In a hurry today, so this’ll be quick (but I hope great fun).
As reader of this blog, you might now that I’ve been playing around with Win Produced game splits and adding in defense. The driver for this has been the fact that currently Wins Produced divides defense up at the team level for all stats that are not in the boxscore (see the Basics for detail). So I have been trying for a while to get at this. And it’s been hard but we’re making it work.
Science! (Image courtesy of xkcd.com)
This is revision #2 (powered by NerdNumbers) which is as good as it’ll get until we get play by play data (which Andres is working call it super NerdNumbers powers) . What I’ve done is look at Player data for the last ten games for each team and map how each opponent did by position. I’ve also accounted for the effect of altitude,rest and home and away. That looks like so:
How do you read this? Let’s give two examples. Devin Harris and Jordan Farmar get worked by opposing Point Guards as raw productivity (ADJP48) for their opponents goes up a straggering 63%. By contrast, Kobe Bryant is an excellent defender as his opponents raw productivity (ADJP48) drops 28% (I did just throw up a little in my mouth thanks for asking).
Fear the defensive power of the duckface!
So what happens when I work out WP48 for the last 10 games adjusting for position based defense? And then use my nifty points equations (last seen here) to work out point margin contributions? And then work this out for all player in the league and rank them by value? And when I stop with the questions and just post the damn table :-)? This:
Note: I was off by a factor of two initially on some of the point stuff (Long explanations here, short waht is I used the sum not the average fuction in excel). Fixed now (As always your refund is in the mail). Player Value last ten games (for all players >100 MP as of 12/21/10 Take 2)
The big winner?
Arturo's Silly Little Stats v2.0 
Leon
12/21/2010
Steve Nash a bad defender huh?
arturogalletti
12/21/2010
It was very surprising.
some dude
12/22/2010
again a situation like ron. Nash doesn’t always guard the opposing PG. They try to hide him on the weakest perimeter player. The often put Hill on the PG. That’s why you see everyone else so high on the team.
nerdnumbers
12/21/2010
So interesting note. It’s possible to be good with subpar defense but impossible to be bad with good defense. This is awesome stuff Arturo. Kevin Garnett is pretty amazing at holding down the fort. It seems wrong that Miami and Boston have been playing so well and neither has really fielded their full strength at any point in the season.
Baytown Ryan
12/21/2010
Nice Work Arturo! Still seems like the sample size might be a little small to get an accurate reading. Anecdotal case in point, Kevin Martin is pretty much universally renowned for his horrendous defense, but as your numbers suggest the Rockets as a team have stepped up their collective game as of late and I believe it may be skewing his individual defensive numbers. Just a first glance thought. Still awesome job. Keep up the good work.
arturogalletti
12/21/2010
I did 10 games as a first pass. I’ll do bigger data sets in the future.
T Pain
01/04/2011
82 games has Battier playing about 3/4 of his minutes at SG. I’d guess that by this evaluation Martin rides his defensive coattails. Finally a table/metric to justify my mancrush on Battier, kudos.
Mike
12/21/2010
It’s surprising that Opposing SGs are a staggering 67% worse against the Houston Rockets. Is it a function of help defense? Or is Kevin Martin really a defense beast?
arturogalletti
12/21/2010
This really might be a function of Battier taking the better man.
Fred Bush
12/21/2010
The guard tables have the highest variance by far. I would suspect it’s because guard efficiency depends a lot on how well they hit 3s and jump shots, which will vary a lot from game to game.
Fred Bush
12/21/2010
Small forward also has a fair amount of variability (not quite as much as the guards), but look at how smooth the big man results are compared to the others.
