Analyzing the Golden State Warriors Offense in 2010-11: A Synergy Perspective

For my first front page story on GSOM I want to delve into (what else?) some hardcore stats. This is not going to be an ezPM post. Instead, I want to look into Synergy offensive stats for the Warriors this past season (the defensive story will be featured in a sequel). For those who are unfamiliar with Synergy, it is a service that breaks down game film and records statistics for every play that ends in a shot, turnover, or free throw attempt. The hope is that these stats can help give us a deeper insight into the strengths and weaknesses of our favorite NBA team. 

 

I recently did a league-wide analysis of Synergy stats in a couple of posts (here and here) over at my personal blog. You may want to read those before proceeding, but it should not be necessary, as I'll review some of the main findings in this article.

Synergy classifies plays into 11 different types:

 

  • Isolation (ISO)
  • P&R Ball Handler (BALL)
  • Post-Up (POST)
  • P&R Man (ROLL)
  • Spot-Up (SPOT)
  • Off Screen (SCREEN)
  • Hand off (HAND)
  • Cut (CUT)
  • Offensive Rebound (REB)
  • Transition (TRANS)
  • All other plays (OTHER)

Synergy records the efficiency and frequency for each play type for each player. The efficiency is given in terms of points per play (PPP). The frequency is simply the percentage of total plays in which that particular play type was used (or defended) by the player. The total team efficiency is, thus, a summation of all the individual efficiencies and frequencies (I often call this "rate" interchangeably) for each player. Here are a couple of plots that summarize the league-wide rates (Plot 1) and efficiencies (Plot 2) by play type: Off_play_type_bw_medium

Off_ppp_play_type_bw_medium

These are called "box and whisker" plots. The black line across the middle of each box represents the median value for the league. The top and bottom of the box represents the 25-75th %-iles. The "whiskers" represent the 5th and 95th %-iles, respectively. Any circles found outside that range represent outliers (crazy bad or crazy good teams, depending on which side of the whiskers they reside). In these plots, I have added a blue line which represents the Warriors. 

Here are the data for the Warriors in tabular format (sorted in descending order by play frequency):

 

PLAY

RATE

SDRATE

PPP

SDPPP

SPOT

20.2%

0.141

1.05

1.132

ISO

14.0%

0.763

0.85

0.265

BALL

13.7%

0.710

0.88

1.544

TRANS

13.4%

0.685

1.20

1.278

CUT

7.4%

-0.355

1.19

-0.952

OTHER

6.6%

-0.841

0.39

-0.867

POST

6.6%

-1.118

0.76

-1.626

REB

6.1%

0.102

0.98

-1.899

ROLL

5.2%

0.049

0.96

-0.611

SCREEN

4.6%

-0.287

0.86

-0.369

HAND

2.2%

-0.225

0.99

8.796

 

 

The two columns titled SDRATE and SDPPP, respectively, represent the RATE and PPP values in terms of standardized units (S.U.) with respect to the league. For example, look at the row labeled ROLL. The Warriors run that play 5.2% of the time. That seems like a very small amount, right? Well, the standardized rate of 0.049 tells us that it is actually just ever so slightly above the league average (a value of 0.0 would be exactly league average). On the other hand, look at the POST row. You'll see that POST plays are run 6.6% of the time, which is more than ROLL plays, in absolute terms; however, the SDRATE for POST plays is -1.118, which means that we run those at a rate that is about one standard deviation less than the league-wide average. In other words, we don't run post plays that much compared to other teams. These same data are represented graphically in the plots above. So, the blue line for ROLL should be right about where the black line is in the frequency plot, whereas the blue line for POST is actually located below the yellow box, but above the bottom whisker.

 

Hopefully, all that made some sense. If not, feel free to ask for further clarification in the comments section.

 

So, now with those data, we can begin to make objective judgements about where the team performed well with respect to the league, and where they did not perform so well. In terms of efficiency, the strengths of the team offensively appear to be in running SPOT, BALL, and TRANS plays, while the main weaknesses are REB (plays coming immediately after offensive rebounds, for example, tip-ins), POST (no kidding?), and CUT. Overall, the Warriors tied for 11th in the league with a team PPP of 0.95 (0.376 S.U.).

