2013-14 MVP Discussion Thread Pt II

[SideshowBob]

And BTW, it's not true that NPI 55 game sample of RAPM isn't usefull. That's more than half season and mysticbb cleary have said that 25% of season is enough to have relaible result in NPI RAPM.

I recall him saying this for PI RAPM, not NPI. In any given season I'm probably going to put more stock in the PI data even with the knowledge that its influenced by the previous year (not in the manner a multi-year model is).

^
Doc is right, SIdeshowBob. Basically only metrics with past season(s) element have LJ over Durant. But if we look only at what happened this year, then KD is clearly better. And it's not surprise considering LJ coasted first half of the season.

J.E.'s SPM is only using 2014 data, but I don't disagree with yours and Doc's general point.

[kabstah]

Some would argue undoubtedly that this is why you need prior-informing so that you don't overreact to small sample size. However, given Durant's clear superiority in box score metrics, LeBron showing clear signs that he's coasting, and how shocking it was that Durant did what he did with Westbrook out, to me all this is just showing why it's problematic to only look at prior-informed data.

The methods geared toward giving us the best results on average over the NBA do so with more convoluted approaches which at times can lead us far from the mark.

LeBron coasting is reflected in his 13-14 PI dRAPM being quite a bit worse compared to last year. This matches what we know about Miami as a team: the offense picked up right where it left off while the defense has taken a step back. Durant's PI RAPM makes sense compared to last year too, with a significant improvement on offense and hardly any change on defense.

It seems the only incongruity is LeBron's sizeable edge in oRAPM despite Durant's (crushing) advantage in box score stats. Of course, a disparity between RAPM and box score measures isn't uncommon, and that disparity isn't necessarily indicative of RAPM being overly biased towards the prior.

[Doctor MJ]

Some would argue undoubtedly that this is why you need prior-informing so that you don't overreact to small sample size. However, given Durant's clear superiority in box score metrics, LeBron showing clear signs that he's coasting, and how shocking it was that Durant did what he did with Westbrook out, to me all this is just showing why it's problematic to only look at prior-informed data.

The methods geared toward giving us the best results on average over the NBA do so with more convoluted approaches which at times can lead us far from the mark.

LeBron coasting is reflected in his 13-14 PI dRAPM being quite a bit worse compared to last year. This matches what we know about Miami as a team: the offense picked up right where it left off while the defense has taken a step back. Durant's PI RAPM makes sense compared to last year too, with a significant improvement on offense and hardly any change on defense.

It seems the only incongruity is LeBron's sizeable edge in oRAPM despite Durant's (crushing) advantage in box score stats. Of course, a disparity between RAPM and box score measures isn't uncommon, and that disparity isn't necessarily indicative of RAPM being overly biased towards the prior.

Using this site:
http://talkingpracticeblog.com/2012/10/24/hello-world/

PI numbers:
Durant +4.66, LeBron +0.77

NPI numbers:
LeBron +5.88, Durant +4.92

That is an insane difference for LeBron. The Durant stuff is well within range of things like study variance, but this is quite clearly a study where the year-to-year difference is extreme enough for LeBron that it makes the difference between him looking like a megastar and him looking pedestrian.

Do you have similar types of data that say something very different?

[SideshowBob]

Well, it's good you responded and gave these other points of you.

I'll re-phrase:

All stats that rely solely on this year's +/- data show clear signs of Durant being much more impactful than LeBron.

Data that relies on priors sees this edge completely disappear. Basically telling you that if you go in assuming LeBron's making everything happen based on previous year, the actual machine learning won't show any indication that there really isn't any correlation to speak of when looking just at this year.

Some would argue undoubtedly that this is why you need prior-informing so that you don't overreact to small sample size. However, given Durant's clear superiority in box score metrics, LeBron showing clear signs that he's coasting, and how shocking it was that Durant did what he did with Westbrook out, to me all this is just showing why it's problematic to only look at prior-informed data.

The methods geared toward giving us the best results on average over the NBA do so with more convoluted approaches which at times can lead us far from the mark.

But wasn't this exactly what you argued last season as well? That his 2013 numbers (+3.0, #33) were still underwhelming because they were being informed by 2012 (+3.4, #25). The 2012 numbers were underwhelming because they were informed by 2011 (+4.5, #12), etc. Why does this seem to be a continuing issue? The prior is an influencing factor to the extent that it provides context or gives the player a "role", but the regression is done using current season data, and of course the more in season data we have, the less weight the prior holds, and yet Durant continues to underwhelm? At some point, he should rank very high by this metric as well no? What's the issue?

First, I'd point to two things; one being the fact that he's consistently playing on strong teams, so there's an adjustment to be made (a +3.0 player on a +9.9 team is about a +4.0 player in a vacuum). The other being the heavy shared minutes between OKC's starters, particularly Durant/Westbrook. The regression has difficulty in attributing impact between the two.

The problem now is, he's had a good amount of minutes without Westbrook now, and he's still underwhelming (granted he looks better than any prior season). TalkingPractice's most recent set has him at +4.9, ranked 9th. So what's the explanation here?

Well it's not really a problem, its just that close to all of his impact is explained by the boxscore. This was the explanation before, this is what I argued last year, and this is the point mystic's been making for quite some time.

mystic in 2012:

How do you think Durant compares to Dirk, in terms of providing that spacing that you're talking about? It would seem that guys who are deadly shooters that move very well off the ball (Miller, Allen, Durant, Dirk) would provide spacing that would be hugely beneficial to an offense.

Well, it is different, because Durant, Miller or Allen are actually helping themselves more with that by getting into good shooting position than helping their team overall. I hardly see plays in which the Thunder use Durant as a decoy to get someone else a good shot. I don't see Harden getting better looks when being with Durant on the court, and while it made a difference for Westbrook last season (at least my impression), I think that Westbrook can break the defense down this season as well. For me the effect Harden has on Westbrook is bigger. Well, when Harden and Durant are on the court, Westbrook gets constantly those open midrange shots, that is a result of spacing. The effect is smaller for Durant than for Nowitzki. One of the reasons is the fight for the post position. While teams are trying to not give Nowitzki position in the post, they aren't that concerned about Durant on this part.

So, overall the ability to move without the ball is helping a team, but in the case of Durant, Miller or Allen we see a big part of that help already in their scoring efficiency. It is also a different effect when comparing a big and a small. So, when Durant plays the PF position, it becomes a bigger factor. We could also take a look again at Radmanovic and the 2008 Lakers. The offense got especially great when Radmanovic played PF next to Bynum or later Gasol. The reason here is that most times the more agile bigger defender is going out to the perimeter, in a lot of cases is that the defender being most important for the help defense. Now, when you combine that big-shooter-effect with being a post-threat, we are getting a player being really, really helpful for the offense even when he does not have the ball in his hands. The defense just has to react to the size and skills of such a player. Well, when you have a smaller player, the defense can just say that they give the smaller player a bit more room while defending the teammates more, that can work. But do that with Nowitzki and he kills your team from the midpost area.