Guy
12/21/2010
This is interesting data, Arturo. But aren’t you double-counting quite a bit here? A rebound taken from the opposing player will be counted twice. Block shots will show up in reduced opponent shooting efficiency. Steals/turnovers and FT/fouls could pose a problem too. Or have you already adjusted for all of that? Mainly, you just want to capture opponent eFG%.
arturogalletti
12/21/2010
Guy,
I’m adjusting the position averages for the opposition to actual position values. So Point Margin generated = Actual Point Margin Generated. My one adjustment is to take out the effects of Homecourt,altitude and rest (this is significant but the numbers track). So this is really neutral site, defense adjusted WP48.
guy
12/22/2010
I’m finding it a little hard to follow the process, but it seems like you credit players if opposing position has below-avg adjp48. But that creates a huge double-counting problem. If Love grabs a reb, his opponent gets one less. If he blocks a shot, opponent efficiency declines. Steal becomes turnover. So you need to remove all of those elements. Might be easier to just use opp eFG.
But this data is a great way to see (sorry) diminishing returns at work. It appears Love’s opposing position players have average productivity. If so, that means his WP48 is extremely inaccurate. If WP48 is capturing Love’s contribution accurately, his opponent adjp48 should be way below average.
Guy
12/22/2010
OK, I read your prior post, and I think I follow what you’re doing here. Opponent AdjP48 obviously can’t be treated as an additional defensive measure, because so much of it is already accounted for in WP48 (PF=opponent FT, rebound gained = opponent reb denied, STL = oppTOV, etc.). However, I think your data is extremely valuable as a way of validating WP48 (or other metrics). Each player’s OppAdjP48 should be a kind of mirror image of his own WP48. Not exactly, of course, because the player’s shooting efficiency shouldn’t have much impact on OppAdjP48, but there should be a high correlation. A player contributing a lot of steals should have opposing players with high TOV rate. And most importantly, high REb48 players should have low-REb48 opponents, if they are really taking rebounds from opponents rather than teammates.
But look at your data: the high rebound guys don’t have the OppAdjP48 numbers that they should. Love appears to have no impact on opponents at all. Camby’s opposing centers are actually above average, even though Camby is allegedly contributing 6 rebounds above average per 48 minutes. (And Camby is also above average on BLK, TOV, PF, and STL — so his opposing players should have a far-below-average OppAdjP48.) And even in the case of players who do have a positive “defensive value,” that point differential tends to be far lower than it should be if WP48 were correctly measuring their value.
I think this is a very important discovery, Arturo, if perhaps not the one you were expecting. If you can increase your sample size, a logical next step would be trying to find patterns in this data: which players really are having the impact on opponents that WP expects, and which are not? Great work…..
Fred Bush
12/22/2010
Guy: Guards and SF for MIN are +24%, PF/C is -3%. Looks like a pretty big difference between Love and his teammates.
Guy
12/22/2010
Fred: I don’t think I follow. Are you saying this means Love is better than MIN’s guards and SF? Perhaps, but why is that important?
More importantly, shouldn’t we be able to see an impact on opposing players by the league-leader in WP48, a guy whose value derives a lot from his rebounding? Shouldn’t his opponents’ rebounding numbers, and thus their AdjP48, be suppressed by Love’s rebounding skill? How can such a great player leave hardly a ripple?
Fred Bush
12/22/2010
If you model team defense, MIN team defense is terrible. If you model individual defense, you see that Love is apparently significantly better than his teammates.
There’s a .48 correlation between SG/PG average defense and PF/C average defense for a team. That’s pretty high, so that means that terrible team defense is going to affect the individual stats a lot, and therefore you can’t say much about Love because his team is so bad on defense. Your case is much stronger with Camby in Portland.