 

A good question to ask at this point is how do we know which plays are the most important for having a high overall efficiency rating. I tried to answer this question in one of those blog posts mentioned earlier. To do this, I took all the efficiency and rate data for each team and threw them into a multiple linear regression model. Here are the results of that analysis:

Call:
lm(formula = TOT ~ SPOT PPP + REB PPP + TRANS PPP + TRANS RATE +
    ISO PPP + POST PPP + OTHER PPP + ISO RATE, data = PPP2011_off_pivot)

Residuals:
     Min       1Q   Median       3Q      Max
-0.39784 -0.13765  0.02817  0.11230  0.35331 

Coefficients:
             Estimate Std. Error t value Pr(>|t|)
(Intercept)  0.004622   0.042871   0.108  0.91516
SPOT PPP     0.526211   0.056530   9.308 6.69e-09 ***
REB PPP      0.288688   0.048475   5.955 6.54e-06 ***
TRANS PPP    0.253230   0.055811   4.537  0.00018 ***
TRANS RATE    0.247018   0.048504   5.093 4.82e-05 ***
ISO PPP      0.205189   0.062470   3.285  0.00353 **
POST PPP     0.197200   0.052158   3.781  0.00110 **
OTHER PPP    0.172761   0.049844   3.466  0.00231 **
ISO RATE     -0.139141   0.049104  -2.834  0.00995 **
---
Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1 

Residual standard error: 0.2348 on 21 degrees of freedom
Multiple R-squared: 0.9601,	Adjusted R-squared: 0.9448
F-statistic:  63.1 on 8 and 21 DF,  p-value: 5.974e-13

It turns out that 95% of PPP can be explained by the 8 variables listed in the table, 6 of which are efficiencies, and only 2 of which are rates. The first column of values represents the standardized coefficient. So, by comparing each coefficient, we can judge the relative importance of each parameter. It turns out that the efficiency (PPP) of SPOT plays is by a fairly wide margin the most important variable. Fortunately, the Warriors were very good at this (>1 S.U.). The next most important variable was REB PPP, which the Warriors were the worst in the league (-1.899 S.U.). The next two categories are the TRANS PPP (1.278 S.U.) and RATE (0.685 S.U.), both of which were relative strengths. The team was a little better than average in ISO efficiency (0.265 S.U.), but very bad in POST efficiency (-1.626 S.U.). The team was not good in OTHER plays (-0.867 S.U.), but it's hard to know really what that even means. How can you improve in OTHER? I have no clue or insight to offer, unfortunately. The last variable in the model, ISO RATE, is perhaps, the most interesting. First, it's the only rate in the model other than TRANS. Second, it's the only one with a negative coefficient. That means, in theory, anyway, that overall efficiency at the team level actually suffers the more that ISO plays are run. In other words, there is an inverse correlation. Unfortunately, for the Warriors, that's not good, because we run ISO plays at a significantly higher rate than the league average (0.763 S.U.). This is something that is often discussed at GSOM. Too many isolation plays is not a good thing. Here are the stats that actually validate that theory. 

It should be noted that, since there are only 8 variables in the model, there are many other play types that are apparently not important. Where are the pick and roll plays? Screens? Cuts? All those are neglected by the model. Well, let me simply quote my own summary from the blog:

What these regression models suggest is that for the most part, efficiency — much more than play frequency — accounts for overall team efficiency. In other words, whatever plays you run or defend, the key is to run them efficiently not simply more. It's not how many post plays you run, but how efficiently you can run them. It's not how many spot up plays you generate, but how efficiently you hit those shots. And so on — at least, within the range of play frequencies that NBA teams typically run. I'm certainly not suggesting that a team could run all post plays or no post plays and still hope to compete. That's not how it works. What the data show comes as a result of years of optimization by players, coaches, and GM's of personnel and strategies. What I would suggest, however, is that the models shown here represent the current state of the NBA as of 2011. If I had access to previous years of data, my strong guess is that the models would look vastly different. Regression models are meant to explain the data that are fed to them, and should not be used to extrapolate or predict the results of parameters outside that range. This should go without saying, but I say it, nevertheless to shield myself from those obvious questions.

 

In other words, for example, I'm not saying pick and roll plays are unimportant, but it could be that given the rate at which they are run and the capability of most point guards in the league, this is simply not much of a differentiating factor between teams.