Obviously, this effect depends on the teammates being able to use the opportunities given to them, but without the opportunities they aren't helping much anyway. A good example would be the difference here between Bryant and Nowitzki in terms of time with the ball. Nowitzki doesn't have the ball in his hand as often, but that also means that someone else has, likely a smaller guy, who can play to his strength with the ball. But when someone like Bryant is controlling the ball, how is a smaller player next to him in a position to play to his strength? That is an important aspect here, also with the Thunder and Durant. When Durant is moving without the ball, even if he doesn't give the same spacing effect, he still allows Westbrook to play with the ball. Imagine Durant would be more like Anthony and would demand more iso plays, Westbrook would likely be in a worse position. How much that effect is worth, is not easy to determine and I wouldn't try to put a number on it, but that is at least something we should consider as well.

mystic in 2013:

That is based off analyses from the last 1.5 seasons. In that the results indicate that James and Paul are +8 players, Durant, Ginobili, Duncan, Garnett +6 while Westbrook is about +3.5 (all per 100 possession).

That means the difference between Westbrook and Durant is smaller than the difference between Westbrook and Paul/James.

Overall, most of Durant's impact can be explained by his efficiency advantage, but he is not really making it much easier for his teammates. Even though his "playmaking" improved, he is not close to the impact Westbrook has on the roleplayers in terms of "making them better". And that's where Westbrook's impact also comes from. The collapsing defense makes it possible for Westbrook to find open teammates on cuts, on spot ups. That is not a bad thing, it is actually helping the team.

I also think that your argument about the raw +/- numbers is flawed. You said that Durant is now 100+ over a non-teammate, something which doesn't happen very often. But it is also purely due to circumstances, because the best players on the other top teams are playing a lot less minutes. If we adjust for the minutes here, Duncan would be 17 points away and Paul would be just 34 points away. Durant is not leading the league in +/- per 48 min, Ginobili does. The important thing to notice here is that the circumstances are the reason for the abnormality, not some incredible high level of impact by Durant. Cirumstances, which include an incredible streak of health. Since 2009/10 the Thunder missed their Top7 minutes per game getters in just 73 games combined (24 of that is by Sefolosha last season, the #6 in terms of mpg), in 273 games. The Lakers in this season alone have a combined 62 missed games by their Top7 minutes getters, if we exclude Steve Blake from that, who missed so far 35 games alone (and would have probably be 6th or 7th in min per game on the Lakers, but who knows). The Thunder streak of health is the main reason for them to be that good. They have 7 players being constantly healthy and being within 2 points per +/- per 48 min. That is the abnormality here, not the numbers for Durant alone (except of his high scoring efficiency). Right now Durant has +11.2, that puts him at #37 in the ranking since 2001. No, that it is not that outstanding as you made it out to be.
If you want to see some outstanding +/- numbers, go to 2003. Nowitzki was leading the league with 221 over the next best non-teammate (129 over Finley) Tim Duncan while even playing less minutes. Nowitzki was leading the league in +/- per 48 min for all players with 500+ min; that was outstanding.

Overall Durant is just the most visible profiteur of abnormal circumstances, which makes him look like he is much better than he really is. Yes, he is a great shooter and scorer, but his overall impact is not on the same level as that of Paul or James.

Okay, but noting that that makes zero sense to do given Durant's numbers skyrocketing this year unless you have reason to say they are mirage, which it appears you're going to try to do below.

His scoring efficiency increase makes 2pts per 100 possessions. On the other hand his defensive numbers (except for the boxscore) declined. Overall I get 1 point per 100 possession increase in terms of Durant's performance level in comparison to last season. Maybe that will further increase towards the end of the season, because so far he has still more minutes seen in 2012 as in 2013, but overall, just using a smaller sample and declare a player as completely changed, is a fallacy.
Remember, James started last season out with a PER of 33 after 32 games, for his remaining 30 games he averaged 28.2. Durant started out with insane Net +/- numbers, he is down to +11.5 per 100 possessions (11.4 on and -0.1 off).

Okay, but noting that you've gone on in great details about the profound Dirk has simply by his basketball gravity on the spacing and that Durant is the next generation's Dirk. You're fully aware that a player can make it easier for teammates without being a world-class playmaker.

There is HUGE difference between playing SF and PF in that style. When Durant plays PF, his impact in terms of spacing is MUCH bigger than as SF. Drawing out the PF is making it easier for players to cut and drive to the basket. Occupying the SF on the perimeter is just that, the SF is expected to be there.

So because something happened a decade ago that was even more noteworthy, everything else afterwards should be considered essentially noise?

No, that is not the point. The point is that Durant has such a lead based on circumstances, and when we adjust for those circumstances (minutes), that "noteworthy" numbers is going down to what we usually see. What we don't usually see is a team missing their 7 most important players so little as the Thunder have for the last 3.5 years.

The most important thing with all of this really is how OKC has changed from last year, and it's not like injuries were a crippling concern last year.

No, that is not the point. When your players are not injured, you can build up chemistry and have the coach figure out flexible working lineups. That's what Brooks is pretty good at, that's why you see 7 players in terms of +/- per 48 being so close together. Usually the difference even between starters is bigger, here we have a case of two additional bench players entering the picture. They have lineups without Durant playing which are beating the "s**t" out of their opponents. One of the big factors for the increased performance by the Thunder is better coaching, better lineup usage, and that is possible due to the fact that the coach has their most important players available at basically all time, not just for games, but also for practice. Crediting the Thunder improve (which is 2 points so far over last year) completely on Durant is not just not a good idea, but according to my stats analysis also not warranted.

So I'll ask directly: Why is it you're insisting on looking at Durant through a 1.5 year lens when there appears a major shift with the start of this season?

Because of the bigger sample, simple as that. Give me another half season and Durant keeping it up, we can talk about different things. But so far it all looks rather circumstantial. Or let me ask you a different question: Would you, for this season, really take Durant over Paul or James to build your championship team? Just based on their respective performance level?

mystic in 2014:

ardee, that is hardly a proof. It might just be that you and others are underrating the support, especially the coaching staff is doing an extremely good job at putting players into positions where they can succeed. The lineups are very well constructed and consistent. What we see for Durant is that his impact comes basically only from his incredible high efficiency scoring. He lacks the defensive impact as well as the impact to make it easier for his teammates. Just two weeks ago I tested some stuff for OKC and found that based on the efficiency without Durant his teammates are scoring about 2 TS% points below the expected value with him on the court despite the fact that they have a lower usage, which usually should lead to a better scoring efficiency. Also, a typical effect of a volume scorer is a lower turnover rate for the overall team with him on the court, but Durant isn't helping here at all. In fact, Durant is not making passes in which his teammates are per se in better position to score without creating, his playmaking is rather average, making the right passes, but with a couple of passing turnovers as well as not those "Steve Nash passes" included.

Overall the plays are putting Durant rather often in good positions to score, and then the Thunder can run those plays with a different player while still being effective with it (obviously not to the extend as with Durant, because that guy can just shoot the ball so well from basically everywhere). The Thunder also don't use Durant as a decoy like the Mavericks use Nowitzki for example, to open up the court, also, because Durant can't play the PF position as much. But when they have him play that position, they are clearly better offensively, and Durant's impact is really high. But the situations in which that works are limited due to his inability to defend the bigger players.

In the end Durant's incredible boxscore stats are also in part the result of a well-oiled team around him, which puts him into better positions, but Durant is not necessarily helping the other players to be in better positions.