Man of Steele
12/22/2010
Since Landry Fields has been under discussion of late, it’s interesting to note that the Knicks hold opposing SGs to 63% of their normal production
Guy
12/22/2010
Fields is another good example. He’s supposed to be adding 5.5 rebounds every 48 minutes (above an average SG), which means a +5.5 point differential. But NYK opposing SG are only -2 points overall. I suppose he could be taking rebounds from other positions as well. But other opposing players are above average on AdjP48 as a group. And of course the Knicks are a below-average rebounding team overall (on both oreb and dreb), despite Fields’ apparent brilliance. I guess we have to add him to the long list of “great rebounders” saddled with “poor-rebounding” teammates.
arturogalletti
12/22/2010
To be fair he is playing with Amare
Guy
12/22/2010
Amare has a Reb48 of 11.9, just 0.5 below average, so I don’t see how that explains much. But obviously, you can point to teammates who are below-average rebounders, since Fields is very high and the team overall is below average. And that’s precisely my point: it turns out that every great rebounder has teammates who are below-average rebounders.
arturogalletti
12/22/2010
Perhaps Fields is stealing the opposing teams rebounds?
And judging by the numbers Amare, Turiaf and Mozgov are just giving them right back.
Guy
12/22/2010
Arturo: Although we’ve talked about this a lot, I fear we are somehow still talking past each other. Because it seems to me you are just confirming diminishing returns by pointing out these players. Every great rebounder — every single one, without exception — is surrounded by guys who “give them right back” (according to WP). The way to disprove diminishing returns is exactly the opposite of what you’ve done here: find a great rebounder whose teammates do NOT give them right back. So far, no one here has found even a single example.
And in any case, what do you think about these shocking numbers you’ve produced on Camby and Love? Or Duncan? How can these players be securing such a huge number of rebounds for their teams, while having no impact at all on opponents in the same position?
some dude
12/22/2010
in fairness to guys like Duncan and even Camby, help defense plays a big role in the production of the guards/wings too and isn’t attributed to them.
What is more interesting to note is not a single center holds an opponent below 80% of their production. Not 1! And for PFs, we see this happen only once with Boston & KG.
But for the guards/wings, we have 18 such instances (if i counted right). This distribution tells me that bigs are producing more consistently across all levels. It also tells me that holding a big to 90% of his contribution is more impactful than holding a guard to 85% of his contribution.
And if bigs contributions is mostly made up of rebounds, then this is great evidence that rebounding is overstated at the individual level OR that the position adjustment is not correct.
Unless we are to assume NBA bigs are of much higher quality,we should be seeing similar drop rates across the league, but we don’t. Bigs are being overrated in WP48.
Guy
12/22/2010
sd: I don’t disagree that big men may impact the shooting efficiency of opposing players who play different positions. But Arturo is not looking only at opponent efficiency, but at opponent AdjP48, which includes rebounds, steals, blocks, personal fouls, and turnovers. Surely we should see a substantial impact of those in the performance of opposing players in the same position, right? Not one-to-one, but a strong connection. And rebounds is the big factor here: it is simply impossible that Love can be adding 9.5 rebounds to his team every 48 minutes, while opposing PFs put up average productivity.
some dude
12/23/2010
yeah, i agree. that’s what i was getting at in the 2nd half of the post.
some dude
12/22/2010
Arturo – Ron Artest generally guards the other team’s shooting guard and Kobe guards the SF. Or, more specifically, Ron guards the best wing player on the opposing team most of the time.
So the SG numbers are really more Ron’s and Brown’s than Kobe’s.
T Pain
01/04/2011
Is Shannon Brown that good a defender?
arturogalletti
01/04/2011
T Pain,
You’re seeing confunding with Kobe and Barnes somewhat.
Jimbo (Oz)
12/22/2010
Arturo, love your work, but as others have pointed out above, is there really a correlation between a player’s D, and his opponents effectiveness, for most NBA players ? I would say that generally the answer is no, except maybe for very good defenders on very good defensive teams (like Bowen from the Spurs of old). The NBA experts I trust (eg. David Thorpe and Coach Pop), and my own eyes, tell me that it is almost impossible to stay in front of an opponent and impact his play without having a great team defense providing help. Thoughts everyone ??
Philip
12/23/2010
Jimbo,
Depends on the opponent, and how you want to look at team defense. Teams are great defensively because they force turnovers, rebound the ball well, and limit opponent TS%. I assume you’re referring to limiting opponent TS%. (Plus, getting steals and rebounds is generally pretty consistent year-on-year for players; coaching and teammates don’t have a lot to do with it).
