Earlier in the week a familiar story was posted in this forum. Although many scorers are valuable, there are scorers that can be – contrary to what you hear in the sports media – easily replaced. Neil Paine – at Basketball-Reference – posted a comment that appeared to disagree with this argument (although if you read the comments you will see that the apparent disagreement may be because Neil mis-read the WoW post). These two posts led to a comment by Alex Konkel (aka the Sport Skeptic). In case people missed what Alex had to say, his comment has been re-posted below (thanks to Alex for sending this along).
The Value of Scoring
Recently Dave Berri put up a post examining what would have happened this past year if every NBA team lost its top scorer. Neil Paine then put up a post looking at what happens to team offense if they lose their leading scorer. The tone of the article seemed to say that it was refuting Berri’s post (although it became less clear in the comments), so I thought I would go through both and see exactly what panned out.
Berri starts his piece by noting that recently a few teams have lost prized scorers (such as Iverson, Carmelo, and Rudy Gay) and were predicted to fail afterward. Curiously, they did not fail. He points out that if you look at Wins Produced, this would not be surprising; scoring is not the only thing that leads to winning. Then he produced a table that showed how many wins each team was expected to have this season and how many the would have been expected to get if their leading scorer (determined by most points scored) were replaced by an average (WP48=.100) player. Berri notes that 19 teams would be worse off without their leading scorer, 4 very much so. 11 teams would be better with an average player, 1 very much so (take a bow, Bargnani).
The point, in general, is that leading scorers are generally above average, but only generally. If all you look at is scoring, then sometimes you’ll make a good decision and sometimes you won’t. You would be right to play your leading scorer broadly speaking; if you add up the numbers in the table teams would lose about 96 wins, or a little over 3 per team, if they replaced their scorer with an average player. But that isn’t true for all teams; Toronto should be looking for other options (as in almost any other option). The leading scorer is not critical for team success. In contrast, Berri’s previous post found that if you replaced a team’s *best* player as measured by Wins Produced with an average player, they would lose about 223 games (by my calculation from Berri’s table) or about 7.5 games per team. That’s over twice as many as you lose by replacing the leading scorer. Again: scoring is not always important, but good scorers can be good players.
In his post Neil Paine deemed a player to be the leading scorer if they played over half of the team’s games and had the highest points per game. I would imagine that would create the same set as Berri’s, although it might not 100% of the time. Paine also looked at the past 25 years instead of just this season. He gathered each team’s offensive rating (points per possession) in games played by the leading scorer and compared it to the offensive rating in games not played by the leading scorer. The ratings were adjusted by what an average team would have done against that opponent and weighted by the number of games missed by the leading scorer. Neil found that teams lose 2 points per 100 possessions when their leading scorer doesn’t play, although the number was only .9 per 100 this past year. Conclusion: the leading scorer is important.
Neil’s analysis is not the same as Berri’s. First, there’s no connection to winning. Berri’s WP analysis includes defense (even though plenty of people tell you it doesn’t!) to allow for a connection to team wins. Offensive rating alone ignores defense, so we don’t know how the teams did without their leading scorer in Neil’s analysis. Second, Berri ‘replaced’ the leading scorer with an average WP player while Neil looked at what happened with whoever actually replaced the leading scorer. While Neil’s analysis looks to be more practically useful, it also has flaws; it is easily the case that lots of things change when the leading scorer doesn’t play besides just the leading scorer not playing. For example, teams will rest players at the end of the year. That means that not only is the leading scorer not playing, but probably the top two or three scorers. Berri’s analysis explicitly removes only the leading scorer’s influence, whereas we don’t know that to be the case for Neil’s; the change in offensive rating is really more of a highest-possible-impact measure.
In response to that first issue, let’s see what a leading scorer might be worth in terms of wins for Neil. The average pace this season was 92.1, so to account for some overtime games let’s bump it to 95 (perhaps an overly large bump, but that isn’t too important). The important thing is that teams are losing less than 2 points per game if they lose their leading scorer; this past year it would be closer to .8. The general conversion for point differential to wins says that a point is worth 2.5 wins, so leading scorers this past year are worth 2 to 2.5 wins to an average team and over the past 25 years they are worth a little under 5 wins. Those numbers aren’t too different from the 3 we got from Berri.
