Why I'm not a WP fan

[azuresou1]

Great article. I hope Berri actually does wander into the field of baseball and gets reamed hard by infinitely better statisticians.

LOL at just doing multiple regression analysis in a sport where so much isn't captured by boxscore. This is a guy who'd rather have Jamario Moon than Melo and would give Moon starters minutes.

[DSMok1]

What disturbs me is that mainstream economists trying to explain/understand the economy seem to be about as good at that as these sports economists are at sports stats.

[giberish]

I actually expect that some of Berri's excessive certainty comes from being an economist.

Especially as it comes to being heard amoung the political shouting classes, it's considered a huge sign of weakness to show any doubt or have any uncertainty in your position. Meanwhile being relentlessly wrong, but loud and certain makes you an expert.

For his data, it's better than just shouting PPG, PPG, it's more like a generally good scout with known biases. When you know the biases you can use the information realatively well (and other compilation stats have biases as well, and +/- has noise issues so nothing's perfect).

Berri's most proment biases are overrating defensive rebounding - for the team as a whole it's very important, but most of the time if player X didn't get a particular defensive board, a teammate would have; underrating usage - casual fans overrate this, and many if not most NBA teams are oversubscribed in terms of the ability to create shots, but high usage players do allow high-efficiency role players to be effective; and as with all box-score compilation stats, largely ignoring defense.

Lee obviously benifits from the defensive rebounding and ignoring defense, while Camby is helped by the defensive rebounding and low usage.

[Doctor MJ]

Describing it as a scout with known biases is a good analogy. The problem with WP though - there were already better known scouts with less crazy biases out there before Berri even started with this stuff. I think everyone with a predilection for quantitative analysis is sympathetic to someone who creates something that moves things forward despite having clear flaws - but Berri didn't move anything forward he's just needlessly creating something new and trashing everything else that's out there.

[floppymoose]

It should be rather obvious that the amount of rebounds in a game is somehow connected to the missed shots. He didn't account for that.

This reminds me of the time years back (roughly a decade ago, I think) that the Warrior led the league in rebounds. Commentators were always talking about how good a rebounding team the Warriors were.

Except that they actually got less than 50% of the rebounds. They just played at a fast pace, and missed a *lot* of shots. They got lots of rebounds because the Warriors played with the Warriors on the court every night, while the other teams only got to do that a few times a season.

[sp6r=underrated]

Let's see how the Warriors are doing 1/2 through the season.

2010: Warriors 26-56 (SRS: -3.27)
2011: Warriors 35-47 (SRS: -3.03)

The warriors have made a decent improvement by w-l but PD shows almost no significant change. There is no significant difference in injuries compared to last year. Lee did miss 9 games but Biedrins has been healthy.

[ElGee]

Let's see how the Warriors are doing 1/2 through the season.

2010: Warriors 26-56 (SRS: -3.27)
2011: Warriors 35-47 (SRS: -3.03)

The warriors have made a decent improvement by w-l but PD shows almost no significant change. There is no significant difference in injuries compared to last year. Lee did miss 9 games but Biedrins has been healthy.

Close to that 57-win pace.

[floppymoose]

I doubt any other Warriors fans would agree with me, but I think Biedrins health is the key factor. Ws are 29th in rebounding this season, but we actually get 50%+ of the boards when Biedrins is on court.

[sp6r=underrated]

Let's see how the Warriors are doing 1/2 through the season.

2010: Warriors 26-56 (SRS: -3.27)
2011: Warriors 35-47 (SRS: -3.03)

The warriors have made a decent improvement by w-l but PD shows almost no significant change. There is no significant difference in injuries compared to last year. Lee did miss 9 games but Biedrins has been healthy.

Close to that 57-win pace.

I don't even care about the 57 pace being off. It is the fact that it isn't even close. There has been no noticeable improvement in the warriors when you evaluate the team by point differential. They're treading water. I guess Lee isn't as good as Wade.

