plus/minus stat for basketball

[Wareagle]

I was just wondering if anyone here knew if a website tracked a player's plus/minus like they do in hockey.  Not saying it would be better than Ken Pom's efficiency stats and website, but it would be an interesting tool to use when comparing different players on the same team.

One thing I don't think Ken Pom/tempo free stats capture well is how good a player is on defense.  Sure, there are stats on steals, blocks and rebounds, but that doesn't account for if your man blows by you for an uncontested layup or you dip under a screen and your guy hits an open 3.  I would think a plus/minus stat might help provide a more complete picture.

[robmufan]

yahoo does it in the boxscores of its NBA games.  Its kind of useless though, i dont like it!

[MUSF]

+/- is a very telling stat in hockey but I'm not sure it translates all that well to BBall.  Too many points scored and a lot less work put in to score or prevent a score in BBall.

[Henry Sugar]

They track it in the NBA.  Kevin Pelton started doing it first (I think)

The closest I've come is Brian's work from Hoya Prospectus.  He runs through actual play by play data to compile his info.  Has literally written 5K+ lines of code.

http://hoyaprospectus.blogspot.com/2008/11/big-east-hd-box-scores-marquette.html

I've gone back and forth on defensive stats.  I track defensive efficiency, but it's much more of an approximation based on the team defense and then adjusted for blocks, steals, defensive rebounds, etc. 

The reason I don't do it this year is primarily because of James.  His defense is so good that his guy hardly even gets the ball.

[Sweenz]

I think it does have some good bearing of not only defensively what you mean to your team, but also how involved you are offensively...

In the Duke, Georgetown game, Gtown was killing Duke when Monroe was on the court, I think +15 or something like that... but he was in foul trouble most of the game, and Duke went on a huge run every  time he was on the bench... could be a telling statistic for a lot of guys like Monroe, who are integral to the team and the offense runs through them.

[DaCoach]

Unfortunately there does not seem to be any available stats that recognize that cumulative defensive efforts of a player. And while there are some +/- numbers that reflect the difference in team scoring, it obviously reflects the results of everyone on the court, not merely a single player. It also is based upon the level of competition which can widely vary.

[dwaderoy2004]

The reason it doesn't work for basketball is that some guys play so many minutes.  so their +/- will basically be the final scoring margin of the game.  the stat could be useful for bench players, but not really starters.

[MUSF]

The reason it doesn't work for basketball is that some guys play so many minutes.  so their +/- will basically be the final scoring margin of the game.  the stat could be useful for bench players, but not really starters.

True, in a 60 min hockey game the best players will play only about half of that time and only in about 90 second to 2 minute shifts.  If you walk out of a hockey game with a +2, under those circumstances, you are probably having a pretty good game.

[robmufan]

True, in a 60 min hockey game the best players will play only about half of that time and only in about 90 second to 2 minute shifts.  If you walk out of a hockey game with a +2, under those circumstances, you are probably having a pretty good game.

Depending what the final score is +2 is a great game, you get a game 3-0, 3-1, i would say +2 could get you to be one of the stars of the game (depends on if you helped in getting the +2)

[MUSF]

Depending what the final score is +2 is a great game, you get a game 3-0, 3-1, i would say +2 could get you to be one of the stars of the game (depends on if you helped in getting the +2)

I say probably because just because you were on the ice for two goals and zero goals against that doesn't mean you were necessarily great.  You could just be really lucky.  Pucks bounce funny ways sometimes but it is a pretty good indicator.