So I don’t see that we have a disagreement here. According to Neil, leading scorers are worth 4 or 5 wins per season in terms of their offensive contribution beyond whoever comes off the bench; we would need to look at their defensive rating effect to see if that number goes up or down for their total contribution to winning. Dave Berri says the number is about 3 this year when you account for offense and defense compared to an average player. The analyses don’t strictly agree, but they’re also different approaches using different measures on different data sets. Given that, it’s impressive that they came out so close. But I think Berri’s overarching point still stands: it would be better to account for everything a player does than just look at his scoring.
– Alex Konkel (the Sport Skeptic)
Wages of Wins Journal 
MikeMother
May 19, 2011
The internet: where Violent Agreements are somehow all the rage :)
If I can ask a reverse question: I understand that WP48 is a regression from differential, but can the reverse be done? Can you take a WP48 team “score” (sum of WP48 * minutes for all players), and generate a theoretical differential?
The latest posts talk about minute allocation in the playoffs vs regular season, and the changes to wins this creates. Is there a way to go from, say, a 64 win team to a 68 win team and equate that back to a new differential?
In other words, given a reasonable guestimation at minute allocation, can we create a “theoretical” differential to use when comparing two playoff teams / twp theoretical teams (e.g. with and without Melo)?
It would also perhaps quell some of these type of arguments, as rather than comparing WP48 to X metric, where the value of WP48 is always questioned, converting it back to differential, a stat everyone (mostly) agrees on, might help change the debate from the merits of WP48 to an apples to apples comparison.
GovernorStephCurry
May 20, 2011
Oh jeez this metric is so bad. Apparently Dallas only loses 4 games more with an average player replacing Dirk…. the TWolves lose 20 more games with an average player replacing Love….fail, so far off.
fricktho
May 20, 2011
Dirk seems to be some sort of anomaly in this approach. If you swapped Love and Dirk I’m not sure there would be a noticeable difference in wins, realistically speaking. Love would likely regress quite a bit playing with a team of better players. Dirk would look like a superstar playing with a group of pure scrubs. I’m not saying their WP would flip-flop, but they would probably level out at worst. And Love would steal some rebounds from Chandler and Marion causing their WP to decrease which would lead to very little net gain for the team in this swap of PF’s.
sandalstraps
May 20, 2011
I hate to feed trolls, but…
The next step, Guv, is to provide an alternative theory and some data to support said theory. It’s easy to say “fail, so far off.” Anyone who can type can do that. But can you prove your case?
Or, to put it another way, “so far off” from what? And how do you know? How would replacing Dirk Nowitzki with an average player impact Dallas? How would replacing Kevin Love with an average player impact Minnesota? How do you know? What data will you use to support your case?
Say something. Say something interesting. Say something compelling. Support it with data. That would be a conversation worth having. But thus far your comment amounts to a preschool taunt, which is, I hope, beneath you.
Andrew
May 20, 2011
Gov,
Dirk shot 52% which is impressive, but only pulls down 7 rebounds a game. Love shoots 47%, and pulls down 15 boards (including 4.5 on the offensive end).
Every other stat between the two is a complete wash (steals, turnovers, 3pt shooting, blocks, etc.). Dirk is the better as a shooter but Love is twice as good on the boards.
Don’t confuse Dirk’s insane game 1 with how he plays every game.
Tommy_Grand
May 20, 2011
Good article.I think I could find room for Alex skillz in our legal mediation dept.
GovernorStephCurry
May 21, 2011
Dirk has the best regularized +/- in the NBA this season. He is a superstar. His individual rebounding isn’t great but thats because he doesn’t steal rebounds from teammates. He is a good team rebounder. Dirk can also play a bit of defense. Did i mention he’s amazing on offense?
KLove is a very nice player but he can’t create out of the post, or defend his position. Reboundings nice, but some defense would be more appreciated.
