View full size photo »

A Daily Babble Production

The last two months have seen a confluence of factors bring the discussion of statistical evaluative tools closer to the forefront of the conscience of basketball's mainstream following.

Back in February, Michael Lewis published the "No-Stats All-Star" in the New York Times, utilizing the Rockets' Shane Battier and Daryl Morey as catalysts for a discussion about the need for statistics more telling than those found in a traditional box score.  That, along with March's more-popular-than-ever MIT Sloan Sports Analytics Conference (check out Kevin Pelton's recap here) reinvigorated a discussion that has long been raging in the blogosphere centered around the value of statistical analysis and questions of what can and can't be measured on the basketball court. 

Two weeks later, the folks at The Big Lead said they weren't buying the metrics that have come to be known as advanced stats.  Tom Ziller responded with a pro-stats post that led to a debate at Sactown Royalty far more spirited than the Kings' defense has been at any point this season.  TrueHoop's Henry Abbott made a succinct point about the value of seeking the right stats and then chronicled a pertinent disagreement between Knickerblogger and David Friedman at 20 Second Timeout over the improvement or lack thereof of the D'Antoni Knicks.  Ziller then went back for more with his StR readership.

All that said, there is no shortage of recent reading material available written from a variety of perspectives.  Those links are all just from the last eight weeks.  A simple Google search for "APBR metrics" or "advanced basketball stats" will yield plenty more, both old and new.

That brings us to today's topic.  As the home of my various hoops-related ruminations, this Daily Babble space plays host to a wide range of statistical references, many of which are made without further elaboration.  I owe it to each of you faithful readers to provide an explanation of how statistical analysis is approached at Babble Central.

No matter what statistics one uses, they are a complementary tool to observation, not a replacement.  Paper exists in two dimensions, and basketball is played in three. 

I've long said that I'm fortunate that my archives at my old site were wiped out because of all my rambling postulations, two or three were exceptionally dumb.  One of the silliest things I wrote - an ill-informed diatribe about Kevin Martin for which Tom Ziller rightfully killed me - came courtesy of not watching enough and using the wrong statistics.  At the time, I thought I had taken enough time to observe Martin's game, and I simply hadn't.  It was an error in judgment but ultimately a good learning experience for someone still becoming familiar with the responsibility that comes with offering his thoughts to a viewing public. 

I present my writings to you here on a daily basis as a product of my own observations based on as significant a sample size as I can find, my discussions with and readings of writers, watchers, players, coaches and front office members whose insight I respect, as well as pertinent statistics that relate to what I've seen.  Since the eyes do deceive sometimes, when the statistics don't match up with observation, it seems only right to take the time to explore that discrepancy rather than potentially postulating inaccuracies to you. 

The eyes and the stat books serve as good checks for each other rather than mutually exclusive entities.  I've encountered the viewpoint in discussions with friends of mine that people either watch basketball or sit at their computers watching numbers but not both.  That's simply not true.  Without a doubt, there's a danger of becoming too invested in the paper and not paying enough attention to the nuances of the game that lead to success and failure (we don't have specific measures for screening, for how teams play screen-and-rolls or how well a player moves his feet defensively), just as there is of deciding that Mikki Moore is a good shot-blocker because he is tall and you saw him get a piece of the ball once. 

But the people who are worth listening to aren't at either of those extremes.  Kevin Pelton is brilliant.  He's become one of my favorite hoops writers because of his ability to perform in-depth statistical analysis (like this study of close game performance) while also providing exacting insights on the technique of the players he watches (check out this breakdown of Blake Griffin's performance against North Carolina in the Elite Eight).  Ziller watches every Kings game, understands the ins and outs of their offensive and defensive schemes and presents the numbers to back his assessment of their play (not to mention that he's pretty good on the rest of the league as well over at FanHouse).  Those are only two that come to mind quickly on a long list of like writers.  The point, in short: There is an attainable happy medium of both understanding the way basketball is played and the way it is measured.

Along with the issue of general statistics use comes that of using the right statistics.  The money quote from the Lewis article cited above is Rockets GM Daryl Morey suggesting the creator of the box score be shot.  Just because we have used certain performance measures for a long time doesn't mean we can't improve them.  For example, blocks and steals are not, repeat not, grounds for definitive assessments of the totality of a player's defensive contribution.  Over a period of several years, the APBRmetrics community (named for the Association for Professional Basketball Research) has made great strides in making those improvements.

