Is pau statistically more important than kobe?

Moderator: Doctor MJ

User avatar
amb1ent
Veteran
Posts: 2,810
And1: 17
Joined: May 22, 2006
Location: la ciudad
 

Re: Is pau statistically more important than kobe? 

Post#21 » by amb1ent » Mon Nov 9, 2009 10:42 pm

wreck wrote:
NetsForce wrote:All I know is that the Lakers wouldn't be 5-1 if Pau was playing and Kobe wasn't.


You're right. They'd most likely be undefeated. Maybe you should look up the Lakers' record since Gasol has joined the team. Then again, most people have (inaccurately) attributed the improvement to Bryant.


and what's the Lakers' record with Gasol but not Kobe?
mysticbb
Banned User
Posts: 8,205
And1: 707
Joined: May 28, 2007
Contact:
   

Re: Is pau statistically more important than kobe? 

Post#22 » by mysticbb » Thu Nov 12, 2009 7:49 pm

drza wrote:You just gave a long, well thought out rebuttal to the way that Berri interprets his efficiency stat.


Wins Produced is NOT a stats about efficiency. If Berri says something like this, he doesn't even understand his own regression analysis. Matter of fact is that he just took those boxscore numbers and related them to the points and opponents points. That doesn't even say something about winning at all. But that isn't the main problem. Wins Produced has several flaws, which Berri isn't even aware of.

I could go much more into detail, but I pointed several flaws out before in other posts.

Overall I would say a player can be described in numbers by using his production, his efficiency and his overall impact (some sort of +/- numbers). Gasol's production last season was close to Bryant's, but Bryant still produced more, Gasol had the better efficiency. So, overall Win Shares is taking those two things into account which gives the result that in the combination of production and efficiency Gasol was statistically speaking more important than Bryant. Well, the last part are the +/- numbers and Bryant was better in that department (raw, Net and adjusted) last season than Gasol. And here starts the interpretation of those results. While Gasol gets the edge in efficiency, Bryant is better in production and impact. There are two main reasons for Gasol's better efficiency. First of all the TPO will give a post player usually the better opportunity for an easy score, that is just how this kind of offense works. The 2nd reason is the opponents defense which had a stronger focus of not letting Bryant score too easy.
So, for an overall conclusion you have to take those things into account. And that is the reason why I say that Bryant was not only the best, but also the most important player on the Lakers last season.
User avatar
rrravenred
Retired Mod
Retired Mod
Posts: 5,906
And1: 425
Joined: Feb 24, 2006
Location: Pulling at the loose threads of arguments since 2006

Re: Is pau statistically more important than kobe? 

Post#23 » by rrravenred » Fri Nov 13, 2009 12:37 am

Well... Wins Produced values efficiency over several different raw stats. It's all (in my understanding) about how many possessions a player obtains for a team and what they do with those possessions.

Rebounding, missed shots, steals, blocks and turnovers are all influences on the raw numbers of possessions a player generates / loses, wheras made baskets and assists all impact on team efficiency. (this is based on a simplistic read on WP, so forgive any glaring idiocies)

It's bad on defence, but standard box-score stats don't really measure "proper" defence in any case.

You say that Kobe is better than Pau in terms of "production" and "impact".

Let's unpack those terms:

Production: I assume you're talking about individual point production, which isn't (for mine) necessarily the most important part of the game. Please correct me if I'm wrong. I'm an unabashhed fan of efficiency in terms of offence. You get the ball to a player who is most likely to score, either yourself (a good shooting percentage) or another player (a good Assist / TO percentage). If you turn the ball over or take a low-percentage shot, you're not helping your team.

Impact: +/-? Noisy, noisy, noisy. Starters generally play with the best players on their team, who are most likely to have the best chance to score / stop the opposition scoring. In theory, it works its way out over the course of a season. For mine, it never escapes the interference generated by the other players the coach puts on the floor. If you're talking about defence, you're on stronger ground, but I'd question whether on average Kobe is a substantially better defender than Pau.

BTW, what are the significant problems that you believe Berri hasn't recognised?
ElGee wrote:You, my friend, have shoved those words into my mouth, which is OK because I'm hungry.


Got fallacy?
mysticbb
Banned User
Posts: 8,205
And1: 707
Joined: May 28, 2007
Contact:
   

Re: Is pau statistically more important than kobe? 

Post#24 » by mysticbb » Fri Nov 13, 2009 2:24 pm

rrravenred wrote:Well... Wins Produced values efficiency over several different raw stats. It's all (in my understanding) about how many possessions a player obtains for a team and what they do with those possessions.


It is not possessions based. Berri just took the boxscore numbers for his regression analysis.

rrravenred wrote:It's bad on defence, but standard box-score stats don't really measure "proper" defence in any case.


