The Knicks this season got off to a very bad start. And early on, Amare Stoudemire was actually the least productive player in the NBA. Stoudemire played so poorly, one person offered the following comment (yes, this was Italian Stallion):
Sorry to take this a million miles off topic, but I have to vent.
Amare Stoudemire can’t play basketball to save his life. He’s athletic as hell and can score when he gets a clear path to the basket or decent mid range shot, but he can’t handle, pass, or make plays at all, is mediocre as rebounder at best, and is a turnover machine.
Thanks. I feel better now.
In response to this comment, Andres (Dre) Alvarez – of newly re-designed Nerd Numbers – offered the following:
First off it won’t stay that way. You guys beat a top team in the East with Amare playing like crap. Right now Conley, Rudy Gay and Melo are playing super well. It won’t last. If Randolph and Amare at least return to average (for big men) and Fields keeps playing well then you guys look in good shape. Right now your roster is actually above average on the whole! 40 wins baby. Still it does amaze me that Amare is dead last in Wins Produced right now.
Just as Dre expected, Stoudemire and the Knicks did get better. In fact, Stoudemire has been playing so well that someone (yes, this was Italian Stallion again) thinks our perspective on Stoudemire would be very different if we ignored his awful start. Here is what IS recently said about Amare:
“I think it might make sense to look at Amare’s stats for the first 11 games (the Knicks were 3 – and since then (17 – 6).. . . It’s sort of like the Heat. When people evaluate the Heat now, they totally disregard the early games when they were playing very inconsistent basketball while working out their roles and style of play etc….”
Typically I don’t offer analysis over just a part of the season. But once again, Dre comes through with a response. And that response (which you should read) indicates that
- Stoudemire would not be considered MVP if one ignored the first eleven games.
- Landry Fields is still the primary reason the Knicks have improved.
- the Knicks do not need to do much to be a contender in the Eastern Conference.
Again, one should read everything Dre has to say on this topic. And don’t forget to check out the automated Wins Produced numbers. Yes, this site has also been updated (and it looks great).
Let me close by noting this short post at NYK Mistakes on Landry Fields. Wins Produced argues that Fields is the primary reason the Knicks have improved this year. John Hollinger’s Player Efficiency Rating, though, disagrees. According to PERs, Fields is actually below average. The problems with PERs have been documented (in this forum and by Wayne Winston in Mathletics), and so I agree that how PERs evaluates Fields is incorrect. But there is a part of this brief post where I think I disagree. The post argues that Hollinger actually “hates” Landry Fields – and again, although I think PER is incorrect – I doubt Hollinger has any feelings for Fields one way or the other. And the same would be true for people who employ Wins Produced (or any other performance metrics). In other words, I doubt people who employ Wins Produced in the evaluation of players “love” Dennis Rodman and “hate” Allen Iverson. At least, that is true for me.
– DJ
Wages of Wins Journal 
brgulker
January 9, 2011
Well, I don’t literally hate Allen Iverson the person. But, given the circumstances around how he became a Detroit Piston … well, hate is too strong, but I do have a very strong distaste.
entityabyss
January 9, 2011
Uh uh. No way. You’re a hater. WP highly overrates rebounding and doesn’t give credit to shot creation. Also, adjusted plus/minus is a better method of player evaluation. Since WP doesn’t correlate well with APM, but rather only with wins, it is wrong.