While volume stats still have a place (there is something to be said for scoring an obscene amount of points in a game or grabbing a lot of rebounds), they aren't the be-all, end-all by any means.  Since every team plays at a different pace and not all players receive the same amount of playing time, it seems fair to normalize for those factors when we evaluate teams and players.  Because of that, it makes a lot of sense to me to use per-possession measures rather than per-game measures when possible.  A game is an artificial concept, whereas a possession is one time down the court with a chance to score. 

Since both teams get roughly the same number of possessions in a game they play against each other, one can predict more accurately how they will compare on offense and defense based on how effectively each uses its possessions.  The same goes for rebounding, turnovers and a host of other parts of the game because possessions are opportunities that are similarly defined for everyone whereas games are not.  Similarly, while per-minute stats should be used with caution (projecting out for a guy who plays a minute a game against LeBron James' figures is idiotic), they also help to normalize for players who receive different opportunities as do certain stats (such as rebound rate) that standardize for pace.

On another note, it bears noting that basketball is far from an individual game.  The ultimate goal is to win basketball games as a team, and a player's individual stat line often doesn't tell the whole story of how well he interacts (in an on-court sense) with his teammates to achieve that goal.  That has led a variety of efforts to quantify a player's value to his team with a single number, such as adjusted plus-minus, win shares, Wins Above Replacement Player, Wages of Wins and on/off court ratings.  John Hollinger's Player Efficiency Rating also attempts to combine a player's contribution into one number to help indicate his value to his team.

On a personal note, I'm still not sure how necessary I feel it is to devise a statistic that even approaches serving as one tell-all of a player's performance and value to his team.  As the proponents of many of those stats attest, they realize that these stats aren't actually tell-alls.  For instance, Hollinger knows that PER doesn't do enough to recognize defensive contribution (because, again, defense goes waaaaaaaaay beyond blocks and steals figures) and will remind you that it is just part of the equation.  But I would like to think we're also capable of looking at the measurable facets of a player's game - scoring efficacy, success on the boards, ability to take care of the basketball, how much he dominates the offense, to name several - understanding that they aren't all of equal importance for each player and assessing those players based on a set of different measures and observations.

But, and this is key so far as what you see in the Babble is concerned, I'm not qualified to say a word more in that previous paragraph because I don't understand those metrics well enough yet.  I jumped on the advanced stats bandwagon late, starting at the beginning of last season, and I've still got a ton of work to do to catch up.  I'm currently in the middle of Dean Oliver's landmark Basketball On Paper, and I've got a ton of reading after that to do on PER, adjusted plus-minus, Wages of Wins, win shares, value added, composite score and a host of other statistics that could well be useful evaluative tools.

There are certain statistical measures that you don't see in this space and won't for the immediate future, not because I don't like those stats so much as it is that I don't understand them.  Adjusted plus-minus sounds like a great idea to me - measuring how a team does with a player on the floor while controlling for the quality of his teammates and opponents.  But I have no idea how someone could effectively control those factors, and I haven't taken adequate time yet to acquaint myself with any of the several forms of APM available on the 'Net.  Until I do, I don't feel comfortable using it.  StR reader Viliphied is right when he says that not understanding a stat isn't a valid argument against that stat.  As someone who relays my opinions on the game to you every day, it is my job here to be understand those stats, and I promise you that I am continuously devoting my time to gain a better understanding of the ones I don't grasp yet.

Until then, you will continue to see what you already have: statistics that I feel I grasp well enough to explain their meanings to any of you who may be less familiar with those measures.  I refuse to throw out numbers that don't represent something I can explain to anyone who questions them.  Since our readership comes from a variety of backgrounds and likely has a wide scope of familiarity levels with the stats, tomorrow I will post a primer of the measurements commonly used here at the Daily Babble as well as a list of resource links for my details on APBR stats.  As I continue my learning about the metrics mentioned above and come to conclusions about which I trust enough to use in my writing, I will add those to the list.

We aren't able to measure everything that happens on a basketball court at this point.  Perhaps we never will be, and that's not necessarily a bad thing.  But it's worth the effort to look for measures that do more justice than the old standbys.  In this small corner of the blogosphere, we'll continue to stress the side-by-side value of observation and statistical evidence, the use of measures that one understands and the value of continued learning statistics that provide more accurate evaluations of the players and teams who compete than the box score does.