[mysticbb]

I should have know better, but somehow I followed a link from another page to Berri's blog a couple of days ago. I read a couple of comments and someone honestly said that WP48 wouldn't have an issue with defense (unlike PER). Now, everyone knows that Berri is just using a team adjustment to account for the defense. I pointed out that Berri has no chance to adress defense on an individual level, because the boxscore just doesn't contain enough informations for this. I pointed two examples out: Kris Humphries (apparently he is franchise player material) and Kevin Love (the by far best player in the league). Both are below average defenders who are skipping their defensive assignments for rebounds. Every non-boxscore based data is showing that. Well, but those numbers are, according to those people, not reliable. Well, whatever. Now, yesterday one of Berri's lemmings felt the need to tell everyone who the "real" All-Stars are: http://nerdnumbers.com/archives/1474

It is a fun read, because obviously Nowitzki is suspect and Kris Humphries should have been an All-Star. Yeah, the guy really wrote: "Kris Humphries – A classic story of an underrated player good at getting boards. Part of his problem is that he is not getting enough minutes. Unfortunately both coaches and fans are swayed by per game numbers and Humphries is below 10 on both points and rebounds. Looking at his per 36 numbers shows how well he’s actually been playing."
So, Humphries is not getting enough minutes. Fine, we can make a crosscheck with the real OnCourt results of a specific lineups with Kris Humphries and how the sum of the WP48 is correlated to the real result. We take the lineup with the most minutes on the Nets: Harris (0.177), Morrow (0.069), Outlaw (0.007), Humphries (0.346) and Lopez (-0.043). They are playing 239 minutes together. According to WP48 those 5 together should generate an average of 0.556 wins per 48 minutes. Well above league average, which suggests they should be able to outscore their opponents in average, right? A team should be able to score more points than their opponents, if they are that good according to a metric. Now, during the 239 minutes those 5 players where together on the court they had an average ORtg of 103.8 and an average DRtg of 115.9. They got outscored by 12.1 points per 100 possessions. Seriously, a team which is supposed to play above league average gets outscored per 100 possessions like the WORST team in the league right now. Are we supposed to believe that those 5 always faced the very best team during their 239 minutes?

Well, we can check a similar thing for the Timberwolves. Their lineup with the most minutes is Milicic, Love, Beasley, Johnson and Ridnour. They are playing 276 minutes together. WP48 tells us that this lineup should generate 0.516 wins per 48 minutes. But in reality they are gettting outscored by 4.5 points per 100 possessions. How can someone based on those numbers even think about the possibility to suggest a minutes distribution? Those people are claiming that coaches are dumb and not able to distribute the minutes right. Really?

We can look at the opposite site. The Mavericks are playing Haywood, Nowitzki, Marion, Terry and Barea the most minutes with 267. According to WP48 those 5 together are generating 0.365 wins per 48 minutes. Yes, that lineup should get heavily outscored by those mentioned lineups of the Timberwolves and the Nets. Well, who would you pick?
But let us do the reality check. During those 267 minutes the Mavs lineup is outscoring their opponents by 11.2 points per 100 possessions. Uh? What? The way below average players on the Mavericks together are trashing their opponents? Who would have thouht this?

Now you can ask how my rating is doing for those players? Well, in average the Timberwolves lineup has 9.8 and the Nets has 9.5. I can translate that via a linear model to 0.47 wins per 48 minutes for the Wolves lineup and 0.44 for the Nets lineup. For the Mavericks I get 11.1 which equates to 0.64 wins per 48 minutes. Looks more in line with the reality, doesn't it?

While I'm at it, Camby, another All-Star snub. Camby, Aldridge, Batum, Matthews and Miller played 323 minutes together. According to WP48 the best possible lineup for the Blazers. Together they should generate 0.862 wins per 48 minutes. In reality they have +0.6 per 100 possession. My rating has them together with 10.8 or 0.599 wins per 48 minutes.

[sp6r=underrated]

1. Very nice lineup analysis

[mysticbb]

The real thing here is Kris Humphries. The guy is doing better per minute than LeBron James. He generated so far, according to WP48 9.3 wins, while Nowitzki has 5.4 in basically the same amount of minutes. The Mavericks traded Humphries last season for salary purposes. According to WP48 and those guys there they should have kept him and traded Nowitzki instead for Ronnie Brewer to improve their weak SG position and get an overall better team.

I just wait for the upcoming free agency when we can watch "The Decision" by Kris Humphries. How many teams will give him a max to build their contender around him? Oh, wait, the decision makers in the NBA are just dumb, they will not do that. Not because they are evaluating players on another level, no, because they are just not capable of using stats properly.

[sp6r=underrated]

The real thing here is Kris Humphries. The guy is doing better per minute than LeBron James. He generated so far, according to WP48 9.3 wins, while Nowitzki has 5.4 in basically the same amount of minutes. The Mavericks traded Humphries last season for salary purposes. According to WP48 and those guys there they should have kept him and traded Nowitzki instead for Ronnie Brewer to improve their weak SG position and get an overall better team.