[Wareagle]

Interesting article in the NY Times about use of advanced statistics in basketball and a quote about plus/minus.  The article is 10 pages long, so I won't post it all.

http://www.nytimes.com/2009/02/15/magazine/15Battier-t.html?pagewanted=1&_r=1&hp



"One well-known statistic the Rockets' front office pays attention to is plus-minus, which simply measures what happens to the score when any given player is on the court. In its crude form, plus-minus is hardly perfect: a player who finds himself on the same team with the world's four best basketball players, and who plays only when they do, will have a plus-minus that looks pretty good, even if it says little about his play. Morey says that he and his staff can adjust for these potential distortions — though he is coy about how they do it — and render plus-minus a useful measure of a player's effect on a basketball game. A good player might be a plus 3 — that is, his team averages 3 points more per game than its opponent when he is on the floor. In his best season, the superstar point guard Steve Nash was a plus 14.5. At the time of the Lakers game, Battier was a plus 10, which put him in the company of Dwight Howard and Kevin Garnett, both perennial All-Stars. For his career he's a plus 6. "Plus 6 is enormous," Morey says. "It's the difference between 41 wins and 60 wins." He names a few other players who were a plus 6 last season: Vince Carter, Carmelo Anthony, Tracy McGrady."

[Henry Sugar]

long and for the stat folks.

First, the NBA approach is really interesting, but can't quite translate to the college game.  There are several reasons.

• The NBA has a financial incentive (salary) that drives their statistical analysis
• With only 30-some teams in the NBA (compared to 341 in the NCAA) the player pool is much smaller
• The player pool is much more static, with productive players having a nice lengthly career vs a short one

This is what allows the NBA to derive their hyper-scouting reports, such as "force Kobe to his left and he only shoots 43% vs 54% when he goes right", or "Kobe is less effective putting the ball on the floor that spot-up" or whatever.

In some of my research, I found that Marquette actually used a system last year called Joctrac.  This allowed the team to come up with their own analysis, such as x% more effective after two passes than one.  However, it was basically only for MU analysis, required a dedicated person, and was difficult to travel with.  I don't believe they are using it this year.  The point is an example of how the type of hyper analysis in the article is difficult in college.

Moving on, I read a blog by Dave Berri, who wrote "The Wages of Wins".  He has a very interesting article in response to the NYTimes article.  By the way, although I really like Michael Lewis' writing, Berri points out that the concept of Morey/Battier was originally covered by a Houston Press journalist.  Within the link, there are several good links to read, including the HP article.

http://dberri.wordpress.com/2009/02/18/back-to-battier/

Cutting to the chase, Berri says that his approach also proves that Battier is a winning player, and that it's based on box score info. 

Theoretical discussion
As for his approach (Win Score / Win Produced), I keep going back and forth.  In Dean Oliver's book, "Basketball on Paper", Oliver argues pretty strongly against the notion of an "all-encompassing" stat like PER or Win Score.  If you don't know, Dean Oliver's work is the basis for just about everything on Pomeroy's site.  The guy not only played in college, but has a PhD in Engineering, and has served as a scout and consultant.  Anyways, the crux of Oliver's argument is that it's too hard to get to an "all-encompassing single number" for basketball because it is so much of a team sport.  Berri gets less flack from Oliver because they have collaborated, and because Berri derived a lot of his analysis using regression techniques.  However, the APBRmetric community is not much of a believer in Win Score.  Like I said... I go back and forth, and Berri's book is still sitting unread on my shelf.

Berri also has some additional commentary about plus-minus, with other perspectives if you're interested.  I have no opinion on it for the college game.

[Blackhat]

Pretty hard to quantify how effective a guy is at help defense also, unless you want a poor SOB to chart 20 million "stops" throughout a game. 

[Henry Sugar]

Pretty hard to quantify how effective a guy is at help defense also, unless you want a poor SOB to chart 20 million "stops" throughout a game. 

That gets to two points about basketball.

The first is that basketball is a team game much more so than baseball.  The canonical example is an assist.  One assist leads to one field goal.  But how do you weight credit for the assist or the field goal?  Was the basket created more by the assist or more by the player that did the shot? 

The second is tracking the impact of defense.  One proposal is to collect additional information in the box score in terms of "Forced Miss", defensive Field Goal Made, Forced Free Throw Missed.  In other words, tracking stats for guarding (or fouling) a player and how well that player shoots.  There was an effort to do it with the WNBA, but I'm not sure where it is.  However, even if you pull that off, like a stat for "Defensive Field Goal Missed", the help defense problem gets you back to the concept of "who gets the credit?".