I can probably find examples of myself arguing along a similar manner. None of that is to say that his impact isn't insane; it is. RAPM undervalues him and we can see why, but I don't think its all that far off from the truth. The biggest point I'd make personally is that I don't think he's made any major jumps from 2012->present. I think he's followed a pretty standard development pattern. I have him as a +5 (or ~5.5 by APM/RAPM scale) in 2012, and right around +6 (or +6.5 by APM/RAPM scale) in 2013 and a small tad higher in 2014. He's made improvements that would be expected from a player of his caliber, and considering he's only 25, continuing to follow a standard development curve would give him a very high peak somewhere in the next 4-5 years (maybe more who knows). But I really don't believe he's made any sort of major jump; he was already transcendent in 2012.

[Doctor MJ]

Well, it's good you responded and gave these other points of you.

I'll re-phrase:

All stats that rely solely on this year's +/- data show clear signs of Durant being much more impactful than LeBron.

Data that relies on priors sees this edge completely disappear. Basically telling you that if you go in assuming LeBron's making everything happen based on previous year, the actual machine learning won't show any indication that there really isn't any correlation to speak of when looking just at this year.

Some would argue undoubtedly that this is why you need prior-informing so that you don't overreact to small sample size. However, given Durant's clear superiority in box score metrics, LeBron showing clear signs that he's coasting, and how shocking it was that Durant did what he did with Westbrook out, to me all this is just showing why it's problematic to only look at prior-informed data.

The methods geared toward giving us the best results on average over the NBA do so with more convoluted approaches which at times can lead us far from the mark.

But wasn't this exactly what you argued last season as well? That his 2013 numbers (+3.0, #33) were still underwhelming because they were being informed by 2012 (+3.4, #25). The 2012 numbers were underwhelming because they were informed by 2011 (+4.5, #12), etc. Why does this seem to be a continuing issue?

Actually all of those facts show consistency in what I'm arguing, but clearly I need to explain better:

Durant as he made the leap to the "next level" was underrated by prior-informed metrics because they set expectation too low.
LeBron this year as he coasts is getting overrated by prior-informed metrics because they set expectation too high.

The continuing issue is that if a player's impact changes significantly between years then using priors will make analysis of that player less accurate. This doesn't mean that using priors in general is a bad idea - and unlike XRAPM style stats, it's pretty easy to "fix" the data in my mind when I'm reading the prior-influenced numbers - but in a scenario like this, I think evaluating LeBron based on anything using the prior could lead you far astray.

The prior is an influencing factor to the extent that it provides context or gives the player a "role", but the regression is done using current season data, and of course the more in season data we have, the less weight the prior holds, and yet Durant continues to underwhelm? At some point, he should rank very high by this metric as well no? What's the issue?

The fact that prior only sets context and yet LeBron looks drastically differently because of it shows you the issue. I honestly would not have thought it possible to see that major of a difference as we're seeing in the NPI vs PI numbers I listed for LeBron, and it disturbs me. I knew it was an issue to be taken seriously, but wow.

Re: Durant continues to underwhelm. That's not what I see really. I mean if you're looking at the numbers wondering why he isn't head and shoulders ahead of everyone not named LeBron, I understand some disappointment. He's right with the top tier though - or perhaps I should say, given the sample size we have thus far, I wouldn't look at the gaps between #1 and Durant as significant enough that I'd use it to draw any conclusions.

First, I'd point to two things; one being the fact that he's consistently playing on strong teams, so there's an adjustment to be made (a +3.0 player on a +9.9 team is about a +4.0 player in a vacuum). The other being the heavy shared minutes between OKC's starters, particularly Durant/Westbrook. The regression has difficulty in attributing impact between the two.

The problem now is, he's had a good amount of minutes without Westbrook now, and he's still underwhelming (granted he looks better than any prior season). TalkingPractice's most recent set has him at +4.9, ranked 9th. So what's the explanation here?

I think it's good to point out the multicollinearity issue with Durant & Westbrook. I see it too - additionally I think it's clear that Westbrook in the past has had direct impact on Durant's ability to help his team. Durant skyrocketed to a superstar in both box score & +/- stats in '09-10, and then took a considerable step back for the next couple years in +/- as Westbrook & Harden took on greater primacy. I think we've seen him break past that though in more recent times so the data isn't as problematic as it was even if it's clearly still not perfect.

Re: Ranked 9th. Again, I don't see this as a significant issue, and I suppose that means I should say something about how i use this stat, and stats in general. When I see that a stat is doing something that needs doing better than anything else available, I use it. That does not mean that player 8 by that metric has a true qualitative edge for me over player 9, or even that player 1 does.

Look at the 8 guys ranked ahead of Durant there:

3 of them are limited minutes guys (Collison, Johnson, Ginobili)
1 of them is a heavy multicollinearity concern (Iguodala)
2 of them are "in perfect situation for them to impact" guys (Nowitzki & Adridge)

That leaves LeBron & Paul, the other two members of the NBA's top 3 the past few years, and both of whom had a big RAPM edge at least until last year over Durant.

Add in general noise, I just look at all that and think "he's in the ballpark". Nothing alarms me about it.

Well it's not really a problem, its just that close to all of his impact is explained by the boxscore. This was the explanation before, this is what I argued last year, and this is the point mystic's been making for quite some time.

mystic in 2012:

How do you think Durant compares to Dirk, in terms of providing that spacing that you're talking about? It would seem that guys who are deadly shooters that move very well off the ball (Miller, Allen, Durant, Dirk) would provide spacing that would be hugely beneficial to an offense.

Well, it is different, because Durant, Miller or Allen are actually helping themselves more with that by getting into good shooting position than helping their team overall. I hardly see plays in which the Thunder use Durant as a decoy to get someone else a good shot. I don't see Harden getting better looks when being with Durant on the court, and while it made a difference for Westbrook last season (at least my impression), I think that Westbrook can break the defense down this season as well. For me the effect Harden has on Westbrook is bigger. Well, when Harden and Durant are on the court, Westbrook gets constantly those open midrange shots, that is a result of spacing. The effect is smaller for Durant than for Nowitzki. One of the reasons is the fight for the post position. While teams are trying to not give Nowitzki position in the post, they aren't that concerned about Durant on this part.

So, overall the ability to move without the ball is helping a team, but in the case of Durant, Miller or Allen we see a big part of that help already in their scoring efficiency. It is also a different effect when comparing a big and a small. So, when Durant plays the PF position, it becomes a bigger factor. We could also take a look again at Radmanovic and the 2008 Lakers. The offense got especially great when Radmanovic played PF next to Bynum or later Gasol. The reason here is that most times the more agile bigger defender is going out to the perimeter, in a lot of cases is that the defender being most important for the help defense. Now, when you combine that big-shooter-effect with being a post-threat, we are getting a player being really, really helpful for the offense even when he does not have the ball in his hands. The defense just has to react to the size and skills of such a player. Well, when you have a smaller player, the defense can just say that they give the smaller player a bit more room while defending the teammates more, that can work. But do that with Nowitzki and he kills your team from the midpost area.

