Advanced Stats?

(bwoods note:  Of course True Hoop ran this really interesting piece on the Cuban Mavs' stat guru right after I posted)

As you probably saw, Prada turned up a provocative quote from the Ernie Grunfeld presser this week, on the subject of the team's use of "advanced stats."  Then came Flip on charting plays.  All of this begs the questions-what are teams doing with "advanced stats"?  And, more basically, what does that phrase even mean?

Now that you've made the jump, I can tell you that I really can't answer the first question, because of:

The Culture of Secrecy

Interest in and use of advanced stats has made a great deal of progress in both NBA fandom and in NBA organizations over the last decade.  It is, however, hard to pin down where the two intersect.  This is mostly because there is a culture of secrecy surrounding organizations' use of statistics. 

Teams have funded development of new data and of new ways of looking at old data, but, rather than publicizing what they find, they have (understandably) sought competitive advantage by keeping their findings private and proprietary.  They are essentially trade secrets.  Organizations seem to treat their numbers perhaps not all that differently from how Coca-Cola treats its secret formula.  (Just wondering, would it be easier to get the CEO of Coke to hand over the formula, or R.C. Buford to hand over whatever data it was that made him think the Jefferson trade was a good idea?)

Indeed, as Henry Abbott wrote yesterday:  "I've talked to many team stat people who don't know which other teams even have stat experts."

So, in this land of secrecy, exclusivity agreements, non-compete clauses, hit men (kidding, I think) and the like, we are left with limited information about who is doing what.  Sure, there is the occasional item like the Michael Lewis (of Moneyball fame) piece on Shane Battier and Daryl Morey, giving us a little, tiny peak at what a famously analytical GM and organization are interested in.  And Mark Cuban loves to seek any advantage, but is also happy enough to do some talking.  And there are basketball stat guys who have gone on to work with organizations, leaving one to assume that perhaps the work that person did before getting hired is something the organization might be interested in.  And, you never know when an organization might drop just a bit of self-serving information in justifying a trade.  But, on the whole, we're left without a lot of information.

The Wizards

Against that backdrop, it is not surprising that Ernie Grunfeld, a GM famous for both the short and long-form versions of answering questions without divulging a thing, wouldn't tell us what the team is up to, even if it were the most stat-invested team in the league.  Even so, I think it is the reference to using "some services" that probably caused Wiz fans (myself included) at least a little concern.  The concern is probably less because it means the team isn't doing anything and more because it suggests that the team isn't doing anything special or unique, or that it doesn't really take these things seriously.*

On the other hand, Flip's comments since coming on board suggest that regardless of how the personnel side of the team might be using stats, the coaching staff is certainly looking at them in evaluating strategy.  That's very good to know.  Designing a bazillion plays (is that the right number?) is of somewhat limited use if you don't keep track of what works and what doesn't.

So, we can be fairly sure that the coaching staff is on board even if we aren't so sure about the guys upstairs.** 

What do we even mean by "advanced stats"?

Ok, now that I've told you that little of importance is known as to what specific teams are really doing, let's run through what is out there and along the way I'll point out where we know things really are being used. 

I think, at this point, "advanced stats" can mean a lot of things to different people.***  Most basically, it seems to refer to anything other than the traditional box score stats.  I think of this broad range as generally falling within several major categories on the current basketball landscape:  (1) stats that are themselves byproducts of or new ways of looking at conventional, box score numbers (let's call these "composites" and "situational stats"); (2) player "stats" that require the collection and presentation of game data beyond that previously collected ("expanded stats"); (3) stats that are based on the team outcomes of what happens on the floor while individual or groups of players are on the floor ("the plus/minus family"); and (4) a catch-all of the murky universe of other observational stats ("designer stats").

Composites:  These have dominated the discussion of hoops stats online and I've used them heavily in prior posts myself.  They cover a lot of things.  Let's split them into two basic groups.  First are the things that are (most basically) just different ways of packaging the conventional stats, while attempting to take out the "noise" from things like playing time, pace of play, the different values of 2 and 3-pt shots etc.  These are stats like true shooting percentage, assist percentage, rebound percentage, and the use of per-36 minute measures.  Basketball Reference is a great place for sifting through these numbers.

The second group of composites are stats that try capture player effectiveness or value on a larger scale, by combining other box score or composite stats.  These are things like John Hollinger's PER, the various permutations of "win shares,"**** the use of "offensive ratings," etc.  There is too much ground to cover in this area for this week's column.  Suffice to say that there are many, many ways you can slice and dice the box score data that has been collected in the last thirty years, and there are a lot of people who are looking at creative ways of doing so.  It seems that all teams are aware of these at some point, but how much teams weigh them is another matter entirely.  We just don't know.

