And now for something completely different.
What? You were expecting Eric Idle?
I wasn’t sure what to post today and so I decided to take a topic and give a different spin. So I’m going to talk MVP (as promised) but it’s going to be different because rather than looking at the whole season, I’m going to go off and look for a series of good nights.
Let me explain, I felt a little odd yesterday when through the magic of the podcast I discovered that my rankings (done with great care with a perfect blend of maths, excel, a heavy dose of the Basics and powered by NerdNumbers) looked very much like the numbers coughed up by the supercomputers up in Bristol,Conneticut. While it feels oddly cool to hack yet another model (albeit accidently), I felt a little unoriginal.
A little unoriginal can be a good thing though
Today we fix that.
The graph up top is the wins generated for every game played by every player who has played more than 10 games this season. Isn’t that cool? Ok, here’s the rundown for those keeping count at home:
9531 Games in the data set
.059 Wins generated is the average value per player
.165 is the standard deviation
So by those numbers I’d categorize games (and count them) as follows :
| Category | Wins Produced for Game | Count | % |
| Lousy | <=-.271 | 958 | 10% |
| Bad | <=-.106 | 406 | 4% |
| Average | >-.106 and <.224 | 6713 | 70% |
| Good | >=.224 | 1098 | 12% |
| Great | >=.390 | 356 | 4% |
By these numbers only 16% of all the games played were good to great and only 4% were great.
And this is all well and good but what does this mean for my MVP evaluation? Here’s what I’m going to do, I’m going to assign each player 2 points for a great game, 1 for a good game and dock him a point for a bad night or 2 for a lousy night (and yes Dre you did kick my brain down this path). MVP will be the player with the Most Points.
By this logic, Chris Paul and Kevin Love are in a dead heat for MVP. Love has the edge in great games (but he also has a lousy one). One interesting thing to look at is the 17 players without bad or lousy games:
| Player | Lousy Games | Bad Games | Average Games | Good Games | Great Games | Games | Points |
| LeBron James | 18 | 11 | 9 | 38 | 29 | ||
| Pau Gasol | 18 | 15 | 4 | 37 | 23 | ||
| Landry Fields | 19 | 10 | 6 | 35 | 22 | ||
| Marcus Camby | 17 | 12 | 5 | 34 | 22 | ||
| Kevin Garnett | 15 | 7 | 7 | 29 | 21 | ||
| Jason Kidd | 24 | 6 | 6 | 36 | 18 | ||
| Rajon Rondo | 12 | 9 | 4 | 25 | 17 | ||
| Kris Humphries | 20 | 12 | 2 | 34 | 16 | ||
| Joakim Noah | 13 | 8 | 3 | 24 | 14 | ||
| Nene Hilario | 21 | 8 | 2 | 31 | 12 | ||
| Mike Dunleavy | 24 | 7 | 2 | 33 | 11 | ||
| Thabo Sefolosha | 32 | 6 | 38 | 6 | |||
| Ben Wallace | 30 | 4 | 1 | 35 | 6 | ||
| Josh Childress | 23 | 1 | 24 | 1 | |||
| Vladimir Radmanovic | 23 | 1 | 24 | 1 | |||
| Jamario Moon | 20 | 1 | 21 | 1 | |||
| Antonio McDyess | 26 | 26 | 0 |
Here Lebron is clearly the King. If I go the other way (i.e. players with no good or great nights):
| Player | Lousy Games | Bad Games | Average Games | Good Games | Great Games | Games | Points |
| Antonio McDyess | 26 | 26 | 0 | ||||
| Eric Maynor | 2 | 33 | 35 | -4 | |||
| Goran Dragic | 2 | 29 | 31 | -4 | |||
| Greivis Vasquez | 2 | 20 | 22 | -4 | |||
| Baron Davis | 1 | 3 | 17 | 21 | -5 | ||
| Shelden Williams | 3 | 1 | 26 | 30 | -7 | ||
| Jordan Hill | 3 | 2 | 23 | 28 | -8 | ||
| Ryan Hollins | 4 | 19 | 23 | -8 | |||