This is true, and that is the reason why every metric based on the boxscore will not be very successful in representating defense. The highest correlation with the DRtg of a team will be found for defensive rebounds. You will get a correlation coefficient of around -0.5 between the pace and minutes adjusted defensive rebounds and the opponents points per possession. While you have a around -0.1 and close 0 for those adjusted values of steals and blocked shots. So, rating a player by his steals and blocked shots in terms of defense isn't a good way. Rating him by his DRB% is the better way. But the correlation is still pretty low in comparison to those 0.82 between PER and winning. So, PER isn't doing a bad job overall.

rrravenred wrote:Production: I assume you're talking about individual point production, which isn't (for mine) necessarily the most important part of the game. Please correct me if I'm wrong.


Production is not only scoring, but all the other numbers in the boxscore. PER for example has a pretty correlation to usage, more than it has a correlation to TO% or TS%. So, overall PER is more about production than about efficiency (despite of the name).

rrravenred wrote:I'm an unabashhed fan of efficiency in terms of offence. You get the ball to a player who is most likely to score, either yourself (a good shooting percentage) or another player (a good Assist / TO percentage). If you turn the ball over or take a low-percentage shot, you're not helping your team.


Well, if you don't shoot within 24 seconds, your team will lose their possession anyway. Therefore the assumption that taking a "bad" shot isn't helping your team isn't quite accurate. That is also one of the main problems of Berri's regression analysis, which leads to results like Rodman was more important to the Bulls than Jordan. Well, imagine a team with 5 Rodmans vs. 5 Jordans. Rodmans inablity to create a scoring opportunity will hurt the team more than Jordans lesser ability to rebound the ball.

rrravenred wrote:Impact: +/-? Noisy, noisy, noisy. Starters generally play with the best players on their team, who are most likely to have the best chance to score / stop the opposition scoring. In theory, it works its way out over the course of a season.


Uh, starters are also supposed to play against the better players from your opponent, right? So, your idea is basically pointless overall. You need also take into account, that adjusted +/- numbers are including those kind of things, which makes them a good indicator for overall impact. And again, Bryant was better in all three different +/- numbers.

rrravenred wrote:For mine, it never escapes the interference generated by the other players the coach puts on the floor. If you're talking about defence, you're on stronger ground, but I'd question whether on average Kobe is a substantially better defender than Pau.


Well, a big usually has the bigger impact on defense, which is related to the fact that the DRB% has a good correlation to the DRtg and bigs are usually the better rebounder.

rrravenred wrote:BTW, what are the significant problems that you believe Berri hasn't recognised?


A couple of those are related to the 24 second clock, Berri shows that he isn't really understanding Basketball at all. Another point would be his overall assumption that the boxscore can tell him enough about the game and the impact of a player for winning. Well, he is making strong adjustments to get even the numbers of wins for a whole team close to the real number by applying a team factor. He also makes an adjustment for position, which complitely contradicts his purpose of comparing different players on different positions. Basketball is a big men game, lower their numbers to get better values for small guys isn't a good way and can explain why he got Kidd as the more important player than Nowitzki on the Mavericks. Well, Nowitzki had even the better TS% and better TO%, if you use the possession based argument Nowitzki was the more efficient player.

I developed a different player rating, which gives an assists a higher value than the usual metrics. It also includes the effect of offensive rebounds. Overall Nowitzki was responsible for 30.9 points in every 100 possessions the Mavericks had last season, if I use my model. Kidd had 23.4. So, Nowitzki's offensive production was bigger and he done that with a higher efficiency.

Basketball is still a game in which a team wins, if it scores more points than its opponent. Basically that means that mean being able to score more points per possession than your opponent is the key to winning. Bryant has a clearly better TS% and better TO% than an average player, so he in fact helps his team winning by doing what he is doing on the court. His overall scoring efficiency is also hurt by the fact that he usually takes a shot, if the Lakers couldn't get their planned offensive system working for a possession. That will be most likely a bad shot. Gasol took way less of those kind of shots. As Gasol played for the Grizzlies he done that more often and his overall efficiency was lower than on the Lakers. I guess that is a pretty good example to show that only focussing on efficiency while comparing players is the wrong way. The complexity of the game just makes it hard to judge a player in this adequately.
User avatar
rrravenred
Retired Mod
Retired Mod
Posts: 5,906
And1: 425
Joined: Feb 24, 2006
Location: Pulling at the loose threads of arguments since 2006

Re: Is pau statistically more important than kobe? 

Post#25 » by rrravenred » Sat Nov 14, 2009 12:05 am

Well... I've always though Bryant occasionally confuses the Lakers offence with his own offence.