Related to the composites are "situational stats," which are perhaps still the things you're most likely to encounter in an NBA game broadcast.  These can get silly fast, as they are the basketball equivalent of the good old baseball "batting average in day game late inning situations after striking out earlier in the game" obscurisms.  I don't mean to dismiss them entirely.  It can undoubtedly be interesting to know how players really fare on meaningful, game ending possessions, and you can find other examples like that.  I have my doubts about whether front offices are using this group, but one never knows about coaches.

Expanded stats:  Calling these "advanced stats" is in some sense a reach, as many coaches have for a long time tracked their players' performance in areas beyond the box score.  These can be things as basic as tracking "hustle plays," or "hockey assists," etc.  They also might include things like "rotations made/missed" or "FG attempts defensed," or any other player-specific action that you can think to keep tally of.

The Plus/Minus Family: I believe Jeff Sagarin and Wayne Winston are generally recognized as having pioneered plus/minus stats in basketball.  The basic concept is simple:  normal counting and percentage stats can't begin to measure all of the impacts a player can have while on the floor, so instead look at how the player's team performs while the player is on the floor, off the floor, etc.  Roland Beech, of 82 games, has done much to bring plus/minus to the masses, and if you want to know more about plus/minus, that's where I would point you, particularly to the "Roland Ratings."  There are many variations within this theme, as there are many different ways of trying to isolate the impact of a particular player.  The best known organizational use of plus/minus is the Dallas Mavericks.  Mark Cuban has made statements that, in the extreme, seem to repudiate entirely the use of box score data in favor of proprietary plus/minuses.  The most featured stat in the Michael Lewis NYT piece mentioned earlier was plus-minus as used by the Rockets.  The simple version of in-game plus minus has made it into official NBA box scores.  This really seems to be where the money's at right now.  It will be very interesting to see whether one of the proprietary plus-minus systems will bust out into the mainstream and become a "conventional" player measurement.

Designer stats:  I'm probably throwing far too many things into this bucket and maybe I'll need to revisit and create more categories, but, for today's purposes, I'm using this to capture every other application of quantification or statistical analysis that doesn't fit in the above categories.  Everything from Flip's tracking of the outcomes of individual pages from his playbook, to Mark Cuban's charting of the refs, to studies of whether the "hot hand" really exists, can all go here.  Basically, if there is a question you can think to ask about player or team performance or game outcomes, there are ways you can (at least try to) frame it in terms answerable by observation and statistical analysis.  This has probably been the biggest breakthrough of numbers into the NBA. 

Basketball is an enormously, fascinatingly complex game.  It blends so many sets of skills, athletic abilities, and intelligence in so many ways.  Rather than sitting back and assuming that our eyes can tell us all we need to know to best answer questions about the game, more and more bright folks are trying to analyze outcomes to add to or challenge what our eyes tell us.  Every team has a different perspective about what should be challenged, what should be analyzed, or, perhaps, whether numbers can even be trusted.  And so it goes. 

*As I noted in a comment on Prada's Grunfeld post, EG's track record here is interesting. Miller is almost the archetype of what sort of experienced player certain classes of stat-head would want the team to pick up (positional needs aside).  A player deemed under-rated when measured by things like points per game or even PER as compared to certain "advanced" metrics.  That's obvious from WoW's recent love for him. On the other hand, this is the same organization that inflicted years of Etan Thomas's "rebounding" and "toughness" on us, because he looks like a rebounder and tries hard on defense.  So, Thomas aside, it appears that EG has not made any very big "mistakes" that could have been prevented by different use of stats.  Certainly not with the Wizards, anyway.

**Curiously, in baseball it seems the other way around--that advanced (different from situational) stats have been embraced by front offices (in picking players) more than managerial offices (in modifying strategy).  Given the difficulties in overhauling NBA rosters compared with MLB rosters, thanks to the small drafts, shallow player pools, and salary cap, it is not surprising that an NBA coach might find more immediate use for creative stats on a day-to-day basis than would an NBA GM.

***Ok, to me the potentially scary part from Prada's Grunfeld post was when EG immediately jumped to the dismissive mention of them as being tied to "fantasy."  First off, that's just a weird connection to make.  If advanced stats are becoming important to NBA fantasy leagues, someone needs to tell Yahoo.  Second, and more seriously, that could be read as suggesting limited understanding of what the universe of advanced stats has to offer.  Let's hope that's not the case.  I trust that it isn't, but can't help but wonder.

****Looking back, I think I've been overdoing it with the whole win shares thing.  I'm going to try to get away from that for a while, except I'll keep producing team projections that are based on them.