RealGM

does anyone know if there's a measurement for the overall success of a PG-C/PF combo? as in, something to find the best PG to partner with a dominant big man?

thanks for all the great mentions so far, I use many of them. From a fantasy stats perspective, is there anything quite like basketballmonster.com?

What is the best stat to determine a player's defensive impact? I don't like stats that incorporate blocks and steals, and would prefer something focused on the offensive production of the player he is guarding. I realize this is difficult though as zone defenses and picks reduce the clarity of which player is responsible for guarding which player. I would think that, as a result, we would be regulating to looking at team defensive statistics with a player on and off the floor. However, the advanced stats on nba.com due break out an individual players performance against another individual, player, so it seems possible to individualize the impact of a single player. I just can't find anything that quantifies how well a player does at holding other players below their average FG% or EFG% for example. Any thoughts on the best means of determining individual defensive impact?

Rod700 wrote:What is the best stat to determine a player's defensive impact? I don't like stats that incorporate blocks and steals, and would prefer something focused on the offensive production of the player he is guarding. I realize this is difficult though as zone defenses and picks reduce the clarity of which player is responsible for guarding which player. I would think that, as a result, we would be regulating to looking at team defensive statistics with a player on and off the floor. However, the advanced stats on nba.com due break out an individual players performance against another individual, player, so it seems possible to individualize the impact of a single player. I just can't find anything that quantifies how well a player does at holding other players below their average FG% or EFG% for example. Any thoughts on the best means of determining individual defensive impact?

Defense RAPM.

http://stats-for-the-nba.appspot.com/ranking_rec

^How exactly is that calculated? And is there only RAPM stats for this year?

-Kees- wrote:^How exactly is that calculated? And is there only RAPM stats for this year?

Do you want "exactly" or "kinda sorta"? I gave the "kinda sorta" answer in my primer (towards the end):

http://www.d3coder.com/thecity/advanced-stats-primer/

It's a regularized version of adjusted +/- (APM). Basically, it decreases the variance of predictions that you would make with APM, which results in a lower overall prediction error.

Those RAPM ratings I linked to are a weighted multi-year result. Here are the single-year 2012 results:

http://stats-for-the-nba.appspot.com/ranking12_no_prior

I'm also calculating adjusted four-factor +/-, which is another way to do it:

http://www.d3coder.com/thecity/2012/05/ ... g-to-a4pm/

^How exactly does it decrease the variance? What happens to the stat/whats added to make it better than regular APM?

Also, thats a great blog. I read a whole bunch of posts you made, it's great work.

For anyone...I'm looking for a stat, basically, I'm wondering for each turnover, how often does it lead to a fast break basket attempt. So for each turnover, how often does the other team get out on the fast break and attempt a shot? I can't find it.

-Kees- wrote:^How exactly does it decrease the variance? What happens to the stat/whats added to make it better than regular APM?

As part of the regression a Lagrange multiplier ("lambda") is added as a penalty term to shrink variance. The value of "lambda" is chosen through cross-validation to minimize prediction error. It's really very common now:

http://en.wikipedia.org/wiki/Tikhonov_regularization

Does anyone know where I can find play-by-play data for the 2004 NBA Finals?

drejeronfire wrote:Does anyone know where I can find play-by-play data for the 2004 NBA Finals?

Game 1: http://www.basketball-reference.com/boxscores/pbp/200406060LAL.html
Game 2: http://www.basketball-reference.com/boxscores/pbp/200406080LAL.html
Game 3: http://www.basketball-reference.com/boxscores/pbp/200406100DET.html
Game 4: http://www.basketball-reference.com/boxscores/pbp/200406130DET.html
Game 5: http://www.basketball-reference.com/boxscores/pbp/200406150DET.html

Thank you... For some reason I never knew you could find play-by-play data on basketball-reference !

I've always wondered, how do these stat guys compile all of the statistics onto one spreadsheet/data set to be able to analyse/compare?

I've got some good knowledge of SPSS, Excel, SAS... Learning R currently.. But the actual retrieving of the data has always eluded me..

NM

The RAPM is enlightening, but I'd also like to know how they determine Defensive Ratings for individual players. I can find plenty on how it is calculated for teams, but not for individuals. I appreciate any help. Thanks!

Re: Best website to access advanced statistics?

Postby ag101 on Thu Aug 23, 2012 7:46 pm
I've always wondered, how do these stat guys compile all of the statistics onto one spreadsheet/data set to be able to analyse/compare?

I've got some good knowledge of SPSS, Excel, SAS... Learning R currently.. But the actual retrieving of the data has always eluded me..

ag101
Junior

Posts: 364
Joined: Mon Jan 12, 2009 9:09 pm

Top

If you are able to program with IBM SPSS and SAS you are on the forefront of what teams are looking for, however it has been my dealing that truly understanding the mission one is trying to accomplish and how to throw out good stats from misleading ones is truly the highest prize.
Excel is nice because its compartments allow for easy brain memory, the problem is many times it is a computer RAM hog and the program itself seems limited once you enter a good amount of data.

Does anyone know where I can find assist to turnover career stats?

EvanZ wrote:

Rod700 wrote:What is the best stat to determine a player's defensive impact? I don't like stats that incorporate blocks and steals, and would prefer something focused on the offensive production of the player he is guarding. I realize this is difficult though as zone defenses and picks reduce the clarity of which player is responsible for guarding which player. I would think that, as a result, we would be regulating to looking at team defensive statistics with a player on and off the floor. However, the advanced stats on nba.com due break out an individual players performance against another individual, player, so it seems possible to individualize the impact of a single player. I just can't find anything that quantifies how well a player does at holding other players below their average FG% or EFG% for example. Any thoughts on the best means of determining individual defensive impact?

Defense RAPM.

http://stats-for-the-nba.appspot.com/ranking_rec

Thanks for this, but I'm not quite sure how to interpret the numbers on that site. Can you give me a brief description, like the offense is X per Y, the defense is X per Y, and thus the total is X per Y?

EDIT: Let me add one other question that would help me get the basics down. In that link why are we adding defense to offense, instead of subtracting defense from offense like we would with Ortg and Drtg?

The listed value is what the player brings the team over an average player, if he'd play the entire game.
So, if you replaced an average player on any team with Nowitzki, their offense should be 4.8 points better per game if Nowitzki plays 100 offensive possessions every game. If Nowitzki would play 100 defensive possessions per game, he'd improve his team's defense by 3.2 points per game.
For Nowitzki to help your team by 8 points per game he'd have to play the entire game (100 offensive possessions and 100 defensive possessions). In reality, no one probably plays more than 150 possessions total per game.

Whether we add defense to offense or just subtract it: that's just semantics. In case of this metric positive just always means "good"

blabla wrote:
Whether we add defense to offense or just subtract it: that's just semantics. In case of this metric positive just always means "good"

Exactly. Positive on both offense and defense correlates with an *increase* in total point differential, which is what we're after.

Rod700 wrote:Does anyone know where I can find assist to turnover career stats?

I was convinced http://www.basketball-reference.com has it but they don't. Good catch.
Maybe you should suggest them this improvement?
I would guess it will be easy to add and I've proposed too many suggestions ;-)