Why I'm not a WP fan

[Nivek]

I referenced the article in the Warriors board thread on the World Championships, leading up the event. I would have to track that link down. But c'mon, nobody was predicting that the team would go undefeated; they weren't even favored to win gold.

A USA Today online poll with 1289 respondents had the following results to the question, "Will the USA win the hoop World Championships?"

- Yes. No team can touch them. -- 21%
- Yes. But they will be challenged. -- 60%
- No. This is not a championship team. -- 11%
- No. USA is good but others are better. -- 8%

Bear with me while I do the math -- 21 + 60 = 81. Hmm, 81% fall into that "nobody" category, I guess. So, congrats to WoW for predicting what 81% were also predicting. I'm not sure what statistical model those 81% were using, though.

[floppymoose]

It doesn't really matter. Any halfway reasonable way of predicting player performance is going to be right a lot of the time. That's not a difficult achievement.

The issue with WP is that there are clearly identifiable ways in which it could be better. Those ways are pretty obvious. But instead of working to continually improve WP, it's creator and many of it's boosters are dug in: it's WP's way or the highway.

That's not the scientific method, and it dooms WP to eventual irrelevance.

[Twinkie defense]

You don't quite understand science at all. It is not necessary to provide a better model (which indeed can be found in this very thread here!), but to disprove a model.

I understand science just fine - do you believe in Newtonian physics? They work just fine right? Except for when they don't (hello, Einstein ).

Easy to cherry pick examples that suit your case, but without a system with its own rigor and predictability you're just pissing in the wind. If someone thinks adjusted plus/minus is so great we can just as easily poke holes in that - if Carmelo Anthony turns the ball over and it turns into two points on the other end, who owns that -2? 'Melo or Chauncey? I would say 'Melo.

[floppymoose]

There is no 'rigor' in WP. That's the whole issue here.

[Twinkie defense]

I referenced the article in the Warriors board thread on the World Championships, leading up the event. I would have to track that link down. But c'mon, nobody was predicting that the team would go undefeated; they weren't even favored to win gold.

A USA Today online poll with 1289 respondents had the following results to the question, "Will the USA win the hoop World Championships?"

- Yes. No team can touch them. -- 21%
- Yes. But they will be challenged. -- 60%
- No. This is not a championship team. -- 11%
- No. USA is good but others are better. -- 8%

Bear with me while I do the math -- 21 + 60 = 81. Hmm, 81% fall into that "nobody" category, I guess. So, congrats to WoW for predicting what 81% were also predicting. I'm not sure what statistical model those 81% were using, though.

0% of respondents said the team would go undefeated; a distinct minority simply agreed that they would win gold. I would interpret "challenged" as meaning they are probably not going undefeated. So 79% of people explicitly said they would not go undefeated.

It's my recollection that Spain was favored to win it all.

You're grasping for straws man.

Flop, what was adjusted plus/minus' prediction on the World Championships?

[floppymoose]

What's up with the adjusted plus/minus stuff? huh?

Did I publish a book somewhere that I've forgotten about that said that I could explain 95% of nba wins by using adjusted plus minus? And if I did that, did I try to silence critics who pointed out flaws in that?

[Twinkie defense]

instead of working to continually improve WP, it's creator and many of it's boosters are dug in: it's WP's way or the highway.

That's not the scientific method, and it dooms WP to eventual irrelevance.

If that is true, then it is definitely following the scientific method... and if that sounds strange, check out Thomas Kuhn's seminal work on the history of science, The Structure of Scientific Revolutions.

[floppymoose]

I'm aware of the works of Thomas Kuhn. Quoting Kuhn is not going to convince me that WP isn't a sham.

[Twinkie defense]

Well Floppy you're a big critic of the WoW methodology and have put a lot of weight on plus/minus:

"Those who have discussed nba stat with me before will know I'm much more a fan of metrics that measure on/off court distinctions, than I am of metrics that are purely box score based."

While I have big problems with reliance on plus/minus, which, until we can generate more data in a very complex set of interactions, is guaranteed to be wrong. And you would say that WoW is guaranteed to be wrong. I just think that right now, +/- is wronger

I am very interested to see what comes out of these 3D cameras in our basketball arena though. They have been using similar systems in European soccer for a while, and I'm surprised it has taken the NBA so long to tepidly get on board.

[floppymoose]

+/- is very noisey. And I rarely ever mention adjusted +/-, because I don't know how it has been calculated in the places I've seen it, and in some of the places I've seen it, i t hasn't passed the eyeball test as well as raw +-.

The nice thing about +- is that it counts everything, so it is a great comparison metric for identifying players who contributions (either positive or negative) might be missed by more traditional box-score based stats.

But you can't just list every player by their +- and expect to have a good player ranking. It's way too contextual for that.

I would add this: those who look to support WP by attacking whatever it is they think the critics prefer... well, they aren't doing WP any favors there. The Fox News approach just doesn't work in sports statistics. The audience here is too clever for that.

[floppymoose]

And I do think on/off (not just +-, but a whole suite of on/off stats) are the future of nba player evaluation. There really is no reason not to include the knowledge of when the player is playing and when they aren't. It's more data, it's clearly relevant, and more relevant data is good.

[Twinkie defense]

Well Floppy you are a great basketball mind, and I appreciate your role in the best RealGM forum. A little too much one - sided bashing in this thread though. And I think you may have taken *personal* offense to WoW because of some of their statements. But I think we can all agree that we look forward to a day when we have even better data and analyses on what makes a good player/team/strategy.

And in the meantime, my favorite path towards success on the court: get damn lucky. True difference makers are few and far between.

[floppymoose]

Well, I did find dberri's original response to be senselessly personal. Why would he tell me that I am "not well read" when he really has no idea what I've read, unless he is just being a dismissive dick?

