Who is the most valuable NBA player?

[EvanZ]

i'm certainly not going to propose the idea that kevin love is an all-D 1st team defender. i don't believe he is, and he gets few steals and blocks for a PF. but this webpage:

http://www.82games.com/1112/11MIN12.HTM#bypos

shows poor PF/C counterpart production when he's played this season, and current Synergy defensive data shows a low PPP and low eFG% allowed by him...

As for Love, the Wolves don't have too much trouble scoring when he's not playing , eFG even improves when he's not playing (47% -> 51%). He obviously improves their rebounding, but they also have a better opponent eFG when he's not playing (46% -> 44%), opponents are getting to the line less and Minnesota creates a whole bunch of more turnovers (13/p48m -> 20/p48m).

You may want to look at my recent post on adjusted four factors.

http://thecity2.com/2012/02/21/new-play ... ctor-a4pm/

The line for love:

Code: Select all

PID	NAME	TEAM	POS	POSS	A4PM	OEFG	OTOR	ORR	OFTR	DEFG	DTOR	ORR(OPP)	DFTR
778 Love	MIN	4.2	10793	-0.03	-0.11	0.56	2.11	1.10	-0.02	-0.50	-0.95	-0.15

According to this, he has little positive or negative effect on eFG on offense or defense. His contributions in terms of turnovers are negative on both sides. He increases the rate of FTA/FGA by 1 per 100 on offense. His biggest contributions by far appear to be in terms of rebounding, where he increases offensive rebounding % by 2 points and lowers offensive rebounding of the opponent by about 1 %-point. (IOW, increases his own team's defensive rebounding rate by 1%.) Just thought I'd throw these out there. Jerry has done similar analysis of three of these factors previously, and I don't think his results differ much.

[Mike G]

When Love is on the floor, about 70% of the other players on the floor are starters.
When he's off the floor, only about 40% of those on the floor are starters.
Do these matter, when you're looking at team and opponent rebound% ?

[EvanZ]

In theory, it doesn't matter. Or matters as much as it would matter for adjusted +/-, anyway. This is what we are "adjusting" for, after all.

[bchaikin]

According to this, he has little positive or negative effect on eFG on offense or defense...

according to Synergy, his defensive PPP allowed is quite low, and his defensive eFG% allowed low. how should one rectify this conundrum, your calculation vs. Synergy charted data?...

this not to mention the 82games.com counterpart data for opposing PF/Cs that shows a low eFG% and a low PER allowed when love was on the floor...

[EvanZ]

Why would you have more faith in a counterpart metric like PER than something that is adjusted for opponent and teammate quality? That seems like one or two steps backwards in terms of analytics to me.

[schtevie]

Not two steps backwards. It's just a jump to the left, then a step to the right...http://video.google.com/videoplay?docid ... 8088648679

[J.E.]

His rating is even higher now at +2.8, but I'd give more weight to the fact that it's at 2.8 instead of it going upwards. Love's rating went up as Rubio's and Milicic's On/(Off) numbers went down. With a lower lambda his rating could definitely be even higher, but this value of lamdbda worked best in the last 10 years

Also remember that we're penalizing for the quadratic difference, so his offensive rating won't go from +2.8 to +3.8 as easily as it went from +0.8 (prior from last year) to +2.8

Obviously, his true rating can be higher than what RAPM says, but I still really doubt that he's having as much impact on a game as the league's best players

[J.E.]

His pick and roll defense is still pretty horrible for someone that might get MVP votes and he's also bad at contesting shots...

love's played 1/7 to 1/6 of the t-wolves' total minutes played, yet the team is better than the league average in defense (pts/poss allowed). if his overall defense was poor, do you realize just how good of defense his teammates playing the other 5/6 to 6/7 of time would have to be for the team's defense as a whole to be better than league average? other than rubio, who else on that team are good defenders such that with love's "horrible" p-n-r defense and "bad" shot contesting the team as a whole is better defensively than league average?...

I didn't say his overall defense was "poor"; just not good enough, in combination with his pretty good offense, to make him an MVP candidate
Why wouldn't it be possible that (a) his P&R defense is not good, (b) he's not good at contesting shots, (c) he's an average defender despite (a) and (b), (d) the Wolves are above average on defense, all at the same time?