Obviously, this effect depends on the teammates being able to use the opportunities given to them, but without the opportunities they aren't helping much anyway. A good example would be the difference here between Bryant and Nowitzki in terms of time with the ball. Nowitzki doesn't have the ball in his hand as often, but that also means that someone else has, likely a smaller guy, who can play to his strength with the ball. But when someone like Bryant is controlling the ball, how is a smaller player next to him in a position to play to his strength? That is an important aspect here, also with the Thunder and Durant. When Durant is moving without the ball, even if he doesn't give the same spacing effect, he still allows Westbrook to play with the ball. Imagine Durant would be more like Anthony and would demand more iso plays, Westbrook would likely be in a worse position. How much that effect is worth, is not easy to determine and I wouldn't try to put a number on it, but that is at least something we should consider as well.

mystic in 2013:

That is based off analyses from the last 1.5 seasons. In that the results indicate that James and Paul are +8 players, Durant, Ginobili, Duncan, Garnett +6 while Westbrook is about +3.5 (all per 100 possession).

That means the difference between Westbrook and Durant is smaller than the difference between Westbrook and Paul/James.

Overall, most of Durant's impact can be explained by his efficiency advantage, but he is not really making it much easier for his teammates. Even though his "playmaking" improved, he is not close to the impact Westbrook has on the roleplayers in terms of "making them better". And that's where Westbrook's impact also comes from. The collapsing defense makes it possible for Westbrook to find open teammates on cuts, on spot ups. That is not a bad thing, it is actually helping the team.

I also think that your argument about the raw +/- numbers is flawed. You said that Durant is now 100+ over a non-teammate, something which doesn't happen very often. But it is also purely due to circumstances, because the best players on the other top teams are playing a lot less minutes. If we adjust for the minutes here, Duncan would be 17 points away and Paul would be just 34 points away. Durant is not leading the league in +/- per 48 min, Ginobili does. The important thing to notice here is that the circumstances are the reason for the abnormality, not some incredible high level of impact by Durant. Cirumstances, which include an incredible streak of health. Since 2009/10 the Thunder missed their Top7 minutes per game getters in just 73 games combined (24 of that is by Sefolosha last season, the #6 in terms of mpg), in 273 games. The Lakers in this season alone have a combined 62 missed games by their Top7 minutes getters, if we exclude Steve Blake from that, who missed so far 35 games alone (and would have probably be 6th or 7th in min per game on the Lakers, but who knows). The Thunder streak of health is the main reason for them to be that good. They have 7 players being constantly healthy and being within 2 points per +/- per 48 min. That is the abnormality here, not the numbers for Durant alone (except of his high scoring efficiency). Right now Durant has +11.2, that puts him at #37 in the ranking since 2001. No, that it is not that outstanding as you made it out to be.
If you want to see some outstanding +/- numbers, go to 2003. Nowitzki was leading the league with 221 over the next best non-teammate (129 over Finley) Tim Duncan while even playing less minutes. Nowitzki was leading the league in +/- per 48 min for all players with 500+ min; that was outstanding.

Overall Durant is just the most visible profiteur of abnormal circumstances, which makes him look like he is much better than he really is. Yes, he is a great shooter and scorer, but his overall impact is not on the same level as that of Paul or James.

Okay, but noting that that makes zero sense to do given Durant's numbers skyrocketing this year unless you have reason to say they are mirage, which it appears you're going to try to do below.

His scoring efficiency increase makes 2pts per 100 possessions. On the other hand his defensive numbers (except for the boxscore) declined. Overall I get 1 point per 100 possession increase in terms of Durant's performance level in comparison to last season. Maybe that will further increase towards the end of the season, because so far he has still more minutes seen in 2012 as in 2013, but overall, just using a smaller sample and declare a player as completely changed, is a fallacy.
Remember, James started last season out with a PER of 33 after 32 games, for his remaining 30 games he averaged 28.2. Durant started out with insane Net +/- numbers, he is down to +11.5 per 100 possessions (11.4 on and -0.1 off).

Okay, but noting that you've gone on in great details about the profound Dirk has simply by his basketball gravity on the spacing and that Durant is the next generation's Dirk. You're fully aware that a player can make it easier for teammates without being a world-class playmaker.

There is HUGE difference between playing SF and PF in that style. When Durant plays PF, his impact in terms of spacing is MUCH bigger than as SF. Drawing out the PF is making it easier for players to cut and drive to the basket. Occupying the SF on the perimeter is just that, the SF is expected to be there.

So because something happened a decade ago that was even more noteworthy, everything else afterwards should be considered essentially noise?

No, that is not the point. The point is that Durant has such a lead based on circumstances, and when we adjust for those circumstances (minutes), that "noteworthy" numbers is going down to what we usually see. What we don't usually see is a team missing their 7 most important players so little as the Thunder have for the last 3.5 years.

The most important thing with all of this really is how OKC has changed from last year, and it's not like injuries were a crippling concern last year.

No, that is not the point. When your players are not injured, you can build up chemistry and have the coach figure out flexible working lineups. That's what Brooks is pretty good at, that's why you see 7 players in terms of +/- per 48 being so close together. Usually the difference even between starters is bigger, here we have a case of two additional bench players entering the picture. They have lineups without Durant playing which are beating the "s**t" out of their opponents. One of the big factors for the increased performance by the Thunder is better coaching, better lineup usage, and that is possible due to the fact that the coach has their most important players available at basically all time, not just for games, but also for practice. Crediting the Thunder improve (which is 2 points so far over last year) completely on Durant is not just not a good idea, but according to my stats analysis also not warranted.

So I'll ask directly: Why is it you're insisting on looking at Durant through a 1.5 year lens when there appears a major shift with the start of this season?

Because of the bigger sample, simple as that. Give me another half season and Durant keeping it up, we can talk about different things. But so far it all looks rather circumstantial. Or let me ask you a different question: Would you, for this season, really take Durant over Paul or James to build your championship team? Just based on their respective performance level?

mystic in 2014:

ardee, that is hardly a proof. It might just be that you and others are underrating the support, especially the coaching staff is doing an extremely good job at putting players into positions where they can succeed. The lineups are very well constructed and consistent. What we see for Durant is that his impact comes basically only from his incredible high efficiency scoring. He lacks the defensive impact as well as the impact to make it easier for his teammates. Just two weeks ago I tested some stuff for OKC and found that based on the efficiency without Durant his teammates are scoring about 2 TS% points below the expected value with him on the court despite the fact that they have a lower usage, which usually should lead to a better scoring efficiency. Also, a typical effect of a volume scorer is a lower turnover rate for the overall team with him on the court, but Durant isn't helping here at all. In fact, Durant is not making passes in which his teammates are per se in better position to score without creating, his playmaking is rather average, making the right passes, but with a couple of passing turnovers as well as not those "Steve Nash passes" included.

Overall the plays are putting Durant rather often in good positions to score, and then the Thunder can run those plays with a different player while still being effective with it (obviously not to the extend as with Durant, because that guy can just shoot the ball so well from basically everywhere). The Thunder also don't use Durant as a decoy like the Mavericks use Nowitzki for example, to open up the court, also, because Durant can't play the PF position as much. But when they have him play that position, they are clearly better offensively, and Durant's impact is really high. But the situations in which that works are limited due to his inability to defend the bigger players.

In the end Durant's incredible boxscore stats are also in part the result of a well-oiled team around him, which puts him into better positions, but Durant is not necessarily helping the other players to be in better positions.

