APBRmetrics

phxgummer · Joined: 20 Jun 2005 Posts: 1 Location: Phoenix

I did not see any comments on Marty Burns article on the inaccuracies of NBA stats this season- lowest average of blocked shots and steals to lead the league. He attributes it to the fact that the league has moved statisticians off the scorers table to seats that are far from the floor, obstructed views and more.

Talking to a friend of mine who does statistics for one of the teams, I find that this is just the tip of the iceberg. The stats for many teams are totally innacurate and the league does not care. He told me to tape a game on TV (told me which teams had historically bad stats) and download a play by play from NBA.Com and compare what I see. He was right. Probably 10-15 discrepancies from what I saw and what was recorded and I don't know the scoring rules (but I do know the players and I can identify a rebound and a block).

How can the NBA allow this? How can people who care about stats allow this to go on- it is the history of the game! I will no longer participate in basketball fantasy now that I know that the numbers are not legit.

deepak_e · Joined: 26 Apr 2006 Posts: 357

I guess the key question is here if these inaccuracies tend to even out over the course of a season, or they consistently overcount or undercount certain stats.

It could be an interesting study to sample 5-10 games from every NBA team, audit the boxscores, and see if some teams tend to be worse than others for various stats.

Mark · Joined: 20 Aug 2005 Posts: 807

I hadnt seen it. Thanks for the note. Here is the link:
http://sportsillustrated.cnn.com/2007/writers/marty_burns/04/06/blocks.steals.notes/index.html

Seat revenue prompted the change but if stats and particularly fantasy stats are part of the sales campaign maybe steps could be taken to address concern. I assume NBA team's own stat analysis relies on same source so maybe accuracy would be deemed important from that perspective.

coach_dbop · Joined: 30 Mar 2007 Posts: 34 Location: California

I couldn't agree more with inaccurate statistics. If you go to most of the websites that are honored here and compare you will find many differences. I did however finally solve the problem with the players minutes played during the season. It appears for the past couple of seasons the NBA has gone to a new way of tracking a players minutes played in a game. In the past if a player played even less than a second he was credited for 1 minute played. Now if a player was to play 0.08 tenths of the final quarter he is credited for a game played but his minutes played with remain 0. In their player game logs if a player played 41:34 it will be rounded to the nearest whole number such as 42 and put under that players minutes for that game on his personal game log.

So for instance if you look at Chauncey Billups stats in the first 2 games against Orlando you will see on Billups game log that he played 41 minutes in game 1 and 42 minutes in game 2. On the Pistons playoff statistics Billups is credited for playing 82 minutes. Obviously 41+42=83 so I was able to finally find out how the NBA goes about doing this.

In game 1 according to NBA.com he played 40:32 which is rounded up to 41 minutes.

In game 2 according to NBA.com he played 41:32 which is rounded up to 42 minutes.

NBA.com i assume to get 82 minutes added 40:32+41:32 which equals 81:64. If you round that to the nearest whole number you get 82.

So while you will look at his game logs and see 42 and 41 the total number of minutes played will be determined on adding the boxscore minutes played and rounded.

This has been the case in the NBA for the past 2 seasons now, but if many teams and players for that matter stats are inaccurate I am very upset because I pride myself on giving people the accurate information that they are asking for.
_________________
Greg Eggleston

Lakewood Lancers Basketball Coach
Director of Basketball Operations

Kevin Pelton · Posted: Wed Apr 25, 2007 10:28 pm Post subject:

Is this really that big of a deal? Marcus Camby had a late surge, and his average of 3.30 bpg ended up better than his 3.29 bpg mark from a year ago and Theo Ratliff's 3.23 in 2002-03. Only once in the last five years has the league's leading shot-blocker averaged more than 3.32 bpg.

As for the steals, steals per 100 possessions are also at an all-time low. I don't see how a reasonable argument can be made that statisticians are missing when the ball is stolen. How else would the ball get from one team to the other? Plus, the per-100 possession rate is essentially the same as last year (0.79 vs. 0.80).

Here's how the top five in steals per game compare the last two seasons:

Dan Rosenbaum · Posted: Wed Apr 25, 2007 10:41 pm Post subject:

Folks complaining about NBA statistics should really spend some time as a labor economist working with self-reported income data. NBA data is a dream relative to what I am used to working with in my day job. And that work that I do in my day job has at times helped drive important policy debates - stuff that is way more important than anything that NBA statistics are used for.

As I tell my students, working with data is like trying to put out a forest fire. When you first approach the forest fire, you concentrate your efforts in the areas that will save the most lives and property. But you allow lots of blazes to keep burning. So yes, it would be better if NBA data was recorded perfectly, but this strikes me as a second or third order problem - akin to the blazes that are not put out until very, very late in the process of putting out a forest fire.

HoopStudies · Posted: Thu Apr 26, 2007 8:27 am Post subject:

There are a lot of ways I think the stat taking can be improved in the NBA, but it is, as Dan says, not a big deal in the big picture. We can still do a huge amount with the numbers as they stand. Errors as they may affect fantasy hoops are probably not as big a deal in real basketball (I haven't played fantasy hoops, but it seems like individual stats matter a lot more there for winning than in real basketball).
_________________
Dean Oliver
Author, Basketball on Paper
http://www.basketballonpaper.com

jeffpotts77 · Joined: 18 Feb 2005 Posts: 142 Location: Cambridge, MA

HoopStudies · Posted: Thu Apr 26, 2007 9:19 am Post subject:

mikez · Joined: 14 Mar 2005 Posts: 33

All of this stuff is a problem. But there are workarounds (for example, one could, if it turns out to be a big issue, see which scorers give out fewer assists, and apply a corrective factor, etc.), and the league is aware of these issues.

What actions can/will be taken to improve official scorers' accuracy is still an open question (though some things are going on behind the scenes that, like Dean, I'm unwilling to discuss here) but my guess is that if/when anything does happen you will not hear about it publicly.

-MZ

deepak_e · Joined: 26 Apr 2006 Posts: 357

HoopStudies · Posted: Thu Apr 26, 2007 1:31 pm Post subject:

jeffpotts77 · Joined: 18 Feb 2005 Posts: 142 Location: Cambridge, MA

John Hollinger · Joined: 14 Feb 2005 Posts: 83

I'll second that -- the stats as they currently exist treat defense as something of an afterthought. There has to be a better way. Tracking simple stuff like charges and goaltends seems to be a step in that direction.

Love the forest fire analogy, will have to remember that one. Actually maybe Dan has hit on something -- we can have self-reported basketball stats! "Ron Artest claimed to have 38 points and 14 rebounds as the Sacramento Kings defeated the Orlando Magic on Tuesday ...."

kjb · Joined: 03 Jan 2005 Posts: 685 Location: Washington, DC

I'd like to see better criteria and consistency for awarding assists. The current awarding is pretty haphazard. It would also be good to get more consistent guidelines on crediting steals.

Beyond that, one possible way to increase accuracy would be to generate the pbp live, but then have it checked on tape the next day. I catch errors that way, and it would be easy to fix them if I had access.
_________________
My blog