I think you're selling Berri short on his "understanding" of basketball because his conclusions conflict with your own "understanding".

Using box-score measures is limiting, I'd agree, which is one reason I'd happily commit a major felony to get the full dataset of someone like Daryl Morey, who collects data with a far wider breadth and collecting for specific game-affecting events that standard measures don't. But Box Scores do contain a lot of data which can be used to draw conclusions about the effects that particular players have on outcomes. Humans are notoriously bad observers who jump to conclusions on the basis of hundreds of ingrained (and invisible) biases. I know what I enjoy watching in basketball, and it's not always things that positively impact on winning. I almost always think my Bucks have done what it takes to win the game, if it wasn't for that damn scoreboard run by independent(ish) observers telling me otherwise.

That data of course will remain in-house in terms of the Rockets organisation, although I'd love to see what sports economists like Berri and Hollinger et al could do with something slightly more complex like the 82games data, which includes stuff like shot-type and phase of the shot-clock (and offensive fouls drawn, etc) and is based on far more accessible datasets.

Interestingly, based on 82games data, Bryant shoots far better in the final shot-clock phase (21 seconds plus) at .542 than he does at any other phase of the shot-clock. So unless you think the Lakers are methodical to the point of risking a 24 second violation, the "bail-out-shots-hurt-Kobe's-percentage" doesn't really appear to be supported by the data.

Code: Select all

Secs.    Att.    eFG%    Ast'd   Blk'd   Pts
0-10      37%     .531      56%     5%     9.9
11-15      28%     .449      45%     2%     6.3
16-20      22%     .462      47%     3%     5.1
21+      14%     .542      23%     0%     3.7
ElGee wrote:You, my friend, have shoved those words into my mouth, which is OK because I'm hungry.


Got fallacy?
User avatar
rrravenred
Retired Mod
Retired Mod
Posts: 5,906
And1: 425
Joined: Feb 24, 2006
Location: Pulling at the loose threads of arguments since 2006

Re: Is pau statistically more important than kobe? 

Post#26 » by rrravenred » Sat Nov 14, 2009 12:08 am

Oh, and before I forget, a minor sidebar.

mysticbb wrote:Basketball is still a game in which a team wins, if it scores more points than its opponent. Basically that means that mean being able to score more points per possession than your opponent is the key to winning.


PLUS obtaining more possessions than your opponent, both in terms of retaining possession (offensive rebounds, steals) and not losing possessions (defensive rebounding). If you turn the ball over consistently, or allow your opponent to towell you up on the offensive glass, brilliant shooting won't save you.

More possessions at higher efficiency = consistent wins.
ElGee wrote:You, my friend, have shoved those words into my mouth, which is OK because I'm hungry.


Got fallacy?
mysticbb
Banned User
Posts: 8,205
And1: 707
Joined: May 28, 2007
Contact:
   

Re: Is pau statistically more important than kobe? 

Post#27 » by mysticbb » Sat Nov 14, 2009 8:52 am

rrravenred wrote:I think you're selling Berri short on his "understanding" of basketball because his conclusions conflict with your own "understanding".


LOL, no! ;)

rrravenred wrote:That data of course will remain in-house in terms of the Rockets organisation, although I'd love to see what sports economists like Berri and Hollinger et al could do with something slightly more complex like the 82games data, which includes stuff like shot-type and phase of the shot-clock (and offensive fouls drawn, etc) and is based on far more accessible datasets.


Uh, what? You can get those kind of data out of the pbp. The motivation for Hollinger and Berri is a complete different one. They wanted to make a "simple" metric by using the boxscore numbers.

rrravenred wrote:Interestingly, based on 82games data, Bryant shoots far better in the final shot-clock phase (21 seconds plus) at .542 than he does at any other phase of the shot-clock.


Sorry, but are you serious? Come on, did you ever heard about the term sample-size? Go back and check the numbers for previous seasons and tell me again that a eFG% of .378, .488, .416, .418, .454, .362, .436 are far better than Bryant's average eFG%. Those bail-out-shots are one of the reason a #1 option can have a lower ts% than the #2, #3 and so on.
Get your facts right, before you try to answer those kind of stuff. Using those first 6 games from this season is just laughable.

rrravenred wrote:Oh, and before I forget, a minor sidebar.
PLUS obtaining more possessions than your opponent, both in terms of retaining possession (offensive rebounds, steals) and not losing possessions (defensive rebounding). If you turn the ball over consistently, or allow your opponent to towell you up on the offensive glass, brilliant shooting won't save you.


Honestly, think about those stuff first, before you make a comment. A team can get at max two more possessions than the opponents in the NBA in a game. A offensive rebound doesn't lead to a new possession. Also, turning the ball over is part of the equation to get the points per possession. The equation is pretty simple: points/possessions = points per possession
You are mis-interpreting those stuff as points per scoring possession, that is a different thing.