I can probably find examples of myself arguing along a similar manner. None of that is to say that his impact isn't insane; it is. RAPM undervalues him and we can see why, but I don't think its all that far off from the truth. The biggest point I'd make personally is that I don't think he's made any major jumps from 2012->present. I think he's followed a pretty standard development pattern. I have him as a +5 (or ~5.5 by APM/RAPM scale) in 2012, and right around +6 (or +6.5 by APM/RAPM scale) in 2013 and a small tad higher in 2014. He's made improvements that would be expected from a player of his caliber, and considering he's only 25, continuing to follow a standard development curve would give him a very high peak somewhere in the next 4-5 years (maybe more who knows). But I really don't believe he's made any sort of major jump; he was already transcendent in 2012.

So there's a lot here. Good for people to read. My most salient thoughts:

An edge Dirk used to have on Durant was that he a direct manipulator of space as opposed to an indirect. He would act as a hub that would then either choose to attack or pass. Durant played more pure off-ball. While this isn't entirely in the past tense, I think we've clearly seen that Durant can play as more of a director now - it's just that Westbrook's place out there has always seem to require that Westbrook take on more of that role. This leads to a case where when the two play at times Durant is literally less valuable due to a poor fit, but which can thus essentially "underrate" Durant relative to his capabilities.

Part of the conversation here was about Durant's massive raw +/- lead last year. That's a good thing to bring up.

So back then I was saying "that's a really big deal", "it's not something that can just randomly go away". Of course by season's end it did go away, but not randomly, LeBron led Miami on an incredible winning streak. I maintain that Durant's lead was a big deal, but LeBron earned that MVP by doing something that was an even bigger deal.

Secondly, mystic talks about team context, and of course he's not wrong. I would never make an MVP list just based on raw +/- numbers. My point was always though that while less outlying raw numbers doesn't necessarily mean anything, clear outliers DO mean something. Was that enough to say with certainty that Durant should have been the leader in the MVP race early-to-midseason last year? Not certainly no, but it was enough that it should grab everyone's attention, particularly given that it was a drastic improvement over the prior year DESPITE Harden no longer being there.

Re: don't think he's made major leaps since 2012. I don't see why you say that. Forget the +/- for a second. His box score data shows a massive improvement too. I mean if you want to say "He hasn't improved, he's just been put in a role with higher primacy and better fit", fine, but I hardly think that in 2012 you were sitting here saying "Durant's fully capable of 30 PPG on 64% TS, he's just scoring much less on much weaker efficiency because of context".

Either he's taken a leap, or he's revealed himself to be a "leap" ahead of what he'd shown before. Do you honestly disagree with it being one of those two scenarios?

[KING JAMES1978]

Durant is still MVP but Lebron is damn close second.
Durant deserve it but some people act like KD was a beast and Lebron was nothing.
LBJ has 58% fg career high.So you can't say that he sucks.
But is a very tough interesting race this year.

[MisterWestside]

In the end Durant's incredible boxscore stats I can probably find examples of myself arguing along a similar manner. None of that is to say that his impact isn't insane; it is. RAPM undervalues him and we can see why, but I don't think its all that far off from the truth. The biggest point I'd make personally is that I don't think he's made any major jumps from 2012->present. I think he's followed a pretty standard development pattern. I have him as a +5 (or ~5.5 by APM/RAPM scale) in 2012, and right around +6 (or +6.5 by APM/RAPM scale) in 2013 and a small tad higher in 2014. He's made improvements that would be expected from a player of his caliber, and considering he's only 25, continuing to follow a standard development curve would give him a very high peak somewhere in the next 4-5 years (maybe more who knows). But I really don't believe he's made any sort of major jump; he was already transcendent in 2012.

In terms of global goodness, how do you rate this year's version of James and Durant? I have James as a +9 (+7 offense, +2 defense) and Durant as +7-7.5. Then again I tend to be liberal with the numbers so you probably rank both of them a bit lower

[SideshowBob]

The fact that prior only sets context and yet LeBron looks drastically differently because of it shows you the issue. I honestly would not have thought it possible to see that major of a difference as we're seeing in the NPI vs PI numbers I listed for LeBron, and it disturbs me. I knew it was an issue to be taken seriously, but wow.

Indeed, it is pretty striking. I wish we had slightly more extensive NPI data, I'd want to see the offense/defense splits.

Back to Durant though, what I'm trying to get at is that if last year the argument was that the prior season's data was holding him back in a PI RAPM study, how are we still making the same argument this year? At some point, players show drastic shifts in PI studies year on year (James from 08->09 went from +6.2 to +9.3 and we know he made a jump that year, Nowitzki 10->11 goes from +3.4 to +8.6, we know that there were some fit/context changes). I'm not looking at the specificity of those examples, I'm just trying to point out that drastic year-on-year shifts have been observed with prior-informed data, so what would be the reason for it not showing up for Durant if he's making these kinds of leaps? If 2013 was initially held down by 2012, shouldn't a 2014 informed by 2013 show some relative breakout?

Re: Durant continues to underwhelm. That's not what I see really. I mean if you're looking at the numbers wondering why he isn't head and shoulders ahead of everyone not named LeBron, I understand some disappointment. He's right with the top tier though - or perhaps I should say, given the sample size we have thus far, I wouldn't look at the gaps between #1 and Durant as significant enough that I'd use it to draw any conclusions.

I think it's good to point out the multicollinearity issue with Durant & Westbrook. I see it too - additionally I think it's clear that Westbrook in the past has had direct impact on Durant's ability to help his team. Durant skyrocketed to a superstar in both box score & +/- stats in '09-10, and then took a considerable step back for the next couple years in +/- as Westbrook & Harden took on greater primacy. I think we've seen him break past that though in more recent times so the data isn't as problematic as it was even if it's clearly still not perfect.

Re: Ranked 9th. Again, I don't see this as a significant issue, and I suppose that means I should say something about how i use this stat, and stats in general. When I see that a stat is doing something that needs doing better than anything else available, I use it. That does not mean that player 8 by that metric has a true qualitative edge for me over player 9, or even that player 1 does.

Look at the 8 guys ranked ahead of Durant there:

3 of them are limited minutes guys (Collison, Johnson, Ginobili)
1 of them is a heavy multicollinearity concern (Iguodala)
2 of them are "in perfect situation for them to impact" guys (Nowitzki & Adridge)

That leaves LeBron & Paul, the other two members of the NBA's top 3 the past few years, and both of whom had a big RAPM edge at least until last year over Durant.

Add in general noise, I just look at all that and think "he's in the ballpark". Nothing alarms me about it.

That's cool, we're on the same page here, though I'd point out that both James (+7.3) and Paul (+5.2) still held a considerable edge over him last year (+3.0).

Snip

So there's a lot here. Good for people to read. My most salient thoughts:

An edge Dirk used to have on Durant was that he a direct manipulator of space as opposed to an indirect. He would act as a hub that would then either choose to attack or pass. Durant played more pure off-ball. While this isn't entirely in the past tense, I think we've clearly seen that Durant can play as more of a director now - it's just that Westbrook's place out there has always seem to require that Westbrook take on more of that role. This leads to a case where when the two play at times Durant is literally less valuable due to a poor fit, but which can thus essentially "underrate" Durant relative to his capabilities.

I agree there, that's the point I was getting at earlier. It's beyond mere multicollinearity, Westbrook's presence obviously marginalizes some of Durant's on-ball role. That being said, I think its also reasonable to say that OKC is still better suited with Westbrook in that role than Durant; he plays it better.

