Today’s post is about the Rookie of the Year. And the post is going to begin by covering some very familiar territory.

Blake Griffin was a unanimous choice by the media for Rookie-of-the-Year in 2011. This was not a surprising choice. As the following table illustrates, Griffin scored more than 700 points than John Wall, who finished second in voting for this award.

As is often noted, scoring dominates player evaluating in the NBA. And this award illustrates this observation. Every player who received votes scored at least 750 points in 2010-11. And all the players who did not receive votes scored fewer than 750 points.

Such a vote leads one to wonder (okay, led me to wonder), what would this award look like if we focused on all that a player did via Wins Produced? When we look at these players via WP, we see that Griffin is easily the most productive player. Wall, though, was only the 5^{th} most productive rookie. And DeMarcus Cousins – whose WP48 [Wins Produced per 48 minutes] was in the negative range — was only the 46^{th} most productive rookie.

If we look at the top 5 in WP – in the following table – we see that Landry Fields, Greg Monroe, and Ed Davis might have received a bit more attention from voters.

All of this is a familiar story. Once again, Wins Produced is not completely consistent with popular perception.

Rather than spend more time on that story today, I thought I would focus on something that has not been explored in the past. The Rookie of the Year is the only award that looks at how players compare to other players with the same level of experience. So what would happen if we looked at the top players in 2010-11 at each level of experience?

The following table reports the top 5 players who — at the conclusion of the 2010-11 season — had played 2 years, 3 years, 4 years, 5 years, and 6 years in the NBA.

If we look at Sophomores, the top player is Serge Ibaka. Tyreke Evans – the Rookie of the Year in 2010 – only ranked 17^{th} (of course, Evans was hurt this past year).

The class of players with three years of experience should be led by Derrick Rose. After all, Rose was the number one pick in 2008. And the media declared he was the Most Valuable Player in 2010-11. Although Rose is a “good” player, he is not nearly as productive as Kevin Love (Rose, though, does have much better teammates than Love).

Rose was not the only player taken first in the draft to fail to reach the top of these rankings. In fact, when we look at experience levels from 7 years to 13 years, we see that Dwight Howard and LeBron James are the only former number one picks who also appear at the top of their respective experience rankings.

So what do we learn in looking at these rankings?

- As noted, the first players chosen in the draft isn’t always the most productive NBA player in that draft class. This is not only true when you take a snapshot of a single season. It is also true when you look at career-performance (something we discuss in Stumbling on Wins). So if you follow a team that didn’t make the playoffs this year, it is not the end of the world if that team doesn’t get the first pick in the 2011 draft.
- There is also a significant drop-off as we move from the productivity of the first person listed at each experience level down to the 5
^{th}player listed. This result reflects the observation that most wins in the NBA are produced by a small number of players. And that also tells us that most players taken in the draft are not going to help much (a story you will not hear told on draft night). - And finally, these results reveal that each level of experience didn’t have the same level of productivity in 2010-11. Players in the 2
^{nd}and 11^{th}season in 2010-11 failed to produce one player who produced more than 10 wins this past season. Players in their 5^{th}year, 9^{th}year, and 13^{th}year only had one player produce more than 10 wins. In contrast, 7^{th}year and 10^{th}year players had four such players.

So the top pick isn’t always the best. Every players isn’t going to help. And every draft class isn’t equal.

Let me close by noting that I suspect these stories are the same if we focus overall career performance or performance on a per-minute basis. But I think I have made enough tables for one day.

– DJ

*Basketball Stories*

Jevan

May 9, 2011

I’d like to know how Kevin Love produced 25.6 Wins this season when his team only won 17 games?