Tolliver and especially Rubio are good defenders

define "best" - top 10 players in the league? top 5 among PFs?

I don't know.. "good enough to be considered for MVP"?

In line with what Evan reported, most negative impact comes from not creating turnovers on defense and not helping to avoid turnovers on offense. His defensive rebounding impact (according to adjusted rebounding) appears to be quite a bit smaller than what his individual numbers suggest, too

[bchaikin]

Why would you have more faith in a counterpart metric like PER than something that is adjusted for opponent and teammate quality?

where did i say i "...have more faith..." in a counterpart metric like PER? i asked a simple question - how should one rectify the fact that your calculation, according to you, shows kevin love...

...has little positive or negative effect on eFG on offense or defense...

when two other metrics - both counterpart PER and charted Synergy defensive data - show, or rather infer, his man defense has been better than average to good...

That seems like one or two steps backwards in terms of analytics to me.

it was a simple question...

His pick and roll defense is still pretty horrible for someone that might get MVP votes and he's also bad at contesting shots...

I didn't say his overall defense was "poor"; just not good enough, in combination with his pretty good offense, to make him an MVP candidate. Why wouldn't it be possible that (a) his P&R defense is not good, (b) he's not good at contesting shots, (c) he's an average defender despite (a) and (b), (d) the Wolves are above average on defense, all at the same time?

so now his p-n-r defense has improved from horrible to just not good, and his shot contesting has improved from bad to just not good?? which is it? and if a player is pretty horrible or just not good at p-n-r defense, and bad or just not good at contesting shots, and his rates for steals and blocked shots are less than that of just the league average player at his position, just what is he doing positive on defense to counteract all that such that overall he is an average defender?...

Tolliver and especially Rubio are good defenders...

no argument from me about ricky rubio. he forces turnovers via steals and offensive fouls drawn at one of the highest rates in the league. he's on pace for over 200 ST/3000min and over 60 offensive fouls drawn per 3000 minutes...

what evidence do you have that anthony tolliver is a good defender, or at least a better defender than kevin love? i ask as the Synergy defensive data shows a much lower PPP allowed for love vs. tolliver, and an eFG% allowed that is almost 2% lower...

unless you can show that tolliver has routinely guarded much better offensive players than love has, which might be difficult considering love has played close to 3 times as many minutes as tolliver has, how have you determined he's the better defender?...

In line with what Evan reported, most negative impact comes from not creating turnovers on defense and not helping to avoid turnovers on offense.

tolliver has a higher/worse rate of turnovers per touch than does love, and both force turnovers via steals and offensive fouls drawn at similar rates. so does tolliver have a negative impact too?...

just out of curiosity, this webpage:

http://www.82games.com/1112/11MIN12.HTM#onoff

shows that the t-wolves on defense allowed a 48.3% eFG% with love on the floor (1068 minutes), but just a 44.8% eFG% with him on the bench (323 minutes)...

this webpage:

http://www.82games.com/1112/11MIN8.HTM#onoff

shows the t-wolves on defense allowed a 44.1% eFG% with tolliver on the floor (426 minutes), but a 49.0% eFG% with him on the bench...

so love has played over 3 times as many minutes compared to how often he sat, and tolliver has played just 1/2 the time compared to how much he sat. how does this affect your calculations that determines how good each is as a defender?

in other words, had the t-wolves allowed a 54.8% eFG%, and not a 44.8% eFG%, in the 323 minutes kevin love sat, would that affect the defensive rating you give him? would he then be considered a better defender based on something that occurred when he did not play?...

[J.E.]

His pick and roll defense is still pretty horrible for someone that might get MVP votes and he's also bad at contesting shots...

I didn't say his overall defense was "poor"; just not good enough, in combination with his pretty good offense, to make him an MVP candidate. Why wouldn't it be possible that (a) his P&R defense is not good, (b) he's not good at contesting shots, (c) he's an average defender despite (a) and (b), (d) the Wolves are above average on defense, all at the same time?

so now his p-n-r defense has improved from horrible to just not good, and his shot contesting has improved from bad to just not good?? which is it?