But this thread started before that happened. My problems with WP were not personal - they were based on an examination of how it got the correlation it touted.

[Idunkon1stdates]

I referenced the article in the Warriors board thread on the World Championships, leading up the event. I would have to track that link down. But c'mon, nobody was predicting that the team would go undefeated; they weren't even favored to win gold.

A USA Today online poll with 1289 respondents had the following results to the question, "Will the USA win the hoop World Championships?"

- Yes. No team can touch them. -- 21%
- Yes. But they will be challenged. -- 60%
- No. This is not a championship team. -- 11%
- No. USA is good but others are better. -- 8%

Bear with me while I do the math -- 21 + 60 = 81. Hmm, 81% fall into that "nobody" category, I guess. So, congrats to WoW for predicting what 81% were also predicting. I'm not sure what statistical model those 81% were using, though.

0% of respondents said the team would go undefeated; a distinct minority simply agreed that they would win gold. I would interpret "challenged" as meaning they are probably not going undefeated. So 79% of people explicitly said they would not go undefeated.

It's my recollection that Spain was favored to win it all.

You're grasping for straws man.

Flop, what was adjusted plus/minus' prediction on the World Championships?

While the poll did not explictly ask if respondendents thought the team would go undefeated, the 21% saying the team would be untouchable is pretty close. Further, 79% did not explicitly say the team would go undefeated. They said the team would be challenged. It is your interpretation that `challenged`means probably losing a game or two, but I am sure that was just a slip on your part and there is no reason for me to even argue semantics. That said, someone has presented you with evidence while you are simply relying on your recollection, so you might want to treat that person with more respect.

That said, I have no doubt that Wins Produced tends to be more accurate in its assessment of basketball than the typical respondent to an online USA Today Poll. The question is how Wins Produced holds up to educated scrutiny and how it compares to other abprmetrics. You can find answers to those questions in this thread and around the web if you are interested in looking. If you are only comparing Wins Produced to the lowest common denominator (which Berri and his acolytes tend to do - it is very easy arguing against a straw man), then you will no doubt be satisfied with it and it will continue to produce revelatory conclusions for you.

[Twinkie defense]

"More respect?" What on earth are you talking about dude.

These days, the World Championships begin with four groups with six teams per group, with the top four teams advancing through a series of single-elimination rounds.

So while I can't say what was in the minds of the poll creators (but you gotta love those infographics huh?), or the people responding to the poll as they sat at their Embassy Suite breakfast buffets reading their complimentary USA Todays, it is a fact that 1) the US hadn't won it all since 1994 with Dream Team 2 and Shaq, Reggie Miller, KJ, Mark Price et al.; 2) the '98 team finished 3rd, the 2002 team came in 6th, and the 2006 team came in 3rd; and 3) obviously, a team can win gold without going undefeated.

I'll give those poll respondents the benefit of the doubt of knowing that, though admittedly that could be a mistake

[Twinkie defense]

Let's take Monta Ellis.

Through March 15, Monta had a PER 18.78, which puts him among about the top 40 players in the League.

As of mid-February Monta had a WP48 of 0.106, almost perfectly average.

And in early March Monta was at -5.9 net on/off points per 100 possessions, with only Andris Biedrins worse among Warriors starters.

Will the real Monta Ellis please stand up?

[floppymoose]

The +- is interesting for him. Earlier in his career he was consistently positive, including scoring 20 points a game in the "we believe" season. Then he missed a year hurt, and then he came back and played tons of minutes and was also asked to carry the team. He's been consistently negative since then. On the one hand the minutes are so big that a lot of his "off" minutes are garbage time. But on the other hand it's a two season long trend.

I'm not a big believer in PER or WP, so I don't really have a much to comment on those. Monta is still asked to take the lions share of the "bail out" shots on the team, so that is going to hurt his efficiency (which will affect both PER and WP). So it's fair to say he's no Jordan or Kobe, who have carried bad teams before and who did it better than Monta has. But that doesn't really mean he's a bad player. It just means he can't carry an otherwise bad team.

This season his efficiency has improved and his play-making has improved. I've seen some real growth. But it hasn't really translated into a ton of wins for the Warriors. I don't blame just him for that, although I do think he has certain habits which are hurting the team, like never boxing anyone out when he rotates to a player under the basket.

But the Warriors are never going to be more than a 30-something win team until they have a front line that can both rebound and defend.

[floppymoose]

Jeremy Britton responded to someone on the original wordpress thread:
http://dberri.wordpress.com/2010/07/14/ ... n-50-wins/

He chides the guy, and also me, for an "i told ya so" mentality. Ok, so he can't just admit he was wrong, whatever. But I thought this part was funny:

You cannot forget that thinking with data feels odd a lot of the time. Correlation does NOT equal causation.

The irony had me laughing. We've been telling him all along that WP's correlation with team wins was not anything that proved it was measuring player value appropriately.

[ElGee]

Jeremy Britton responded to someone on the original wordpress thread:
http://dberri.wordpress.com/2010/07/14/ ... n-50-wins/

He chides the guy, and also me, for an "i told ya so" mentality. Ok, so he can't just admit he was wrong, whatever. But I thought this part was funny:

You cannot forget that thinking with data feels odd a lot of the time. Correlation does NOT equal causation.

The irony had me laughing. We've been telling him all along that WP's correlation with team wins was not anything that proved it was measuring player value appropriately.

Epic irony. I laughed. Of course, we could point that out to him, but we've all been banned for asking too many questions over there.

[EvanZ]

Didn't know there was a stats forum here. Otherwise, I probably would have joined a long time ago.

I'm not a WP fan either.