I just wait for the upcoming free agency when we can watch "The Decision" by Kris Humphries. How many teams will give him a max to build their contender around him? Oh, wait, the decision makers in the NBA are just dumb, they will not do that. Not because they are evaluating players on another level, no, because they are just not capable of using stats properly.

check the link again and read the comments

[floppymoose]

<dberri>
mysticbb, the numbers you are citing are +-, which are notoriously inconsistent and hence unsuitable as a tool for player analysis. This has all been discussed and put to rest before. I think you are not very well read. If you would only purchase my book "Stumbling into Basketball Analysis", all of this would become clear to you.
</dberri>

[mysticbb]

Ah, sp6r=underrated, nice.

@floppymoose

Well, I got a rather weird e-mail from him after he deleted one of my comments. But nothing really surprising at all.


I checked the numbers for the Top50 lineups in terms of minutes. I posted the result on my blog: http://bbmetrics.wordpress.com/reality_check/
The data can be found behind the link, everyone can just download the xlsx.

Overall WP48 is doing better than the first couple of lineups suggested. But overall it is worse than WS48 and my own rating. With a 0.58 correlation coefficient for the adjusted +/- data and 0.63 for the unadjusted OnCourt ratings WP48 has the lowest correlation of those three. WS48 has 0.64 for the adjusted and 0.68 for the OnCourt ratings. My own rating has 0.68 for the adjusted values and 0.71 for the OnCourt ratings.

Those numbers aren't suggesting that someone should call coaches dumb or should think basing on a boxscore metrics he can make a proper lineup. Especially WP48 is last thing someone should use for this.

[Doctor MJ]

It's really fascinating to watch Berri and his followers. It's like a cult. I marvel at how certain they all are about the superiority of their techniques. It would be stunning even if Berri was the lone mathematical voice out there for them all to have such confidence, but the fact that he's not at all unique just blows the mind. How do these people choose to follow this voice when there are so many others, equally as mathematical, and far less radical? Clearly, Berri's insistence on being so radical is part of the draw. People want to be in that crowd that gives both an easy answer, and a feeling of being part of a revolution.

I'm glad people here are bringing up that the Warriors are not behaving at all like the WP people predicted. I also find it so amusing that Love is evidentially by far the best player in the league, yet his team is horrendous, and not much improved over how they were with him as a back up.

[floppymoose]

floppy, read up on the thread, and I think you sound very reasonable. This will be a good test for WP, thought stuff like this frustrates me:

Do the doubters of this not see the 500 ifs?

Many of these ifs won’t happen. Nelson isn’t going to allocate the minutes like the author states most likely, they never do.

Berri and his followers are very well versed at picking the slightest deviations from what they think the coach/GM should do, and using that to blame for under-performance.

Let the excuses begin!
http://dberri.wordpress.com/2010/07/14/ ... ment-79449

[Doctor MJ]

floppy, read up on the thread, and I think you sound very reasonable. This will be a good test for WP, thought stuff like this frustrates me:

Do the doubters of this not see the 500 ifs?

Many of these ifs won’t happen. Nelson isn’t going to allocate the minutes like the author states most likely, they never do.

Berri and his followers are very well versed at picking the slightest deviations from what they think the coach/GM should do, and using that to blame for under-performance.

Let the excuses begin!
http://dberri.wordpress.com/2010/07/14/ ... ment-79449

Oh now you've done it.

[sp6r=underrated]

It's really fascinating to watch Berri and his followers. It's like a cult. I marvel at how certain they all are about the superiority of their techniques. It would be stunning even if Berri was the lone mathematical voice out there for them all to have such confidence, but the fact that he's not at all unique just blows the mind. How do these people choose to follow this voice when there are so many others, equally as mathematical, and far less radical? Clearly, Berri's insistence on being so radical is part of the draw. People want to be in that crowd that gives both an easy answer, and a feeling of being part of a revolution.

I'm glad people here are bringing up that the Warriors are not behaving at all like the WP people predicted. I also find it so amusing that Love is evidentially by far the best player in the league, yet his team is horrendous, and not much improved over how they were with him as a back up.

I don't think being radical is a significant factor for his appeal amongst the Gladwell types. For them Berri's work has significantly more credibility because it has been published in peer-reviewed journals.

[Doctor MJ]

I don't think being radical is a significant factor for his appeal amongst the Gladwell types. For them Berri's work has significantly more credibility because it has been published in peer-reviewed journals.

Oh god. Yeah, THAT aspect too.

Gladwell bothers me so much because when he talks on other subjects, he makes a ton of sense to me. Yet, he comes to an area that I really know a lot about, and he's utterly duped by a hack with an academic backing. The unfairness to other basketball analysts is one thing...but how am I supposed to take seriously Gladwell's research in any other area when he has such analytical vertigo?