Back to reality. What happened to the people that said amare wouldn’t be able to score well without steve nash? They said he needed steve nash to be productive. Now (although his stats are nearly the same as last year), people believe he’s the man. I guess he doesn’t need nash to score so well. As for his rebounding, there will probably be people who say landry fields is stealing them. Why would they say that? Who knows.
dberri
January 9, 2011
And you forgot that I am just unwilling to admit the shortcomings of WP. Plus, if you think WP is correct, it is just because you a) don’t understand statistics and b) don’t understand basketball. At least I think I have read stuff like this somewhere (along with the usual personal attacks and stuff like that).
ilikeflowers
January 9, 2011
prof, do players typically experience a temporary decline in productivity when they switch teams when controlling for team mate quality (diminishing returns or the lack thereof)? I would assume so, but is this seen in the data?
dberri
January 9, 2011
There is less consistency (the correlation coefficeint — I think – declines from around 0.8 to around 0.7). But I don’t think there is a pattern to the change we observe. Some players get better. Others get worse. But the correlation coefficient is still quite high. One needs to remember that even if you stay on the same team you will see different teammates every year. Given that injuries do occur (so there is a good reason why performance should change), the high correlations in year-to-year performance suggest that the numbers we see from each player are primarily about the player.
dberri
January 9, 2011
I found where I did this analysis…
Using data set with 6,549 player observations from 1977-78 to 2007-08, we see 1,906 players switch teams. The correlation in ADJP48 from season to season for players who switched teams in 0.76. For players who stayed on the same team it is 0.85 (and for all players it is 0.83).
On average, the players who switched teams saw their ADJP48 decline by 0.007. If you stayed on the same team, the average declines is 0.003 (this small decline reflects the fact that for more players see a negative effect from aging as opposed to a positive effect).
So again, I think players are very consistent in the NBA. Changing teams doesn’t seem to change that story.
Alex
January 9, 2011
I still don’t think Amar’e can score efficiently (if that’s different from well). His true shooting % and effective field goal % are the lowest they’ve been in five years, and the lowest they’ve been since his first two seasons (Nash-less) excluding the year he was hurt and played three games.
Michael
January 9, 2011
Did Landry Fields dramatically improve after the first 11 games?
Greyberger
January 9, 2011
You guys are preposterous. If you’re going to cite TS%, Alex, please learn what a good mark is and a bad mark. Shooting 58% TS while using the number of possessions Amar’e is is historically good. If you don’t think Amar’e can score efficiently, then you don’t know how to read basketball statistics. I’m not just trying to be glib – what exactly would you consider an efficient scorer? 60% TS?
Only eight players in the league this season are shooting 60% TS or better while using more than 20% of their team’s possessions. If Amar’e isn’t considered efficient at 58% and 30% usg, I don’t know what is.
Michael
January 9, 2011
Also Alex, Amare’s TS% is above average, despite being one of the leagues highest usage players. So your position appears without merit, unless your definition of ‘efficient’ is particularly strict.
Michael
January 9, 2011
Ah Greyberger beat me to it!
Greyberger
January 9, 2011
Beat you to it, and with ten times the snark.
Chicago Tim
January 9, 2011
PER does seem to capture conventional wisdom about good and bad players. I sometimes wonder if Hollinger did that because he was afraid of the reaction if he didn’t tilt the stats to favor scorers. I’m not saying that cynically, I’m just saying that he may have honestly thought the stats needed to be adjusted to conform with conventional wisdom. And Hollinger is a journalist first, a statistician second.
I know that I recently made the mistake of saying that Landry Fields was a top twenty player this year on a non-WoW blog, and the reaction was volcanic. I’m still known as the Landry Fields man on that blog, and pretty much everything else I say is discounted on that basis.
dberri
January 9, 2011
Alex was apparently comparing Amare to himself, not the league average. I thought that seemed fairly clear from his brief comment.
dberri
January 9, 2011
Chicago Tim,
I have always thought that Hollinger sought to create a metric that captured conventional wisdom. He really made no effort in his book to tie his metric to wins (and as we have seen, that is because it really is a poor predictor of wins). Given that objective, PER is a successful metric. But it is also redundant. NBA Efficiency also captures conventional wisdom and it is easier to calculate.
Greyberger
January 9, 2011
You should read one of Hollinger’s books on the subject of basketball statistics. You might be pleasantly surprised by his credentials.