When it comes to perimeter isos, and we want to put the offensive player’s options into the trichotomy of drive right/drive left/shoot, then the best a defender can do is usually to take away one option and limit the other two. High-efficiency, high-volume scorers can usually do a couple of things really well and are passable otherwise – Manu can really drive left and shoot, Lebron can really drive left and right, etc.
(Heck, even only being able to do one thing really well if you’re smart about it. Maggette’s only great talent is driving right, and he’s averaged 16.5 points on 58% TS for his career. But he really picks his moments, burning people on sloppy closeouts, if they don’t take away his right, etc.)
So yeah, I’m with Pop and Thorpe on this.
Post isos are mostly just a case of ability discrepancy – some players contain others flat out, others contain until the double-team arrives. No one could consistently solo guard Shaq in the post in his prime. Anyone can guard a Collins brother. Some people can guard Amare, others can’t. Pretty straightforward.
But it really isn’t that simple, because most NBA basketball isn’t just isos. It’s tons of screen-roll coverage, in which case the man guarding the screener (who hedges out to contest the ball-handler) and the weak-side man (who sags middle to guard against the roller) are arguably more important than the defender covering the ball-handler. When it comes to off-ball screens, Ray Allen’s defender isn’t the only one responsible for him when KG sets a screen; KG’s defender needs to jump out at Allen to contest the 3 if Allen’s defender goes over and Allen fades, or take away Allen’s layup if Allen’s defender stays on Allen’s back and forces him to curl.
And this doesn’t even address rotations, which is what sets apart a lot of good teams. The Thunder and Heat are good defensively in part because they rotate so well, not because they’ve got a superb basket deterrent like Howard is now, or KG, Duncan and Ben Wallace used to be. A lot of teams tend to “stay home” more so rotations aren’t so important, including some good defensive teams like the Bobcats. But it’s a part of every team.
Take a player like Bargnani. Fine individual defender. Horrible off-ball. (And at rebounding). So he’s a terrible defender. Then take a player like Marion, who really couldn’t cut it against a lot of NBA 4s, but kept the Suns respectable because he was so good off-ball. (And at rebounding).
some dude
12/23/2010
Bargnani is not a good on ball defender. Look at his stance. Players easily go by him 1 on 1 in post-ups all the time. He can occasionally use his body to thwart the guy, but usually anyone with any quickness goes around him because his feet and knees are never set right. He’s not the worst at it, but pretty bad. But off-ball is probably. the worst in the league.
Lakers, when they care, are an excellent rotate and hedge team, as well. Especially the Bynum-Odom combo. I thought the Bobcats were more of a “switch” than a “home” team, which is quite rare.
Dallas uses a lot of zone this season, which is also unique. Wonder if teams will copy. They are more clever with their zones, unlike certain teams who randomly use it. It’s also hard to tell when they’re in the zone at times.
Shawn Ryan
12/23/2010
Wow, this is great work Arturo (and shoutout to Nerd Numbers). It seems a bit early for conclusions, but these are concrete findings that can be built from.
random dude
12/27/2010
Well yes, Kobe is a great on ball defender. He may not get 2+ Steals or Blocks a game, but he is one the best lockdown defenders.
The Lakers opponents SGs underperform when they play them.
For those of you that said it’s Ron Artest, think again. Ron usually guards the opposing player small forwards. LeBron James, Paul Pierce, Carmelo Anthony, Kevin Durant, etc those Elite Players are all SFs and covered by Artest.
Kobe probably covers most SGs, and it shows that he is a good defender. Steals and Blocks don’t tell the whole story.The people he covers, including Dwyane Wade, Ray Allen (even though most of the time he covers Rondo for Fisher), etc.
random dude
12/27/2010
On the other hand, great job Arturo. Never thought there would be stats like this.