Secondly, mystic talks about team context, and of course he's not wrong. I would never make an MVP list just based on raw +/- numbers. My point was always though that while less outlying raw numbers doesn't necessarily mean anything, clear outliers DO mean something. Was that enough to say with certainty that Durant should have been the leader in the MVP race early-to-midseason last year? Not certainly no, but it was enough that it should grab everyone's attention, particularly given that it was a drastic improvement over the prior year DESPITE Harden no longer being there.

No quips here

Re: don't think he's made major leaps since 2012. I don't see why you say that.

I think I should rephrase: I don't think he's made leaps that are outside of the ordinary development cycle of a superstar player. The difference I see in terms of ability between 2012 and 2014 is ~1.5-2 points,

Forget the +/- for a second. His box score data shows a massive improvement too.

Either he's taken a leap, or he's revealed himself to be a "leap" ahead of what he'd shown before. Do you honestly disagree with it being one of those two scenarios?[/quote]

Oh I'm not considering +/- or the box-score here. I'm talking strictly what I see on the court, which is greatly improved ball-handling, this year he's added some upper body weight that hasn't come at the cost of explosion and in effect, he's become a much better finisher (and he was already the cream of the crop), his passing has improved and he's taken a more active playmaking role (but isn't an elite playmaker, and has become even more turnover prone as a result), he's gotten better at getting himself into his spots (this was a major issue in 2010 and less so in 2011), and he's worked on his shot creation on the ball. All that adds up to what I consider to be a 1.5-2 point improvement in him as a player, all of that is also explained by the box-score. That's pretty significant; he's definitely entering all-time territory now, but again, that's not out of line with what we see from players of his caliber at his age (Jordan from 85/86 -> 88/89, James from 08->09 was an even bigger jump).

I mean if you want to say "He hasn't improved, he's just been put in a role with higher primacy and better fit", fine, but I hardly think that in 2012 you were sitting here saying "Durant's fully capable of 30 PPG on 64% TS, he's just scoring much less on much weaker efficiency because of context".

I think its a combination of both; standard improvements and current situation, which is why I don't want to ever make too much of any striking stat outliers. I think just looking at it as 30PPG on 64% TS is problematic, but I think saying that he would go on to be capable of it back in 2012 would have been more than reasonable. The lockout league environment is definitely something to consider here, 61% in that season is not quite the same as 61% now, and he spent the second half of the season and the playoffs at around 62-63% anyway. A bit more volume with Westbrook/Harden now out coupled with the improvements he should have been expected to make and what he's doing box-score wise right now isn't that outlandish at all.

[SideshowBob]

In terms of global goodness, how do you rate this year's version of James and Durant? I have James as a +9 (+7 offense, +2 defense) and Durant as +7-7.5. Then again I tend to be liberal with the numbers so you probably rank both of them a bit lower

Maybe about half a point lower for both, but yeah that's about where I have them right now. Lebron's defense has been on the rise though, so that's something to watch for.

[NinjaSheppard]

LeBron being a +2 defender this year is really really really generous

[kayess]

Well, it's good you responded and gave these other points of you.

I'll re-phrase:

All stats that rely solely on this year's +/- data show clear signs of Durant being much more impactful than LeBron.

Data that relies on priors sees this edge completely disappear. Basically telling you that if you go in assuming LeBron's making everything happen based on previous year, the actual machine learning won't show any indication that there really isn't any correlation to speak of when looking just at this year.

Some would argue undoubtedly that this is why you need prior-informing so that you don't overreact to small sample size. However, given Durant's clear superiority in box score metrics, LeBron showing clear signs that he's coasting, and how shocking it was that Durant did what he did with Westbrook out, to me all this is just showing why it's problematic to only look at prior-informed data.

The methods geared toward giving us the best results on average over the NBA do so with more convoluted approaches which at times can lead us far from the mark.

But wasn't this exactly what you argued last season as well? That his 2013 numbers (+3.0, #33) were still underwhelming because they were being informed by 2012 (+3.4, #25). The 2012 numbers were underwhelming because they were informed by 2011 (+4.5, #12), etc. Why does this seem to be a continuing issue?

Actually all of those facts show consistency in what I'm arguing, but clearly I need to explain better:

Durant as he made the leap to the "next level" was underrated by prior-informed metrics because they set expectation too low.
LeBron this year as he coasts is getting overrated by prior-informed metrics because they set expectation too high.

The continuing issue is that if a player's impact changes significantly between years then using priors will make analysis of that player less accurate. This doesn't mean that using priors in general is a bad idea - and unlike XRAPM style stats, it's pretty easy to "fix" the data in my mind when I'm reading the prior-influenced numbers - but in a scenario like this, I think evaluating LeBron based on anything using the prior could lead you far astray.

The prior is an influencing factor to the extent that it provides context or gives the player a "role", but the regression is done using current season data, and of course the more in season data we have, the less weight the prior holds, and yet Durant continues to underwhelm? At some point, he should rank very high by this metric as well no? What's the issue?

The fact that prior only sets context and yet LeBron looks drastically differently because of it shows you the issue. I honestly would not have thought it possible to see that major of a difference as we're seeing in the NPI vs PI numbers I listed for LeBron, and it disturbs me. I knew it was an issue to be taken seriously, but wow.

Re: Durant continues to underwhelm. That's not what I see really. I mean if you're looking at the numbers wondering why he isn't head and shoulders ahead of everyone not named LeBron, I understand some disappointment. He's right with the top tier though - or perhaps I should say, given the sample size we have thus far, I wouldn't look at the gaps between #1 and Durant as significant enough that I'd use it to draw any conclusions.

First, I'd point to two things; one being the fact that he's consistently playing on strong teams, so there's an adjustment to be made (a +3.0 player on a +9.9 team is about a +4.0 player in a vacuum). The other being the heavy shared minutes between OKC's starters, particularly Durant/Westbrook. The regression has difficulty in attributing impact between the two.

The problem now is, he's had a good amount of minutes without Westbrook now, and he's still underwhelming (granted he looks better than any prior season). TalkingPractice's most recent set has him at +4.9, ranked 9th. So what's the explanation here?

I think it's good to point out the multicollinearity issue with Durant & Westbrook. I see it too - additionally I think it's clear that Westbrook in the past has had direct impact on Durant's ability to help his team. Durant skyrocketed to a superstar in both box score & +/- stats in '09-10, and then took a considerable step back for the next couple years in +/- as Westbrook & Harden took on greater primacy. I think we've seen him break past that though in more recent times so the data isn't as problematic as it was even if it's clearly still not perfect.

Re: Ranked 9th. Again, I don't see this as a significant issue, and I suppose that means I should say something about how i use this stat, and stats in general. When I see that a stat is doing something that needs doing better than anything else available, I use it. That does not mean that player 8 by that metric has a true qualitative edge for me over player 9, or even that player 1 does.

Look at the 8 guys ranked ahead of Durant there:

3 of them are limited minutes guys (Collison, Johnson, Ginobili)
1 of them is a heavy multicollinearity concern (Iguodala)
2 of them are "in perfect situation for them to impact" guys (Nowitzki & Adridge)

That leaves LeBron & Paul, the other two members of the NBA's top 3 the past few years, and both of whom had a big RAPM edge at least until last year over Durant.

Add in general noise, I just look at all that and think "he's in the ballpark". Nothing alarms me about it.

