Why +/- can’t be trusted for basketball analysis

When Ersan Ilyasova entered Sunday night’s game in Detroit, the Bucks were down 2-0 and Luc Mbah a Moute was headed for the locker room for some stitch work.  (See play-by-play log at NBA.com).

When Ilyasova left the game ten minutes later after picking up his second foul, the Bucks were down 29-21.   That’s a -6 in the world of box score +/- calculation. Pretty simply, right?

The Bucks were down 12 (70-58) when Ilyasova entered the game for Moute in the 3rd quarter and began a 13-6 run to end the quarter, whittling the Pistons lead down to eight (86-78) with 8:28 to go in the game.  That’s a +4 for Ilyasova in that stretch.

Negative 6 and +4 = -2 for Ersan (he had 17 pts at this point).  Yet when I looked at the boxscore online, the Yahoo-NBA boxscore from the Associated Press had Ilyasova at a whopping MINUS 21!

Even if you were just following along with the play-by-play online, you’d know that the +/- numbers for this game were off.   When Ilyasova’s was “corrected” with just under eight minutes left in the game, the box had Ersan at minus 3.  Better, but still wrong.

Then — somehow — as the Bucks fell down by 10 points, Ersan’s +/- actually improved to -1.

A few minutes later, the Bucks cut the lead to 3 (92-89) with Ersan still on the floor and his +/- had miraculously plunged again to -10.  What’s going on here?

All statisticians agree that Ilyasova had 21 points at that point.

The official NBA box, by the way, was even further off.  With the score tied at 94, they had Ilyasova at minus 17.   The correct number was +6.

We may be getting some idea how Ekpe Udoh was among the league leaders in adjusted +/- for Golden State last season.   (There was even some speculation that the trade of Andrew Bogut for Monte Ellis and Ekpe Udoh was inspired by young Udoh’s seemingly awesome +/- and his potential to do even greater things as he matured.  In fact, Bucks GM John Hammond even said words to this effect. Brewhoop addressed the phenomena in October 2012, almost three years ago as of this revision. The explanation for Udoh’s freakishly good +/- turned out to be that in Golden State he played a significant portion of his minutes with Steph Curry, and that those minutes added up to a small sample size.  Here, I’m obviously annoyed that the +/- numbers aren’t even correct some of the time). 

The final tape of our Bucks-Detroit game, despite a good game from Ersan, read Detroit 96, Bucks 94 as  Monta Ellis missing a jumper and a runner in the final 32 seconds.

Ilyasova finished with 24 points, 17 of those coming as he led the Bucks back from a 12-point 3rd quarter deficit.   The official box score at last read Ilyasvoa +3.   The correct +/- number for Ersan in this game +4.

And Ekpe Udoh?  He played 11:39, did not take a shot or grab a rebound, was largely invisible yet was credited with two steals I’m not sure he actually had.   His +/- for those invisible minutes read +7.  This +7 is the number that some will use to show how important Udoh was to the Bucks in this game, citing invisible and unmeasured basketball activities like setting good screens and not getting in the way of others.

It’s nice to have teammates.

Leave a Reply

Your email address will not be published. Required fields are marked *