| Xavier Henry | 4 | 1 | 20 | 25 | -9 | ||
| Will Bynum | 4 | 1 | 19 | 24 | -9 | ||
| Chris Duhon | 4 | 1 | 18 | 23 | -9 | ||
| Ronnie Price | 4 | 1 | 17 | 22 | -9 | ||
| Josh Powell | 5 | 19 | 24 | -10 | |||
| Sebastian Telfair | 5 | 1 | 18 | 24 | -11 | ||
| Jason Maxiell | 5 | 1 | 17 | 23 | -11 | ||
| Solomon Jones | 7 | 19 | 26 | -14 | |||
| Jose Barea | 7 | 1 | 28 | 36 | -15 | ||
| Steve Blake | 7 | 1 | 28 | 36 | -15 | ||
| Dante Cunningham | 6 | 3 | 24 | 33 | -15 | ||
| Willie Green | 8 | 1 | 22 | 31 | -17 |
| Category | Wins Produced for Game |
| Lousy | <=-.271 |
| Bad | <=-.106 |
| Average | >-.106 and <.224 |
| Good | >=.224 |
| Great | >=.390 |
Antonio McDyess shows up here too (oddly enough). And If I look at a topical player?
| Player | Lousy Games | Bad Games | Average Games | Good Games | Great Games | Games | Points |
| Carmelo Anthony | 3 | 1 | 15 | 7 | 1 | 27 | 2 |
Let’s maybe save that for later.
So I’m hoping this was somewhat original.
No re-enactments here (Image courtesy of xkcd.com)
The insanity continues in Part 2.
Arturo's Silly Little Stats v2.0 
nerdnumbers
01/12/2011
Awesome Arturo! Ok so question though as a GM do you prefer Paul or Love? I know my answer, but let’s see what you think.
EntityAbyss
01/12/2011
I prefer Paul, because he’s my favorite player. :-). Also, because he’s the man. Once those kneepads come off, it’s over. That’ll probably be next season though. Also, he’s just the man. Chris Paul for MVP.
some dude
01/12/2011
This is a silly way to pick an MVP. Just be a statwhore on a crappy team and voila. Don’t even need to try on defense. Kevin Love is 2nd in the rankings on a team that is 9-30. That should be reason enough to have started over.
You’re not that valuable if you’re not winning games.
Chris Paul is one of my favorite players, but he hasn’t played that well the past 10-15 games or so (for his standards, mind you). He looks…slow. I’d be very afraid he’s becoming Brandon Roy. I don’t see how he’s in the MVP conversation with the way the hornets have been playing the past month.
Dirk is below Elton Brand…
Seriously, this is a really bad way to rank MVP candidates.
SM
01/12/2011
SD – What do you think team wins should be worth for each player as applied to this ranking system. Would that make this a good way to rank MVP candidates, or what would be a good way in your view to rank mvp candidates?
Also your complaints about this way to do ‘pick’ and mvp seem very petty and not even all that accurate. Brand has played a bunch more games than Dirk and is barely ahead of him. Even is Chris Paul is slow for the last few weeks (how did you measure this?) isn’t at least possible he was extraordinary for the other 70% of the games in which he has played?
some dude
01/13/2011
Team wins have to play a part, no doubt. How, based on the scale above, I dunno.
As for Dirk below Brand, uh, isn’t it obvious how undervalued Dirk is to this team, now? The teamm cannot score without him. It’s averaging 91 points per game the past 8 without him. Dirk is what makes the engine run, so to speak. Dallas becomes a very easy team to defend without him. WP is not capturing this, unfortunately.
Would any of the top teams with high ranking MVP guys (Lebron, Wade, Gasol, Howard, Horford, Fields, CP3) likely go 2-8 against the same teams? Only with CP3 could I imagine a yes. I didn’t put Love in there because they’re likely to go 2-8 with him anyway.