I don’t think he’s tweaking the model just so Jordan’s seasons come out on top. He just made different decisions and considerations hen making his metric, and ended up with a different set of compromises and concessions than the WP48 authors.
PER puts a lot more emphasis on scoring production and being a contributing player on offense. This reflects the variety of opinions you’ll find about the relative value of scoring production.
Greyberger
January 9, 2011
WP summing to wins doesn’t make it necessarily more meritorious than PER or anything else. It can correctly sum player WP to wins without correctly identifying who’s responsible for what. In fact that’s exactly what the discussion will be about with posts like this – whether Amar’e or Landry is responsible for what’s happening in New York.
PER says one thing, WP48 says another. But of course it’s really the underlying assumptions that went into making these metrics that account for the difference here. It’s not because WP was done well, and sums to wins, while PER smells.
Michael
January 9, 2011
“I have always thought that Hollinger sought to create a metric that captured conventional wisdom. He really made no effort in his book to tie his metric to wins (and as we have seen, that is because it really is a poor predictor of wins).”
According to a commenter called Jon on Arturo’s blog, Hollinger’s predictions are about 85% accurate right now, whilst the wins produced based predictions are at about 72%, with the WP residuals being 20% higher.
nerdnumbers
January 9, 2011
Greyburger,
This is one of my favorite posts (http://nerdnumbers.com/archives/69) I agree Amare is a great scorer. This season he is still doing that well (although not as well as he has before). The problem is to be a really great player you need to do other things well too. Also the credential game does not prove a point. For fun though. . . using your own argument, have you read Dr. Berri’s book or know his credentials? Also, the answer on the Knicks is both Landry Fields and Amare are responsible. They are the top two players on the Knicks right now
DJ,
Thanks for the shout outs and compliments on the new site. Also thanks to you and IS for the inspiration on the Amare post.
nerdnumbers
January 9, 2011
Michael,
Not quite. A bunch of WoW analysts made preseason predictions. We also looked at Hollinger’s predictions. Arturo made a big predictive model using the Wins Produced metric, which many of us used. Hollinger did not use PER and also as an ESPN analyst had insider information that casual sports fans did not. Feel free to explain how using non-PER methods with extra information against a Wins Produced method with less information proves PER superiority.
Also I assume that since the start of the season no injuries have happened, no trades have occured and no rookies have performed differently than expected. Devin Dingmen of the NBeh? blog does a great breakdown of some things that have hurt our predictions (http://thenbehteam.blogspot.com/2011/01/pre-season-predictions-how-have-i-done.html)
Michael
January 9, 2011
“Feel free to explain how using non-PER methods with extra information against a Wins Produced method with less information proves PER superiority.”
I wasn’t aware that this was my argument. Perhaps you could expand upon how John Hollingers ‘insider information’ gives him an advantage over you guys when he apparently knows nothing about basketball?
nerdnumbers
January 9, 2011
Michael,
“I wasn’t aware that this was my argument.” Your last comment started with a comment explaining how Hollinger’s metric didn’t work and then used his predictions as proof to his superiority over WP.
The WP predictions were based on expected minutes played for players. In basketball certain players can make or break a team (e.g Chris Paul and Brandon Roy). Having inside information on certain players injury info or expected play time can be tremendously valuable (e.g. Brandon Roy suffered a knee injury and his knees are insurable). There is also an element of luck involved (e.g. David Lee gets bitten by a zombie)
My problem with Hollinger is that he sells a metric that he doesn’t use himself. He may be a brilliant basketball analyst but he the metric that is all over ESPN (his employer) is not a good metric.
Italian Stallion
January 9, 2011
I’d like to clarify my positions about Amare’s performance this year.
Typically, I don’t believe in evaluating players off just a handful of games because of the small sample .
However, I do think when circumstance change dramatically looking at the short term can actually give you better insights into what’s going on than the long term will.