And that is the reason why I mentioned ts% and to% when I talked about efficiency.
User avatar
rrravenred
Retired Mod
Retired Mod
Posts: 5,906
And1: 425
Joined: Feb 24, 2006
Location: Pulling at the loose threads of arguments since 2006

Re: Is pau statistically more important than kobe? 

Post#28 » by rrravenred » Sat Nov 14, 2009 12:37 pm

Fair enough. You value your own opinion. I'm sure Berri does as well.

Probably right re: Hollinger and Berri's reasonably simplistic data-set for metrics. One reason being, of course, that they want to sell books to people who can do it themselves. You say it's easily obtainable via phb. Is there a public repository of that information as far as you're aware?

Having said that, Berri has had at least a minor play with the data.

About bail-out shots...

Small sample size for this season, true. A whoops moment for me. But you look at 07/08 and:

Code: Select all

Secs.      Att.   eFG%      Ast'd   Blk'd   Pts
0-10      39%     .573      45%     3%     9.3
11-15      29%     .447      45%     4%     5.3
16-20      19%     .454      32%     5%     3.6
21+      12%     .488      24%     4%     2.5


Shucks, looks like an 82 game sample to me. Get your facts right before dismissing my point out of hand. :P

Now that was the MVP season, and certainly an outlier in terms of his usual efficiency in that phase of the shot clock. Looking at the comparison between Pau and Kobe (which this thread was, at least originally) about... using the 08/09 season as a base:

* Kobe takes 2.5 shots per game in that phase of the shot clock, connecting at 37% percent.
* Pau takes 1.2 shots per game in that phase of the shot clock, connecting at 47%

and for kicks...

* Fisher took 1.26 shots per game in that phase of the shot clock, connecting at 49.5%

So there's maybe... one additional shot per game in terms of poor-offence buzzer-beating shots potentially mashing his FG%, opposed to Pau?

Kobe's assisted on 16% of made shots, whereas Pau is assisted on 70%. That cuts both ways. You can argue that Pau getting those assists means that he's getting the ball in a better position than Kobe, but you can also argue that this means that Kobe has held onto the ball for longer, meaning it's his poorer shot selection causing the bad percentages. I'm certainly not going to make that call, but it's worth considering.

Possessions? I think you're really being pedantic on definitions. And I wish you'd actually read what I'd wrote and made some attempt to understand the point I was trying to make. Well... let's compromise and call it shooting opportunities per possession. If you can:
(a) maximise your possessions in terms of taking care of the ball and stealing it from your opponent and
(b) maintain an advantage in shot attempts per possession over your opponent (denying your opponent offensive rebounds by efficient defensive rebounding whilst obtaining as many offensive rebounds as you can yourself) and
(c) maintain offensive efficiency in terms of shooting well as a team and ensuring that your most efficient players are the ones taking the shots

then you're giving yourself the best chance of winning.

I really think we're just on different trains of thought here, however.
ElGee wrote:You, my friend, have shoved those words into my mouth, which is OK because I'm hungry.


Got fallacy?
mysticbb
Banned User
Posts: 8,205
And1: 707
Joined: May 28, 2007
Contact:
   

Re: Is pau statistically more important than kobe? 

Post#29 » by mysticbb » Sat Nov 14, 2009 12:54 pm

rrravenred wrote:Shucks, looks like an 82 game sample to me. Get your facts right before dismissing my point out of hand. :P


Uh, his average eFG% was .503 in this season. Again, get the facts right, before you are going further. ;)

rrravenred wrote:And I wish you'd actually read what I'd wrote and made some attempt to understand the point I was trying to make. Well... let's compromise and call it shooting opportunities per possession.


Lol, sorry, I read what you wrote, that is the reason I answered the way I did. You were wrong about the possessions. Everything else is just not worthy to comment. I probably should just give you the suggestion to read Dean Oliver's "Basketball On Paper" first, that is very easy to understand. After that you are probably able to understand a little bit more. ;)

rrravenred wrote:I really think we're just on different trains of thought here, however.


Well, that might be right. Anyway, good luck with your ideas.
jinxed
Starter
Posts: 2,098
And1: 1,239
Joined: Oct 11, 2009

Re: Is pau statistically more important than kobe? 

Post#30 » by jinxed » Thu Nov 19, 2009 4:37 am

How do you calculate WP? I don't have the book.
Check out my book! "The Awakened Ape :A Biohacker's Guide to Evolutionary Fitness, Natural Ecstasy, and Stress-Free Living"

https://www.amazon.com/dp/B01MSJN3Q4?ref_=pe_2427780_160035660

Return to Statistical Analysis