As for Cp3, he was great for the first 3 weeks, but just good since. New Orleans record since the 8-0 start isn’t very good and unless things turn around, I’d have to believe it was an fluke. He will always be in the conversation because of his amazing numbers and the fact that his team probably falls apart without him, but it’s been fairly established that you need to be on a contender to have MVP merit (as proven by Nash’s MVPs), and they aren’t one right now. And as I said, CP3 has seemed off for about 3 weeks now.
arturogalletti
01/12/2011
SD,
My soapbox dude 🙂
I’m think about how to add in a sliding scale for team wins.
some dude
01/12/2011
The no “lousy or bad” games list in interesting, too. Pretty much all of them are good rebounders for their position. In other words, in games where they mostly play like crap, they rebound enough to keep them from being “bad” according to this.
http://www.basketball-reference.com/boxscores/201012300ORL.html
How was this not a bad game for Landry, exactly?
arturogalletti
01/12/2011
SD,
-.05 WP48 that night for Fields. Makes him average (i.e. within one std deviation of the mean). He had a a bad night for him (in fact that was his worst for the year so far) but lets look at the boxscore:
0-3 FG, 1 ORB,3 DRB,1 AST, 0 Blks,0 Steals, 0 Turnovers, 2 Fouls in 28 minutes. He didn’t help you but he didn’t kill you with bad stats.
And Jason Richardson (playing opposite him) put up a -.02 WP48 performance in 35 minutes. So I think it’s very fair to call it a wash for the Knicks.
some dude
01/13/2011
So basically you reward a person or standing around and doing nothing for 28 minutes.
So Joel Anthony’s 0-0-0-0 game was “average?” http://www.basketball-reference.com/boxscores/201101090POR.html
The logical conclusion of your claim is that if I was playing for an NBA team, or rather a traffic cone, it would provide an average effort because it didn’t “hurt the team” in a measurable box score way. That is an absurd way to define the “average performance.”
5 Joel Anthony’s would make an average team!
Howabout we put in a caveat that if you aren’t actually doing anything while on the floor, you’re hurting your team. If your team is playing 4 on 5 offense because you’re not shooting or finding open guys, you’re hurting the team. WP needs to capture this, otherwise it concludes that 5 traffic cones would give an average performance.
Also, what if I told you he didn’t defend J-Rich so much as turkoglu and JJ Redick?
arturogalletti
01/13/2011
Joel Anthony against Portland on 01/09/11,-.250 wins, bad game.
You’re Wrong,Wrong, Wrong.
You have to perform at close to an average level to be average. Almost putting up a trillion is not average (particularly for a center).
0 Raw Productivity or ADJP48 at the 5 equals about -.300 WP48 or an inverted superstar.
some dude
01/13/2011
Then it’s an issue of position adjustment, no?
What if Joel Anthony’s line was for a SG? What would be his WP?
I mean, just because Landry Fields had a whopping 4 rebounds as a SG didn’t make his night average.
Fred Bush
01/12/2011
Noah’s 0-rebound game wasn’t “bad”?
Also, any idea why the distribution is weird?
arturogalletti
01/12/2011
Fred,
Player Joakim Noah
Team CHI
Opp ORL
AdjP48 0.24
WP48 -0.06
Wins -0.03
PerMinUsage 25.3
Usage 13.4
MP 25.43
Pg 0
Sg 0
Sf 0
PF 0
C 25.43
P2 6
P2A 11
P3 0
P3A 0
FT 4
FTA 4
ORB 0
DRB 0
AST 4
TOV 0
STL 0
BLK 1
PF 4
Points 16
Average.
The second question (which has inspired today’s post) seems to have to do with opportunity. Bad teams are forced to play bad players regardless of performance. So a lot of the lousy games have to do with talent and depth.
Fred Bush
01/12/2011
I’m not sure about talent/depth being the only explanation. For example, “good” players, arbitrarily defined as scores >=2, have 77 lousy games, 96 bad games, 570 good, 258 great. That’s still more lousy games than expected.
arturogalletti
01/12/2011
Fred,
I think part of it has to do with opportunity and talent and part with players being “stars”. A “star” player will still get his reps even if he’s having a lousy game. It’s interesting actually that as you look at player variability it implies that some teams have increased variability built into their rosters.