Seems like a major flaw in these ratings.

nerdnumbers

May 9, 2011

Jevan,

Valid concern. https://dberri.wordpress.com/2011/04/22/the-tragedy-of-kevin-love/

His support was very bad and his team was unlucky. Pretty much worst case scenario for a budding superstar.

arturogalletti

May 9, 2011

Jevan,

Here you go.

http://arturogalletti.wordpress.com/2011/03/14/1-in-10-million/

Jevan

May 9, 2011

Arturo,

In your model you have the Timberwolves winning -4 games without Love. But that is impossible. A team cannot win -4 games. You could send 5 corpses out there and they would at least go 0-82. Obviously, something is wrong with the numbers. Which goes to my point. You can’t produce 25 wins..and that’s what wins produced is supposed to mean right? How many wins you added to your teams win column, if your team only wins 17 games. If your team wins 17 games, the maximum wins produced (if taking that literally) that you could possibly produce is 17. Right? What am I missing here?

Or maybe the metric isn’t accurate…and that Kevin Love exploits it…

Like discounting defense..(Half the game) which was the reason Kevin Love was benched by his coach in the early season..because he doesn’t play D, and this metric doesn’t account for it. Or that Love pads his rebound numbers, and takes rebounds from teammates. His RAPM is 0.0, making him an average NBA player according to that metric. But Wins Produced has him as the best player in the league…

nerdnumbers

May 9, 2011

Jevan,

I love new readers. Let me try and give you a tour of the facilities. We have a ton of articles (check the Wages of Wins Network on the side) I believe every site has a search option. We also have some FAQs, link on the top. I know this can be daunting but this site has been going on five years strong and has had some smart people contribute. Take time to see what we’ve done.

If you do that and come back I’d be happy to have a discussion. Unfortunately I’ve had the current one you’re making too many times and we’ve answered a lot of questions in the FAQ. Enjoy!

Philip

May 10, 2011

Jevan,

Wins Produced is based (correct me if I’m mistaken) on Pythagorean Wins derived from point differential, which correlates strongly with win/loss record. A team with a +10 differential on the season would expect to win about 70 games. A -10 team would win about 10. A -50 team would correlate with negative wins, while a +50 team would correlate with over 82 wins.

Your issue isn’t specifically with WP; it’s with Pythagorean Wins. So how about this addendum: Pythagorean Wins fairly accurately predicts team record accurately, until it gets to 82 wins or losses, at which point the record can’t rise or fall.

The point isn’t just to predict team record – it’s also to gauge how good a team is. A team with a +15 differential would win every single game it plays, or at least would have about a 99% chance of doing so. Ditto a team with a +20 differential. But the +20 team would still be much, much better.

In either case, does it really matter? A +12 team is almost a virtual lock for a championship unless it faces a similar juggernaut. A -12 team would be comically terrible.

Nate

May 10, 2011

It would be interesting to redraft each of the classes based on these numbers as well as some kind of metric that takes into account their age and potential years of productivity. (Would you draft a super productive 23 year old, or a slightly less productive 19 year old?)

ilikeflowers

May 10, 2011

Looking at some of the Dallas players on those charts, it’s interesting to see what Cuban has done. They basically have a large collection of 0.200 – 0.250 players (the ghosts of Shawn Marion and Jason Kidd are clinging desperately to greatness). This is a similar approach to the way that LA has won their last two championships (well that and thanking their lucky stars that KG is old and Mr. Glass can play every once in a while). If you can give 3+ borderline superstars enough minutes, then in the absence of a truly great team like the 2008 Celtics you can win a championship. It’s gonna be real tough for them to win both the Western Conference Finals and the Finals though given the competition.

ilikeflowers

May 10, 2011

Oh and on a side note, what’s this I hear about Kobe having problems with his dominant big man? Hmm, so if you get divorced from your dominant big man twice, does this mean that the problem might be you?

marparker

May 10, 2011

The negative wins are an issue. The model is based on efficiency differential. We’ve never seen a team win more than 72 or lose more than 73. That is the data we have to go with. Lets theorize that two zero teams occur in the same season. How would you quantify which players produced or didnt without using negative numbers? An exercise I found useful was to change the endpoint of the dataset and recalculate the numbers.

