Statitics evaluation

[pad300]

All right, the following link may be of interest for those who enjoy basketball stats, and those who like to argue about which stats are better.

http://www.uncg.edu/eco/rosenbaum/nessis.pdf

Merry Christmas, Stats people...

[ShoogarBear]

Thanks, outstanding find. Finally, someone is attempting to do this in a rigorous fashion.

The bottom line from this paper:

1. PER is no better than PPG or MPG in predicting future team wins (tables 4-7), and in fact, it's somewhat worse.

2. PER does do a good job at predicting Adjusted +/- in future years.

3. However, and this is the surprising thing, Adjusted +/- does a crappy job at predicting future team wins.

[sendman]

Here are some interesting PG stat numbers to compare and to discuss. I find that $ per point numbers mind boggling.
http://images6.theimagehosting.com/Pict0118.45c.th.GIF (http://server6.theimagehosting.com/image.php?img=Pict0118.45c.GIF)

[FromWayDowntown]

Thanks, outstanding find. Finally, someone is attempting to do this in a rigorous fashion.

The bottom line from this paper:

1. PER is no better than PPG or MPG in predicting future team wins (tables 4-7), and in fact, it's somewhat worse.

2. PER does do a good job at predicting Adjusted +/- in future years.

3. However, and this is the surprising thing, Adjusted +/- does a crappy job at predicting future team wins.

What's intriguing to me is the general unwillingness of the statheads to accept that observation, irrespective of statistical analysis, is a useful evaluation tool. David Berri's statement that one could watch basketball for a day or for a thousand years and still be no better able to determine what causes teams to win or lose games without sophisticated statistical analysis strikes me as nothing short of hubris. It's seems somehow fitting that his pet tool is most readily assailed by Rosenbaum's study.

People who are trained to evaluate basketball players and build teams based on those observations can be helped by statistics, but it would be folly to think that statistics alone could provide a useful analytical tool for building teams. The Houston Rockets might be learning that very lesson at this moment.

[ShoogarBear]

What's intriguing to me is the general unwillingness of the statheads to accept that observation, irrespective of statistical analysis, is a useful evaluation tool. David Berri's statement that one could watch basketball for a day or for a thousand years and still be no better able to determine what causes teams to win or lose games without sophisticated statistical analysis strikes me as nothing short of hubris. It's seems somehow fitting that his pet tool is most readily assailed by Rosenbaum's study.

People who are trained to evaluate basketball players and build teams based on those observations can be helped by statistics, but it would be folly to think that statistics alone could provide a useful analytical tool for building teams. The Houston Rockets might be learning that very lesson at this moment.I think the statheads had a point with their observations about baseball. Baseball offense is very modular in that it's straightforward to determine the relative value of walks, singles, doubles, triples, outs, etc. and combine them in a unified measure. So I think they is some real value to the way they have introduced new measures of offesive productivity.

In basketball, it's questionable at best and more likely ludicrous to think one can come up with a single measure that combines scoring, rebounding, assists, steals, etc. into one number which can be used to rank players.

And Berri's statement is idiotic.

[FromWayDowntown]

In basketball, it's questionable at best and more likely ludicrous to think one can come up with a single measure that combines scoring, rebounding, assists, steals, etc. into one number which can be used to rank players.

Simplistically, Bruce Bowen's successes (and the Spurs' successes with Bowen) as an NBA player would seem to defeat the statisticians' efforts to quantify basketball success.


And Berri's statement is idiotic.

Berri's statement is an overt attempt to justify his own existence. And it fails miserably.

[Quasar]

Simplistically, Bruce Bowen's successes (and the Spurs' successes with Bowen) as an NBA player would seem to defeat the statisticians' efforts to quantify basketball success.

It could be argued that those statistics focus on a player's offensive ability and not his defensive ability. The only 'defensive' statistics measured today are defensive rebounds, steals, and blocks, which despite being important, are not always a good gauge of defense. E.g. taking out the main scorer by denying him the ball will fail to make a dent on today's "classic" scoreboard statistics. However, +/- does show he plays well, from what I remember :)

Tracking stats of the opposing player's performance when Bruce is NOT on him and when he actually IS may be a good way to determine the efficiency of players like him. I know this is how many on this forum look at the stat lines to rate Bruce's game... I'm referring to he opposing player's :
-FG% (easiest one...)
-Number of FG attempted (denial defense)
-O Rebounds (How well did Bruce box out? Defensive rebounds are not tracked as he won't be preventing the opponents from getting them. This will mainly work for players such as Bonzi...)
-FTs attempted (i.e. conceded by Bruce, or by a big if Bruce missed his assignment)
- forced TO's - offensive fouls, missed passes, etc.
- time with the ball
- time without the ball

These would then need to be compared to the player's average performance in all of these areas... I suspect one would see a large deviation whenever Bruce guards those players.

While this kind of statistic will be skewed due to factors like the opposing player being in a funk, or being "defended" by a bad defender, it should still provide a far better gauge of a player's defensive play. Perhaps this kind of statistic injected into PER or the NBA efficiency, with a high weightage would provide the necessary balance between defense + offense.

[greyforest]

Here are some interesting PG stat numbers to compare and to discuss. I find that $ per point numbers mind boggling.
http://images6.theimagehosting.com/Pict0118.45c.th.GIF (http://server6.theimagehosting.com/image.php?img=Pict0118.45c.GIF)

it's $ per average points per game

kind of a useless statistic

[sendman]

it's $ per average points per game

kind of a useless statistic

Maybe to you, but I wonder if guys that are paying Eric Snow, Luke Ridnour, Marcus Banks and Stephon Marbury are sharing your opinion.
I see this stat useful if you are trying to find out if somebody is producing according to the salary he gets paid. It's nice tool to pinpoint the underachievers.
Next thing that I find interesting is that a lot top PG's are getting OLD. It looks like we will witness a big changes in PG positions all over the NBA in the next few years.

[Quasar]

I see this stat useful if you are trying to find out if somebody is producing according to the salary he gets paid. It's nice tool to pinpoint the underachievers.

Assists or assist/TO rate are generally considered to be a better measure for the PG position than points.

[sendman]

Assists or assist/TO rate are generally considered to be a better measure for the PG position than points.
I agree with you to a point, but I still believe that you should consider overall effort. Assists are just one aspect of a game. Take Kidd for example, his Assists or assist/TO rate are very good, but his ability to be useful in so much different ways makes him special. Another example, Iverson's ability to score like a maniac makes him special. I think you get my point. And after all that being considered, you look at that players salary and you can find out if he deserves what he is getting.