When the season started, there were questions being asked about whether Amare would be as productive with the Knicks and Felton as he was with Nash because he did so much of his high efficiency scoring in the pick and roll. Nash and Amare were regarded as the best PnR tandem since Malone and Stockton. Felton was coming out of completely different offensive system in Charlotte. If you think every player can execute this play equally well, just watch Toney Douglas try to play PG for a few games. ;-)
When they got off to a horrid start together and were turning the ball over like crazy, it was far from certain they would turn things around and allow Amare to get back to the peak levels he achieved with the Suns.
Certainly, I wasn’t expecting Amare to remain dreadful forever (and believe me he was dreadful for about 10-11 games), but that 100m contract was looking like a terrible decision. As a frustrated Knicks fan I was concerned and venting.
However, if you watched all the games, it was clear that Felton suddenly figured out how to execute the play properly and the Knicks made a few other adjustments that helped. Once that happened, Amare’s efficiency exploded again and his turnovers started to drop.
So IMHO a proper analysis of the situation now is to disregard those first 10-11 games as an aberration where two players that had never played with each other were struggling and could not execute the key play of the offensive system being used.
To understand this and draw the correct conclusions you would have had to actually watch every game so you could see that the early sub par performance by Amare was not just some random downswing likely to revert to the mean over time. There was a specific cause that had to be corrected in order to fix things. That was not certain to happen . However, when that cause was corrected, it changed the way to evaluate him now and going forward.
While I haven’t crunched the numbers, they probably suggest that he’s playing fairly close to the best level he achieved for the Suns. Barring injury, IMO he is likely to stay on a path that more closely matches his productivity from game 12-35 than 1-11 or the aggregate 1-35.
Fred Bush
January 9, 2011
I don’t think anyone is saying that Hollinger doesn’t know anything about basketball, just that he doesn’t know anything about statistics.
ESPN contracts out to various companies that do hardcore simulations, which cater primarily to sports bettors (and possibly NBA teams). If Hollinger uses his access to those models to help do his picks (and if I were him I certainly would) it obviously tells us little about the value of his own model.
Italian Stallion
January 9, 2011
I want to add one other thing.
IMHO, looking at aggregate stats can sometimes be harmful to your understanding even though it’s usually helpful.
Within aggregate data there are often subsets that are totally different than the group. That’s a lesson I have applied successfully to gambling on horses for many years and it’s a lesson I am applying successfully to basketball gambling now (at least so far).
I think this Amare situation is an example of that kind of thing.
Michael
January 9, 2011
I wasn’t using his predictions as proof of his superiority over WP, only that perhaps he does actually have some idea what he is talking about. Obviously a few predictions less than half way into the season are inadequate to ‘prove’ anything. I will take responsibility for not making my intention clear though.
Michael
January 9, 2011
“I don’t think anyone is saying that Hollinger doesn’t know anything about basketball, just that he doesn’t know anything about statistics.”
You’re joking right? Based on what?
Adam C. Morrison
January 9, 2011
Kinda funny to see IS become sort of a running joke here.
dberri
January 9, 2011
IS has not become a running joke around here. His comments have simply led further discussion.
mr obvious
January 9, 2011
Wait, i’m confused. I thought it was already common knowledge that hollinger was an idiot. That guy said that toni kukoc was more productive than dennis rodman, and it’s not even close. Now if anyone can defend that assertion please go ahead.
nerdnumbers
January 9, 2011
All,
Sigh ok let’s clarify some things. Hollinger is not an idiot. Hollinger does know about basketball. The major complaint about Hollinger is that he is “selling” a metric that doesn’t work and that he doesn’t use. To be fair this is nothing new to the world of sports, how many athletes have endorsements for products they probably don’t use or care about? That said it is actually quite deceptive to make predictions etc. not using PER and then publish articles using PER to “prove” a point e.g. the Kukoc > Rodman
Adam C. Morrison
January 9, 2011
“IS has not become a running joke around here.”