Well it's not really a problem, its just that close to all of his impact is explained by the boxscore. This was the explanation before, this is what I argued last year, and this is the point mystic's been making for quite some time.

mystic in 2012:

Well, it is different, because Durant, Miller or Allen are actually helping themselves more with that by getting into good shooting position than helping their team overall. I hardly see plays in which the Thunder use Durant as a decoy to get someone else a good shot. I don't see Harden getting better looks when being with Durant on the court, and while it made a difference for Westbrook last season (at least my impression), I think that Westbrook can break the defense down this season as well. For me the effect Harden has on Westbrook is bigger. Well, when Harden and Durant are on the court, Westbrook gets constantly those open midrange shots, that is a result of spacing. The effect is smaller for Durant than for Nowitzki. One of the reasons is the fight for the post position. While teams are trying to not give Nowitzki position in the post, they aren't that concerned about Durant on this part.

So, overall the ability to move without the ball is helping a team, but in the case of Durant, Miller or Allen we see a big part of that help already in their scoring efficiency. It is also a different effect when comparing a big and a small. So, when Durant plays the PF position, it becomes a bigger factor. We could also take a look again at Radmanovic and the 2008 Lakers. The offense got especially great when Radmanovic played PF next to Bynum or later Gasol. The reason here is that most times the more agile bigger defender is going out to the perimeter, in a lot of cases is that the defender being most important for the help defense. Now, when you combine that big-shooter-effect with being a post-threat, we are getting a player being really, really helpful for the offense even when he does not have the ball in his hands. The defense just has to react to the size and skills of such a player. Well, when you have a smaller player, the defense can just say that they give the smaller player a bit more room while defending the teammates more, that can work. But do that with Nowitzki and he kills your team from the midpost area.

Obviously, this effect depends on the teammates being able to use the opportunities given to them, but without the opportunities they aren't helping much anyway. A good example would be the difference here between Bryant and Nowitzki in terms of time with the ball. Nowitzki doesn't have the ball in his hand as often, but that also means that someone else has, likely a smaller guy, who can play to his strength with the ball. But when someone like Bryant is controlling the ball, how is a smaller player next to him in a position to play to his strength? That is an important aspect here, also with the Thunder and Durant. When Durant is moving without the ball, even if he doesn't give the same spacing effect, he still allows Westbrook to play with the ball. Imagine Durant would be more like Anthony and would demand more iso plays, Westbrook would likely be in a worse position. How much that effect is worth, is not easy to determine and I wouldn't try to put a number on it, but that is at least something we should consider as well.

mystic in 2013:

That is based off analyses from the last 1.5 seasons. In that the results indicate that James and Paul are +8 players, Durant, Ginobili, Duncan, Garnett +6 while Westbrook is about +3.5 (all per 100 possession).

That means the difference between Westbrook and Durant is smaller than the difference between Westbrook and Paul/James.

Overall, most of Durant's impact can be explained by his efficiency advantage, but he is not really making it much easier for his teammates. Even though his "playmaking" improved, he is not close to the impact Westbrook has on the roleplayers in terms of "making them better". And that's where Westbrook's impact also comes from. The collapsing defense makes it possible for Westbrook to find open teammates on cuts, on spot ups. That is not a bad thing, it is actually helping the team.

I also think that your argument about the raw +/- numbers is flawed. You said that Durant is now 100+ over a non-teammate, something which doesn't happen very often. But it is also purely due to circumstances, because the best players on the other top teams are playing a lot less minutes. If we adjust for the minutes here, Duncan would be 17 points away and Paul would be just 34 points away. Durant is not leading the league in +/- per 48 min, Ginobili does. The important thing to notice here is that the circumstances are the reason for the abnormality, not some incredible high level of impact by Durant. Cirumstances, which include an incredible streak of health. Since 2009/10 the Thunder missed their Top7 minutes per game getters in just 73 games combined (24 of that is by Sefolosha last season, the #6 in terms of mpg), in 273 games. The Lakers in this season alone have a combined 62 missed games by their Top7 minutes getters, if we exclude Steve Blake from that, who missed so far 35 games alone (and would have probably be 6th or 7th in min per game on the Lakers, but who knows). The Thunder streak of health is the main reason for them to be that good. They have 7 players being constantly healthy and being within 2 points per +/- per 48 min. That is the abnormality here, not the numbers for Durant alone (except of his high scoring efficiency). Right now Durant has +11.2, that puts him at #37 in the ranking since 2001. No, that it is not that outstanding as you made it out to be.
If you want to see some outstanding +/- numbers, go to 2003. Nowitzki was leading the league with 221 over the next best non-teammate (129 over Finley) Tim Duncan while even playing less minutes. Nowitzki was leading the league in +/- per 48 min for all players with 500+ min; that was outstanding.

Overall Durant is just the most visible profiteur of abnormal circumstances, which makes him look like he is much better than he really is. Yes, he is a great shooter and scorer, but his overall impact is not on the same level as that of Paul or James.

His scoring efficiency increase makes 2pts per 100 possessions. On the other hand his defensive numbers (except for the boxscore) declined. Overall I get 1 point per 100 possession increase in terms of Durant's performance level in comparison to last season. Maybe that will further increase towards the end of the season, because so far he has still more minutes seen in 2012 as in 2013, but overall, just using a smaller sample and declare a player as completely changed, is a fallacy.
Remember, James started last season out with a PER of 33 after 32 games, for his remaining 30 games he averaged 28.2. Durant started out with insane Net +/- numbers, he is down to +11.5 per 100 possessions (11.4 on and -0.1 off).



There is HUGE difference between playing SF and PF in that style. When Durant plays PF, his impact in terms of spacing is MUCH bigger than as SF. Drawing out the PF is making it easier for players to cut and drive to the basket. Occupying the SF on the perimeter is just that, the SF is expected to be there.



No, that is not the point. The point is that Durant has such a lead based on circumstances, and when we adjust for those circumstances (minutes), that "noteworthy" numbers is going down to what we usually see. What we don't usually see is a team missing their 7 most important players so little as the Thunder have for the last 3.5 years.



No, that is not the point. When your players are not injured, you can build up chemistry and have the coach figure out flexible working lineups. That's what Brooks is pretty good at, that's why you see 7 players in terms of +/- per 48 being so close together. Usually the difference even between starters is bigger, here we have a case of two additional bench players entering the picture. They have lineups without Durant playing which are beating the "s**t" out of their opponents. One of the big factors for the increased performance by the Thunder is better coaching, better lineup usage, and that is possible due to the fact that the coach has their most important players available at basically all time, not just for games, but also for practice. Crediting the Thunder improve (which is 2 points so far over last year) completely on Durant is not just not a good idea, but according to my stats analysis also not warranted.



Because of the bigger sample, simple as that. Give me another half season and Durant keeping it up, we can talk about different things. But so far it all looks rather circumstantial. Or let me ask you a different question: Would you, for this season, really take Durant over Paul or James to build your championship team? Just based on their respective performance level?

mystic in 2014:

ardee, that is hardly a proof. It might just be that you and others are underrating the support, especially the coaching staff is doing an extremely good job at putting players into positions where they can succeed. The lineups are very well constructed and consistent. What we see for Durant is that his impact comes basically only from his incredible high efficiency scoring. He lacks the defensive impact as well as the impact to make it easier for his teammates. Just two weeks ago I tested some stuff for OKC and found that based on the efficiency without Durant his teammates are scoring about 2 TS% points below the expected value with him on the court despite the fact that they have a lower usage, which usually should lead to a better scoring efficiency. Also, a typical effect of a volume scorer is a lower turnover rate for the overall team with him on the court, but Durant isn't helping here at all. In fact, Durant is not making passes in which his teammates are per se in better position to score without creating, his playmaking is rather average, making the right passes, but with a couple of passing turnovers as well as not those "Steve Nash passes" included.