Fred Bush
01/12/2011
Also, what is up with Grant Hill? Incredibly erratic!
Man of Steele
01/12/2011
First off, mad props for the Alex Ross Captain Marvel painting!
Second, the most striking player to me in all this is Thabo Sefolosha. He’s never all that good or all that bad, he’s just average every time out. I wonder what percentage of the player population is average in over 80% of their games?
EvanZ
01/12/2011
Vladimir Radmanovic doesn’t have any bad games, let alone lousy ones?
laugh test: failed!
arturogalletti
01/12/2011
Evan,
You’re being impolite and rude but i’ll answer your question (because I was curious myself):
Date Minute Played Wins Produced
11/10/2010 6.1 -0.115
1/1/2011 21.6 -0.099
10/27/2010 6.4 -0.095
1/9/2011 21.1 -0.090
11/24/2010 14.9 -0.078
10/29/2010 19.8 -0.077
11/5/2010 5.3 -0.071
11/13/2010 25.9 -0.064
10/31/2010 8.6 -0.064
12/31/2010 32.3 -0.064
11/11/2010 23.4 -0.060
12/25/2010 17.2 -0.034
11/19/2010 17.4 -0.027
11/8/2010 6.3 -0.014
12/29/2010 22.1 -0.006
11/3/2010 15.9 -0.004
11/21/2010 18.5 -0.002
11/30/2010 11.9 -0.001
11/7/2010 19.9 0.013
11/15/2010 22.6 0.015
12/10/2010 10.7 0.017
12/14/2010 6.8 0.040
1/3/2011 23.0 0.044
12/13/2010 17.3 0.048
12/7/2010 5.2 0.050
11/26/2010 9.6 0.100
12/18/2010 23.3 0.100
12/20/2010 16.4 0.104
12/8/2010 16.2 0.105
1/7/2011 21.6 0.139
1/5/2011 16.6 0.151
12/21/2010 27.8 0.198
12/27/2010 25.6 0.256
His only game that qualified as bad didn’t meet the 10 Minute requirement I used. For the most part he hurts and not helps your team. Lousy implies bad play and ample opportunity.
EvanZ
01/12/2011
I’m being impolite to you or Vlad Rad? Seriously, I know you don’t watch Warriors games (not as much as I do), so I wouldn’t expect you to see what I see.
Knowing what I know about WP, how it doesn’t handle defense well (by your own findings), and what I know about Vlad Rad, I can say honestly and with full confidence, that he’s had more lousy games than average or good ones.
“Lousy implies bad play and ample opportunity.”
Ask any Warriors fan if Vlad meets the above requirements, and you will get a resounding “Yes. Sigh.”
arturogalletti
01/12/2011
I’ll give you that his numbers once we account for individual D will look worse. There are worse options out there. I give you Glen “Big Baby”Davis. Seriously. You don’t even have to give me a pick.
some dude
01/13/2011
i’d so much rather have Big Baby than the Space Cadet.
Ra's Head
01/13/2011
I feel that a lot of the disagreements in the comments here really center around the classification of ‘bad’ and ‘lousy.’ Since your ‘average’ is so large (presumably a standard deviation determined thing), many games that general people would classify as bad are in the ‘average’ range. The difference between -0.106 and 0.224 seems pretty big. The Landry Fields game referenced above, many people would say was a bad game by him. Really, this method isn’t making such subjective judgements, merely stating how many standard deviations away from the mean the performance was. Many average games are significantly better or worse than other average games, so we have to take that into account.
To avoid this, you could have called them “category 1,” instead of ‘lousy’ and so on. That just isn’t as fun, but we probably shouldn’t be so attached to the specific words. As a side note, I love your site.
arturogalletti
01/13/2011
Ra’s Head,
Could not have said it better myself actually. I may divide the average group into help(>0) and hurt (<0) in the future.
Oh, and thanks for the kind words.
some dude
01/13/2011
this would make a lot more sense.