Perception is reality. You’re welcome to disagree with my reality, but to say it is wrong is clearly, well, wrong. (cwutididthar?)
Italian Stallion
January 9, 2011
Adam,
<Kinda funny to see IS become sort of a running joke here.<
Not sure why you think I'm the running joke around here when I'm the only one that has actually explained both the fluctuation in Amare's individual performance and how that contributed to the Knicks performance as well.
There is a tendency by some to simply look at stats without trying to understand the "why". However, the "why" is often critical because it helps you predict BEFORE the fact whether something is a random fluctuation around the mean or a sustainable deviation from the past.
For me, as a gambler, that is critical.
Perhaps much of the conversation is simply outside your range.
Italian Stallion
January 9, 2011
A few thoughts on PER.
I think it has been clearly and irrefutably demonstrated that PER has the potential to overrate inefficient scoring no matter what you think about the usage and shot creations issues.
What is disputed is whether the ability to create and score at higher usage levels has value over and above the value credited due to efficient scoring.
That is what PER is attempting to capture even if it’s seemingly overdoing it.
For example:
When Lebron James was scoring 30 points a game at high efficiency rates, was that also benefiting the efficiency of his teammates because on average they were able to take higher percentage shots than they would have with a more average teammate and higher usage?
Conversely, when a player is a very low skilled and low usage scorer, does the ability of his defender to cheat off him to help, sag and clog the middle, etc… have a negative impact on the efficiency of one or more of the other players on the court.
If those things are true, then players like Amare, Melo, and other high usage scorers are better than they look and perhaps players like Fields is not quite as good at he looks (while still being very good).
The trick then would be to attempt to capture those values.
Capturing those values in an advanced statistics model is well outside my range. That’s why I frequent advanced stats forums, ask questions, get feedback, think out loud, and look at the math of others etc….
However, I am doing something privately that has great value to me.
I look at point differentials adjusted for strength of opponent, home court, rest etc… when certain players are in and out of the lineup for full games due to injury, personal reasons etc… . (I actually think might have some significant advantages over adjusted plus minus and the complex line up issues)
Right now there are about 4-5 advanced stats models I could look at that would give me a different answer as to what Dirk’s value is to Dallas.
So what’s a gamble to do?
What is clear is that Dallas’s point differential has declined significantly since he’s been hurt. At this point I think I have a fairly good line on his value relative to his substitute. That alone is significant both for making estimates in the future when he’s absent again and for projecting how Dallas’s point differential will improve when he’s back.
Adam C. Morrison
January 10, 2011
“Perhaps much of the conversation is simply outside your range.”
umad?
ilikeflowers
January 10, 2011
IS’s perspective is very valuable whether one tends to disagree with him or not. His comments typically lean towards the skeptical and range from plausible to insightful.
It’s also important to remember that the model is describing the productivity that is seen. The data necessarily describes performance and value using the predominant basketball strategies for team construction (as the prof and others have noted in the past). There’s still plenty of wiggle room when dealing with marginal value both strategic and measured for individual player evaluation. The model is good place to start but there are plenty of details left to examine. This is a good thing because ferreting out the details is fun.
Patrick Minton
January 10, 2011
Chicago Tim,
Welcome to my world. I remember even when he was a rookie I said that Kevin Love was a top 5 power forward in the NBA, and last year I said he was a “superstar”. I got laughed at by a lot of people.
This year, he’s putting up MVP numbers in the “conventional” sense (i.e. per-game numbers). What I find amusing is that there is much talk about him having a “breakout” season, when in fact the real difference is that he is playing 36 minutes a game instead of 28. If you look at his 2009-2010 per-game totals and just multiply them by 1.25 you get pretty similar numbers. As I pointed out here on this very blog, Love was already the best rebounder per48 in the NBA last year, just ahead of Camby and Howard. He’s only very slightly better this year (per48 minutes). Why is everyone so “shocked” at his production?