Overall the plays are putting Durant rather often in good positions to score, and then the Thunder can run those plays with a different player while still being effective with it (obviously not to the extend as with Durant, because that guy can just shoot the ball so well from basically everywhere). The Thunder also don't use Durant as a decoy like the Mavericks use Nowitzki for example, to open up the court, also, because Durant can't play the PF position as much. But when they have him play that position, they are clearly better offensively, and Durant's impact is really high. But the situations in which that works are limited due to his inability to defend the bigger players.

In the end Durant's incredible boxscore stats are also in part the result of a well-oiled team around him, which puts him into better positions, but Durant is not necessarily helping the other players to be in better positions.

I can probably find examples of myself arguing along a similar manner. None of that is to say that his impact isn't insane; it is. RAPM undervalues him and we can see why, but I don't think its all that far off from the truth. The biggest point I'd make personally is that I don't think he's made any major jumps from 2012->present. I think he's followed a pretty standard development pattern. I have him as a +5 (or ~5.5 by APM/RAPM scale) in 2012, and right around +6 (or +6.5 by APM/RAPM scale) in 2013 and a small tad higher in 2014. He's made improvements that would be expected from a player of his caliber, and considering he's only 25, continuing to follow a standard development curve would give him a very high peak somewhere in the next 4-5 years (maybe more who knows). But I really don't believe he's made any sort of major jump; he was already transcendent in 2012.

So there's a lot here. Good for people to read. My most salient thoughts:

An edge Dirk used to have on Durant was that he a direct manipulator of space as opposed to an indirect. He would act as a hub that would then either choose to attack or pass. Durant played more pure off-ball. While this isn't entirely in the past tense, I think we've clearly seen that Durant can play as more of a director now - it's just that Westbrook's place out there has always seem to require that Westbrook take on more of that role. This leads to a case where when the two play at times Durant is literally less valuable due to a poor fit, but which can thus essentially "underrate" Durant relative to his capabilities.

Part of the conversation here was about Durant's massive raw +/- lead last year. That's a good thing to bring up.

So back then I was saying "that's a really big deal", "it's not something that can just randomly go away". Of course by season's end it did go away, but not randomly, LeBron led Miami on an incredible winning streak. I maintain that Durant's lead was a big deal, but LeBron earned that MVP by doing something that was an even bigger deal.

Secondly, mystic talks about team context, and of course he's not wrong. I would never make an MVP list just based on raw +/- numbers. My point was always though that while less outlying raw numbers doesn't necessarily mean anything, clear outliers DO mean something. Was that enough to say with certainty that Durant should have been the leader in the MVP race early-to-midseason last year? Not certainly no, but it was enough that it should grab everyone's attention, particularly given that it was a drastic improvement over the prior year DESPITE Harden no longer being there.

Re: don't think he's made major leaps since 2012. I don't see why you say that. Forget the +/- for a second. His box score data shows a massive improvement too. I mean if you want to say "He hasn't improved, he's just been put in a role with higher primacy and better fit", fine, but I hardly think that in 2012 you were sitting here saying "Durant's fully capable of 30 PPG on 64% TS, he's just scoring much less on much weaker efficiency because of context".

Either he's taken a leap, or he's revealed himself to be a "leap" ahead of what he'd shown before. Do you honestly disagree with it being one of those two scenarios?

Great back and forth between you and SSB.

FTR Doc, does this mean you think it's not that Durant's decisively supplanted LeBron as the best player on the planet, but rather ,a combination of LBJ's coasting, Durant's incredible stretch without Westbrook, among other things, leading to Durant's massive edge in NPI +/- and box-score stats?

I remember at the end of last year, OKC without Durant was +5 per 100 IIRC (+12 with him - so a still great +7 improvement)? Which was absolutely mindblowing to me, because bar Westbrook I thought the rest of OKC were scrubs unable to create their own shot, etc.

Miami without LeBron? -2 or -3, and +12 with him. Maybe we are underrating the support Durant has, but I honestly can't believe it since I think Brooks is an idiot. But hey, what do I know.

[Doctor MJ]

Has Durant supplanted LeBron?

I'll let you know in June.

[MisterWestside]

LeBron being a +2 defender this year is really really really generous

I'm basing it on what I think he'll be at by the start of the postseason.

[kayess]

Has Durant supplanted LeBron?

I'll let you know in June.

You tease, you. At least tell us what you think at the moment!

ardee has mentioned (and I concur) that this is his "09 LeBron" season, and comparing those two seasons I'd still say LeBron did more with less.

It's crazy: if it comes down to LBJ vs. KD II and LBJ wins, that means Oden gets a ring before Durant. If I were Durant I'd lock myself in a room for a week if that happened

[EArl]

Lebron flirted with a triple double tonight so not a bad game for him.

[LBJ-ITALY]

And he played onlt 30min!

[ThePersianFreak]

Lebron flirted with a triple double tonight so not a bad game for him.

He was flawless, and did'nt play in 4th quarter.

[CJ_18]

FTR Doc, does this mean you think it's not that Durant's decisively supplanted LeBron as the best player on the planet, but rather ,a combination of LBJ's coasting, Durant's incredible stretch without Westbrook, among other things, leading to Durant's massive edge in NPI +/- and box-score stats?

Oh my goodness.

Maybe if it was a different scenario where you had no clear runaway top guy on the planet with many guys duking it out for that title, than Durant would make his case. But when you have a guy who has been hands down best on the planet for almost 6 years now with 4 MVPs and multiple Championships on his shoulders, no chance in hell some meaningless regular season games can change any of that..thats a damn disgrace.. big players show up in big games, so don't be sayin this stuff about early regular season games meaning anything when a guy has reigned at the top for a long time and is proven in every situation that actually matters.

[MaliBrah]

Has Durant supplanted LeBron?

I'll let you know in June.

Durant's going to have to destroy lebron in June to surpass him and frankly that isn't happening.

[kayess]

FTR Doc, does this mean you think it's not that Durant's decisively supplanted LeBron as the best player on the planet, but rather ,a combination of LBJ's coasting, Durant's incredible stretch without Westbrook, among other things, leading to Durant's massive edge in NPI +/- and box-score stats?

Oh my goodness.

Maybe if it was a different scenario where you had no clear runaway top guy on the planet with many guys duking it out for that title, than Durant would make his case. But when you have a guy who has been hands down best on the planet for almost 6 years now with 4 MVPs and multiple Championships on his shoulders, no chance in hell some meaningless regular season games can change any of that..thats a damn disgrace.. big players show up in big games, so don't be sayin this stuff about early regular season games meaning anything when a guy has reigned at the top for a long time and is proven in every situation that actually matters.

Don't forget, when LeBron came into his own in 2009, he hadn't won anything yet as well.

I get the opposition here, but you don't have to have won anything previously in order to become the best.