the sport.
The NBA followed suit, and advanced stats such as win shares, on-court/off-court, offensive rating, defensive rating and rebounding % now influence the decisions of NBA front offices in the same way. No longer are we bound by the basic stats of points per game, rebounds per game and assists per game, all of which are limited by lack of consideration to opportunity and efficiency of production.
Identifying efficient performers on the court is one of the goals of advanced stats analysis, but the larger goal is efficient use of resources by a team. While baseball has no official salary cap (but does have a de facto cap based on a team's revenue), the NBA does and it became a much harder cap with the ratification of the most recent collective bargaining agreement. This makes accurate valuation of players even more important to a franchise's chances for sustainable success.
We here at BSotS have dabbled in advanced stats analysis; we have some experts, some who are put off by it and some curious novices. Our SBN community has somewhat of a guru in the field in Golden State of Mind's EvanZ, who you might remember from a chat we had last summer. Evan participated in another Google chat with me recently as I picked his brain on some of the advanced data he analyzes when judging player effectiveness.
More after the jump.
Ray: When did you start to analyze NBA basketball using more advanced stats than simple PPG, RPG, etc?Evan: Fairly recently...I think it was spring of 2010.
Ray: What caused you to start?
Evan: Actually, you might not believe this, but it was some time after I joined GSoM. In the beginning I was one of the "anti-stat" guys, but somehow I became convinced after losing countless arguments. After that I was "all in" and it grew from there.
Ray: How is it received at GSoM? Do you still have a wide divide between the stat people and the "anti-stat" people?
Evan: I think it's definitely acknowledged that it's a stat-heavy blog, but there are certainly a minority of folks who don't buy it. They tolerate us though.
Ray: How does analysis of advanced statistical data enhance your understanding of basketball?
Evan: I think it helps you understand "the bottom line" so to speak, what teams need to do to win, vs. what looks good on ESPN or something.
Ray: Does it change the way you watch games?
Evan: Yes, definitely. For example, I cringe when I see "bad shots" that players shouldn't take.
Ray: Haha. Try watching Shannon Brown. Yikes.
Evan: He looks good dunking
Ray: What are some of the data points you find most useful?
Evan: Really...efficiency, that's the number one thing.
Ray: As defined by what?
Evan: Points per possession (PPP). It can be applied to many things, though: at the team level, the player level, the type of play, etc
Ray: Basketball Reference, 82games.com and Synergy Sports are the websites that immediately come to mind with regards to advanced NBA stats. What others have you found to be helpful?
Evan: Hoopdata is the other big one. Oh, and I use Basketball Value a lot for my own metric. It's called "ezPM". There's a brief description of ezPM in "The Primer" (Ray's note: the primer is excellent and I highly recommend it), but basically it puts a value on every possession that a player uses and creates (through steals and rebounds). There's a couple of metrics in the primer that I developed. One is ezPM and the other is PSAMS (position- and shot-adjusted marginal scoring). I also develop metrics here and there when I find something missing in the "field".
Evan: And one more resource...
NBA Appspot
That's where I get the RAPM (regularized adjusted plus/minus) data
Ray: Can you give an example of a player who, according to advanced statistical data, is overrated by most fans?
Evan: Uh, Monta Ellis comes to mind.
(Ray's note: Yes, I am a big Ellis critic. No, I did not pay Evan off for this.)
Evan: Last season, a lot of advanced stats folks thought Rose was overrated - he's actually peforming better statistically so far this season (according to ezPM, anyway). John Wall appears to be highly overrated.
Ray: Underrated?
Evan: A lot of defensive guys like Nick Collison, Luol Deng, Andre Iguodala
Ray: Do you think the existing defensive metrics truly capture a player's value on that end?
Evan: I think RAPM generally does the best job. Unfortunately, there's just not a lot of data aside from blocks and defensive rebounds (which aren't really "defense" per se).
Ray: Obligatory question about the Suns and Warriors: for each team, what player does advanced stats analysis show to be better than we might think? Not as good as we think?
Evan: On the Warriors, it's easily Ekpe Udoh. RAPM has him as an above average player, but according to the box score, he's a zero. But anyone who watches him play see what he brings defensively. That's not captured by the box score. For the Suns, aside from Steve Nash possibly being underrated (I know hard to believe), I'm not sure you have any other players who are that much better than what people think. Markieff looks goooood, BTW, so good on you guys. (Ray's note: Nash is currently rated #6 overall in the league in RAPM.)
Ray: We are VERY excited about Markieff Morris so far.
Ray: Do you see a risk in losing the human element of the game by breaking it down into cold, rational metrics?
Evan: Well, basketball is way behind baseball, so I would ask whether baseball still has scouts? I think so. And that probably means there is still going to be a place for that in basketball for quite some time. Numbers can tell you a lot, but not everything.
Ray: What limitations do you find in statistical analysis of basketball?
Evan: The main limitations all basically boil down to not having enough data. I'll give one obvious example, which would be what I call "potential assists". The box score records an assist when a player scores, but what about when he doesn't score, but should have? Those are not recorded and thus, represent a data limitation. There are many more examples on both sides of the ball. (Ray's note: how many more assists would Steve Nash have this season if he was surrounded by better shooters than he currently is?)
Evan: If folks are interested in ezPM, I am keeping those continually updated on my blog.
Brightsiders, what say you? What value and what limitations do you see in advanced stats? I encourage you to follow the links above; there is plenty more to read.
Please feel free to fire away with your thoughts and questions for Evan, and I'd like to give him a big thanks for his time and expertise.
For more exclusive content, follow us on Twitter @Brightsidesun and "Like" us on Facebook.