The same thing might happen with Fields, as he gets a few more minutes and also gets increased shot attempts (the team is very unlikely to run any plays for him as a rookie), he’ll probably put up numbers similar to what Shawn Marion used to in Pheonix.
ilikeflowers
January 10, 2011
prof,
do you just have the whole season results when players change teams or can it easily be split into halves or quarters? It’d be interesting to see whether or not the wp48 variation from historical norms has a significant trend within the first n-games of the team change.
Italian Stallion
January 10, 2011
ilikeflowers,
Thanks. I try, but I’m learning too.
I’m less skeptical than I was initially. I mostly think the model is very good, but there are some things I would probably tweak, like the extremes of usage and a couple of others.
I would also stress the pre positional adjustment numbers more because I think some players are playing out of position. I’d want their true value reflected and not just their value at the position they are being assigned to play. Some players also have skill sets that are not typical for the position they play but still valuable.
Italian Stallion
January 10, 2011
>umad?<
No.
I'm strongly opinionated, occasionally arrogant, not a very good communicator with the written word, and often play the part of devil's advocate. So I sometimes piss people off and get abused. I get pissed off right back for about 30 seconds and go right back to being me. Despite it all I have good intentions.
GovernorStephCurry
January 10, 2011
WP doesn’t think Dirk is a star anymore? Look at his adjusted +/- and the team’s record with him out. I like WP, but it has it’s serious flaws. There is really no one be it all metric. I just prefer looking at rate stats and +/- for the player. Don’t need PER, WS, or WP. Do my own analysis.
arturogalletti
January 10, 2011
IS,
I echo ilf. Keep the comments up. They are insightful and thought-provoking.
Philip
January 10, 2011
IS,
I enjoy your comments. I wholeheartedly encourage you to keep on trucking.
reservoirgod
January 12, 2011
I support the Stallion. And wasn’t “Adam Morrison” the running joke in the comments earlier this season? I know I found his comments laughable…
todd2
January 13, 2011
IS is a passionate NYK fan—no complaints here! Last night vs the Jazz: both teams over 50%, 13 and 14 turnovers at a fast pace. Was anyone defending? On a side note, Heat vs Clips: 44-26 at the end of 1st and the big 3 playing 40 minutes. The Heat have spent too many games digging themselves out of holes. They play to the level of their competition, which I thought James did with the Cavs last year.
Adam C. Morrison
January 13, 2011
“On a side note, Heat vs Clips: 44-26 at the end of 1st and the big 3 playing 40 minutes.”
Did you watch the game? From the tone of your post it sounded like you didn’t. The Clips played near-perfect basketball.
Adam C. Morrison
January 13, 2011
I think they were shooting nearly 80% at one point in the middle of the 2nd quarter.
kevin
January 13, 2011
Well, it didn’t help that LeBron sprained his ankle. It looked eerily dimilar to what Rondo did when the Celtics played the Knicks. Came right down on the side of it.
Still, the clippers might not be all that good but they sure are fun to watch. Griffin makes at least one “Wow” play a game.
todd2
January 14, 2011
Didn’t see the Clips, props to them. I am in MIA and watch regularly. Point was they’ve been playing catch-up too often.
todd2
January 14, 2011
GSW played perfect basketball against the Heat too—72 points in the 1st half this month.
reservoirgod
January 18, 2011
Todd2:
How “often” should they be playing catch-up? I think your statement is based on recency & not frequency. A New Year’s Day game & once on a 5-game road trip don’t equal “too often” to me. Fact is, the Heat have produced an estimated 7.9 wins in the 1st qtr while their opponents have only produced an estimated 4.6 wins. They’ve been more productive in the 2nd & 3rd qtrs, but the point is that they haven’t been digging themselves out of holes often at all.
todd2
January 18, 2011
Down at the half against the likes of Washington, Houston, Milwaukee and Golden State of late. Shouldn’t happen to purported contenders.