I’ve learned that you can’t take Wins Produced as a stand alone argument. Wins produced is more of a wins above teammates measure. Kevin Love was the most better than his teammates than any other player in the NBA is a more valid argument.

arturogalletti

May 10, 2011

Jevan,

I totally get your question. The simple answer is as follows: the models for performance in the NBA assume linearity (that is straight line behavior, more production= more wins in a defined fixed ratio). This is an approximation which works wonderfully within two standard deviations of the mean.

Let me expand on this point a bit. If we look at wins since the merger and exclude the strike season in 99, we find that wins are normally distributed with a std deviation for wins of 12.6 wins total. This means that we can expect >95% of all seasons to fall between 16 and 66 wins and it also helps the theory that incremental wins/loses outside of these parameters require incrementally bad or good performances.

In essence, the further away you get from the mean the harder it is to move the needle. In fact, there seems to be a hard floor/ceiling at around 9 wins and 72-73 wins.

The point is Kevin Love is good and the Twolves are so bad that it produces funky,weird numbers.

Tommy_Grand

May 10, 2011

J,

When we say the typical American household includes 2.3 kids, we don’t mean there is a cut up baby hidden in the freezer.

Imagine a scenario with 1 top NBA player on a team with four corpses. That team plays against a mediocre NBA team for 100 games. The 4-dead-body-team loses all 100 games, but its live player somehow scores 20 points and grabs 15 rebounds per. Can you imagine a metric to evaluate the losing team’s one good player? Could you convert that measure into a number of theoretical “wins,” even though the dead man’s team generated no real wins?

Let’s say the live player produced .2 wins per game, even though the team lost every game. Over 100 games, he’d produce 20 theoretical wins – even though team zombie lost each contest.

Of course, hypothetical wins don’t count for anything – only real wins court. But assigning a single number to a player’s contribution over time can be a useful and succinct way to compare productive players on sundry team of differing quality.

Schermeister

May 10, 2011

Also it is good to note that negative win players are still very very good basketball players, they are just negative compared to their peers. If every player produced an equal amount of wins they would all have a wp/48 of .1 and every team would be 41 wins in a season. However many players dont even produce at this level. A corpse would still be much much worse.

I dont see how so mnay people have such a hard time at this.

Say you and 2 friends are digging a hole. You dig 2 scoops a second and they throw in 1 scoop each per second. Clearly you are more productive at the task (digging a hole) and they are negative producers.. However you see that you are not making progress. Now someone from far away just reading the progress (hole depth or wins) would say all of you are terrible and making no progress even though it is your friends who are producing negative benefits. While you are quite productive.

Now imagine you are still diggin out 2 scoops a second and your 2 friends each put in 1.5 scoops per second. Apparently you are now making a mound. But lets say you can only read the depth of the hole. Thus the depth would still be at 0. And if you could read the height in terms of your goal it would read -X(X being the height in scoops). There is no way to measure negative wins. Just because we cant see negative wins doesnt mean we can speculate just how far behind the average(Zero depth hole) that you are. Basically a really negative win team would lose to a marginally negative win team.

I hope this helps a little. It is an over simplified version of some logic I use

arturogalletti

May 10, 2011

TG,

The Zombie Twolves and KLove would be a sight to behold.

Nick

May 10, 2011

@arturogalletti:

The ’10-’11 TWolves did a very close impersonation of this hypothetical Zombie team. Perhaps, Bill Simmons’ referring to the Thunder as the Zombie Sonics, should be countered in the WP community, as referring to the ’10-’11 Minnesota Timberwolves. “Zombie TWolves” has a nice ring to it.

@Schermeister:

Very well said. This particular argument comes up very often in this forum. People tend to see in the tangible sense, without really imaging the “mathematical” aspect.