Why are you splitting hairs? "horrible" is a subset of "not good"

[..]
and if a player is pretty horrible or just not good at p-n-r defense, and bad or just not good at contesting shots, and his rates for steals and blocked shots are less than that of just the league average player at his position, just what is he doing positive on defense to counteract all that such that overall he is an average defender?...

Defense has way more aspects than those that were mentioned: Rebounding, smart fouling, taking charges and many other things that I would put under the blanket of "general help defense". I'd imagine a significant part of Kevin Garnett's contribution comes from that last part, with him roaming around a lot and doing it so well that it sometimes feels like there are 6 defenders on the court.

Rebounding is probably the biggest positive contributor to Love's defense

[mystic]

bchaikin, your analysis has the focus only on 1on1 defense, but a huge part is actually the contribution to the team defense (e.g. help, rotation). The overall impact by a defender can't be just seen by the counterpart offense. Love stays with his man, that's is something he didn't do last season for example to that extend, but he is still not good at help defense, still way too concerned with offensive rebounding instead of going back defensively and rather stays underneath the basket instead of closing out on the perimeter when a teammate lost his man.

in other words, had the t-wolves allowed a 54.8% eFG%, and not a 44.8% eFG%, in the 323 minutes kevin love sat, would that affect the defensive rating you give him? would he then be considered a better defender based on something that occurred when he did not play?

That can happen, but the reason for this is that the algorithm would think his teammates would be worse defenders; which is likely to be true, if the eFG% would be 54.8%, don't you think? So, it is a bit silly to ask such a question, imho.

[Mike G]

I understand that there can be player movement, but moving from 0.3 to 2.5 in 30 games or so seems dramatic. ...

Also, if 2.5 is his current rating, doesn't that mean he's been much better than that actually to pull up his prior 3 seasons? He's played about 17% of his career minutes this season. His rating must be tremendous.

Isn't it possible that Love's seeming improvement is largely movement within the range of his standard error? Don't these APM type summaries include a large potential error, that more data (and varying lineups) tend to affect?

So maybe at the beginning of the season, his RAPM was very near the bottom end of his possible value; and it's just shifting to a more normal part of the probability distribution.

What percentage of players will actually be outside the bounds of their SE?
Jameer Nelson has a 1-year APM of +29.8 and a SE of 12.8.
His backup, Chris Duhon, is +33.6, SE 12.2
What do we make of these ratings and their error estimates?
What kind of SE accompany RAPM?

[xkonk]

What kind of SE accompany RAPM?

It's been discussed elsewhere on the site, but the SE for RAPM are guaranteed to be smaller than the SE for APM. However, they are somewhat uninterpretable because of the bias introduced by regularization.

[J.E.]

I went ahead and watched UTA@MIN from two days ago. I didn't find as much mistakes as I found in games earlier this season; maybe he improved in the aspects I looked for, but sample size (1) is obviouslya factor. The team I watched MIN play against earlier were also a bit stronger (HOU, CHI)

Here are some obvious mistakes


His man is free under the basket. Not entirely his fault as Pekovic (or someone else) could have realized that Love isn't covering Millsap and switched, but Love probably caused some confusion by simply not covering the man he covered the entire quarter before that basket.


Harris drives at him and Love tries to take a charge and fails, giving up a free layup. Something similar happened in the CHI game where he tried to take a charge and didn't raise his arms on an opponent shot in the paint (that went in)


Loses track of his man in the zone, giving up a free dunk


Not defense, but he commits one of the worst turnovers in basketball which leads to a 2on1 opponent fastbreak. I'd guess that this kind of situation leads to a layup in 80%+ of the cases. It's the first time I've seen him commit such a bad turnover though, so it's probably not a reoccuring thing.


Kind of late at running back, his man gets by him with two easy dribbles and scores on an easy layup


Harris drives down the lane. Love is right there but does *nothing*.


Very easily gives up ground to an opponent who's in the paint and has already dribbled


As already seen in the other two games, he doesn't seem very motivated to run back in transition. This often leads to easy opponent baskets, not necessarily by the man he is supposed to be guarding. He's not that bad at contesting shots of his own man, but when other players drive at him he kind of freezes.
My opinion is still that, although he has good impact on offense, his overall impact is not good enough to make him an MVP candidate. He'd have to have even more impact on offense, or improve his defense in regard to P&R defense and/or contesting shots