Greyberger
May 21, 2011
“Dirk shot 52% which is impressive, but only pulls down 7 rebounds a game. Love shoots 47%, and pulls down 15 boards (including 4.5 on the offensive end).
Every other stat between the two is a complete wash (steals, turnovers, 3pt shooting, blocks, etc.). Dirk is the better as a shooter but Love is twice as good on the boards.”
This kind of thinking is years out of date. If the only things we had to tell players apart were shooting efficiency, rebounds, steals, TO and blocks it would be very boring to analyze them.
There are, by now, HUNDREDS of measurements, metrics and points of articulation where we can meaningfully say the numbers point to a difference between Dirk and Love; limiting yourself to the box score mainstays is a fantasy-sports way of going about it. And let’s not get started about you reducing the whole of offense into player A’s efficiency versus player B’s…
Nerd Numbers
May 21, 2011
Grey,
Hmmm on the one hand I’m tempted to ask which 100 you’re thinking of but you are actually right. Now that said. . .
“limiting yourself to the box score mainstays is a fantasy-sports way of going about it”
I love your opinion! Now my question is if that’s true why you spend so much time on a niche box score blog (that Matt Johnson over on the Basketball-Reference Dave Berri Love comment thread pointed out isn’t even the most popular) instead of out broadening your horizon. I mean it kind of seems like telling us the car is the new thing while riding by on your horse. Just saying.
Andrew
May 21, 2011
Grey,
If I was using just PPG, I’d agree with you, but I intentionally used shooting percentages, which is a much bigger indication of a players offensive abilities. PPG is the batting average of Basketball stats, where shooting percentages are more like OBP and Slugging. They don’t lie and the do tend to tell a pretty accurate story. Maybe they don’t tell the entire truth, I’ll gladly concede that.
But just from reading these comments, Kevin Love “steals” rebounds, and Dirk is a better rebounder because he doesn’t “steal” rebounds, I have that right? So the fact that he double’s Dirk in rebounds is actually a negative?
That’s worse than using numbers, that is applying a massive bias for a players shortcomings in an attempt to make a player look better (to prove theory). “Dirk is a superstar” was even used a reason for why he’s better, so he’s better because he’s better…right, very scientific.
The reality is, when he’s shooting well Dirk is a fantastic offensive player. There’ nothing wrong with that, but because he doesn’t crash the boards, if he doesn’t shoot 55%+, then how else does he help your team? (I’ll also concede that he does get to the line a lot which has value)
Kevin Love on the other hands, pulls in double the rebounds, and whether people like it or not, forumlas and advance metrics use statistics to determine value, and Win Share uses rebounds. So even if he’s not scoring, he’s doing a lot of other work. Those 4.5 offensive boards that he’s “stealing” from his teammates are very significant.
There’s a reason why two of Chris Paul’s games against the Lakers beat Dirk’s WCF Game 1 performance, it’s because Chris Paul out rebounded, out stole and massive out assisted Dirk. Again, Dirks a great player, but it’s not a coincidence that once they got a better center who can rebound the ball that they are a contender again.
Suomynona
May 21, 2011
hasn’t arturo done some wins producedy work talking about how dirk might be underrated by the system?
Andrew
May 21, 2011
@Suomynona,
I can’t speak to that as I’m relatively new to this site, but there will always be outliers to any system. Specialists or players with unique skill sets can sometimes not rate well (think Bruce Bowen, Tayshaun Prince and Shane Battier come to mind).
The problem is when people try and bend systems to meet conventional wisdom. Thinking that since Dirk’s an all-star, so he HAS to be better can be dangerous thinking.
marparker
May 22, 2011
I think I’ve figured out the secret to NBA superstardom.
Play on a team thats good enough to win without you playing well so that when you play badly you can’t be blamed for a loss. Then when your team wins and you shoot alot receive all the credit.
Secondly, all these metrics are based off Dean Oliver’s work which gives exactly 20% credit to a team’s fg making ability. Yet, anytime its suggested that shooting ability doesn’t matter as much everyone panties get in a huff even though that was already established years ago.
I will probably just post this as my only comment repeatedly from now on.