It is quite difficult to see the difference between a 0-win team who might have an average WP/48 of 0.000 (The Nothings), versus a team with an average WP/48 of -0.100 (The Negatives).

As mentioned, while the results in a particular 82 game season may be the exact same (both teams produce 0 wins), the fact is that The Nothings would have a much greater chance to beat The Somethings (+0.100 WP/48) or The Everythings (+0.200) in any given game, than The Negatives ever would.

What Wins Produced would be saying, is that The Nothings would most likely finish with an equal or better record against The Somethings or The Everythings, than The Negatives would. Of course, MOST of the time, this would result in both teams going 0-82. However, The Nothings much more frequently than The Negatives, would have a 1-81 or better season.

Now if you play a 100 games between The Negatives and The Nothings, now you are giving The Negatives a better chance of winning, but still, they would be the underdog in all of the games (Unless they were the Denver Negatives…then they may win some home games). Of course, since WP is based on the league averages, the fact is that in a league of Negatives, and Nothings, the average would fall somewhere in between, and the Nothings would suddenly become almost Something.

Jevan

May 10, 2011

@arturo

The reason that negative wins is difficult to imagine, and the reason you may want to come up with a new analogy is that no one in the NBA is “throwing sand in a hole his team is trying to dig”..I.e Johnny Flynn may not be a very good NBA player, but at least he’s not shooting at his own basket. In your scenario, you would be better off digging the hole without your friend their at all. He’s just hurting you. But this isn’t true of any NBA player. The Timberwolves would not be better if Flynn just walked away, and they played 4 on 5.

No NBA player is contributing to his teams chances of winning “Negatively” in the sense that the team would be better off w/o him playing at all, he’s at least contributing SOMETHING..now maybe they would be better with someone else playing instead of him, but not NO ONE playing instead of him.

Jevan

May 10, 2011

So I was reading through the “Calculating Wins Produced” page and noticed something interesting that might be contributing to Kevin Love being overrated. Under WP, rebounds are valued highly because they predict winning –> a player who gets lots of rebounds must be contributing to his teams overall rebounds which in turn is contributing to his teams winning. Am I right so far? But with respect to defensive rebounds, is it really the ‘rebound’ itself that is contributing to the teams winning, or is it the MISSED SHOT? See what I’m saying here? That the defensive rebounds impact on winning is mediated by the missed field goal.

Giving a concrete example: Let’s say the Timberwolves are playing the Knicks. The Knicks have the ball and they feed Amar’e. Amar’e blows by Kevin Love, But Darko playing help defense steps in and forces Amar’e to alter his shot. He missed and Kevin Love recovers the rebound.

While Kevin Love secured the rebound, it was Darko who was the one who forced Amar’e to miss the shot. Yet as far as I can tell, under WP Darko would recieve no credit for this, and Kevin Love would get ALL the credit for grabbing the rebound.

Thoughts?

dberri

May 10, 2011

Jevan,

Your point was brought up 5 years ago. And brought up repeatedly since. Read the FAQ page. Think about what is being said. And if you still don’t agree… start a RAPM blog (of course, there are major problems with RAPM, but don’t let that bother you).

nerdnumbers

May 10, 2011

Jevan,

First off that’s actually not a bad question (aren’t rebounds just cheating and accounting for opponent misses?) The answer is no because how the model is derived. Wins Produced was derived using regression analysis, which means if you hold all else equal what does an extra [Fill in the blank with stat] gain you? It turns out opponent shooting is included and as such the value of a rebound was taken into account holding opponent shooting constant.

Good job asking questions. I will say though that if your retort was RAPM then you should be asking an RAPM similar questions. That said, keep at it! Figure out how the models work and add your own judgement to the equation. You’ve figured out step one, having a hypothesis. Move on to step two, testing it. You can also do step zero and just read the previous research but that’s boring right?

Jevan

May 10, 2011

I may have found the reason this metric overrates rebounding.

Tell me where I go wrong.

The logic behind the wins produced formula is that is uses a regression analysis to determine the relative weight of box score stats. So that say ..

The # of 2P FGM has a marginal value of 0.032, rebounds 0.032, and assists 0.022..

Now here is the important point. If I understand this correctly this is saying that defensive rebounding accounts for 0.032 of the variance in wins right? Not that EACH defensive rebound is worth 0.032 wins?

This is a very important distinction. Because when PROD is calculated it treats EACH defensive rebound as being worth 0.032 wins.

As you can see..this makes and ENORMOUS difference. If my understanding is correct, that the regression analysis says that defensive rebounding accounts for 0.03, or 3% of the variance in winning, then Kevin Love’s grabbing of approx .30 of his team’s defensive rebound, should then get credit for .30 of the 3% that defensive rebounding accounts for in producing a win. That would turn out to be 0.9%. And that is ALL his rebounds. Not each rebound. Since Kevin Love grabs 10 defensive rebounds a game, then each defensive rebound would be worth 0.0009 wins.

A far cry from the 0.032 wins each defensive rebound is calculated as being worth in WP.

In addition it also severely overrates rebounders because, instead of saying that all defensive rebounds account for .032 of the variance in wins, and all assists account for .022 of the variance in wins. .. when the illogical leap is then made that EACH rebound is worth .032 wins and EACH assist is worth 0.022 wins then assists get underrated and rebounds get overrated for the simple reason that there are more rebounds in basketball than assists.

An example.

Kevin Love gets 10 defensive rebounds* 0.03, equals .30 wins

Steve Nash gets 11 assists *02= .22 wins

But if we looked at from the perspective of ALL def rebounds are worth 3.2% of the variance in winning, and all assists are worth 2.2% of the variance in winning and Kevin Love grabs 30% of his teams rebounds, accounting for 0.9% of his team winning. And Steve Nash accounts for 44% of his team assists and since assists are worth 2.2% of the variance in winning than Steve Nash’s assists account for 0.97% of his teams wins. He now comes out ahead of Kevin Love.

What’s your take?

marparker

May 11, 2011

Jevan,

You really must get a peek at the FAQ page. Your points have all been hashed and rehashed in this forum in the past. Turns out accounting for defensive rebounding in a different manner doesn’t have much of an effect.

Jevan

May 11, 2011

Would someone be so kind as to show me the regression equation used to determine the individual weights? What was the constant? What was the error? The Adjusted R-square? This is really interesting stuff and I’m trying to understand how the math was used here.

Jevan

May 11, 2011

2 mistakes I’d like to correct. I addressed one of my posts to Arturo instead of shermeister, and I also see that in my second to last post that this metric actually takes each individual rebound to be 0.032 wins am right? I was confused by the wording on the calculating WP page..where it says the marginal value of defensive rebounds is 0.032, which I took to mean defensive rebounds as a SET, instead of being more succinct and saying that each individual rebound has a marginal value of 0.032…

Still, I’d like to see the regression equation …

thanks!

arturogalletti

May 11, 2011

Jevan,

I’m going to point you to :

http://www.stumblingonwins.com/CalculatingWinsProduced.html

That’s Prof Berri’s breakdown of WP

The regression equation is:

Winning %= .535 + 3.442 Offensive Efficiency -3.447 Defensive efficiency

with an R2 =93% using data from 78 to 91

Where

Offensive Efficiency = Points Scored divided by Possessions Employed (PE)

Defensive Efficiency = Points Surrendered divided by Possessions Acquired (PA)

Where

PE = FGA + 0.47*FTA + TO – REBO

PA = DFGM + 0.47*DFTM + REBD + DTO + REBTM

Philip

May 11, 2011

Jevan,

The average NBA team gains a lot more steals than forced jump balls. Is a forced jump ball more valuable than a steal?

Not all stats are equally valuable, nor is rarity a signal of importance.