Given all that as introduction (believe it or not!), I want to now move on to analysis of the Warriors in each of the categories that actually made it into the model, with the assumption that these are the things that we should focus on, because they have been objectively identified to be the most critical.

Spot-Up

The league average is 0.98 PPP. This was our best category (1.05 PPP). You can see that Williams and Curry, not surprisingly to many of us, are monsters in terms of spot-up plays. Radmanovic and Wright were above league average and our team average, too. Ellis was above league average, but this is clearly not his strength. He was below the team average. For all the talk of Lee's mid-range game, his spot-up efficiency is simply not that great. And then it gets worse with Law, Amundson, and Udoh.

NAME

PLAY

RATE

PPP

LEAGUE RK

Williams

Spot-Up

31.4%

1.29

7

Curry

Spot-Up

17.8%

1.25

14

Radmanovic

Spot-Up

40.2%

1.08

75

Wright

Spot-Up

38.7%

1.06

93

Ellis

Spot-Up

11.5%

1.01

142

Lee

Spot-Up

13.2%

0.85

252

Law

Spot-Up

22.0%

0.76

309

Amundson

Spot-Up

8.7%

0.67

-

Udoh

Spot-Up

9.1%

0.52

351

 

Offensive Rebound

League average is 1.08 PPP. Remember, this is our worst category (0.98 PPP). I remember Lee missing a lot of easy putbacks early in the season and after his injury, but he must have improved over the rest of the season, because he comes out looking pretty good here. Biedrins, Udoh, Amundson, and Radmanovic are all bad. This is an area that must be addressed. Thinking about players that might help here, Marc Gasol (1.27 PPP, 8.3%) and Nene Hilario (1.37 PPP, 7.3%) are obvious answers . DeAndre Jordan (1.06 PPP, 27.7%) is better than Biedrins but still below league average.

NAME

PLAY

RATE

PPP

LEAGUE RK

Law

REB

1.8%

1.5

-

Curry

REB

1.9%

1.16

-

Lee

REB

10.9%

1.16

56

Wright

REB

4.0%

1.09

90

Biedrins

REB

16.8%

1

135

Williams

REB

3.9%

0.96

154

Udoh

REB

13.5%

0.93

165

Amundson

REB

27.7%

0.81

185

Ellis

REB

0.7%

0.79

-

Radmanovic

REB

10.2%

0.7

194

 

Transition PPP and Rate

This was one of our main strengths. League average is 1.15 PPP, and we were 1.20 PPP, only trailing Miami (1.22) and OKC (1.21). Law, Ellis, and Wright clearly were the main contributors here. Curry was only average. My opinion on this is that while we excelled in transition, we did it mostly by using risky defensive strategies ("going for steals") that hurt us on the defensive end in other ways. Miami was the best team in transition, but when you have LeBron and Wade, you can create transition plays without weakening the defense as much.

NAME

PLAY

RATE

PPP

LEAGUE RK

Udoh

Transition

5.1%

1.47

-

Radmanovic

Transition

11.0%

1.40

18

Lee

Transition

5.7%

1.27

80

Amundson

Transition

6.6%

1.25

-

Law

Transition

27.4%

1.23

99

Ellis

Transition

13.6%

1.22

103

Wright

Transition

20.1%

1.21

112

Williams

Transition

17.5%

1.18

128

Curry

Transition

16.8%

1.15

151

Biedrins

Transition

2.4%

0.75

-

 

 

Isolation

League average is 0.84 PPP. Warriors were just above league average at 0.85 PPP. Notice how much lower isolation efficiency is compared to other types of plays we have discussed so far. The best team in the league - can you guess? - was Miami (0.92 PPP). Interestingly, and I wouldn't have thought this, Ellis was simply not very efficient in isolation, yet he ran it more than any other player. Isolation plays clearly have a place in the league, but my analysis has shown that they should be minimized, and when run, they must be run effectively (i.e. efficiently). It's clear to me that Keith Smart should have tried to rein in Ellis.

NAME

PLAY

RATE

PPP

LEAGUE RK

Curry

Isolation

14.6%

0.93

44

Udoh

Isolation

5.1%

0.93

-

Lee

Isolation

10.0%

0.91

56

Wright

Isolation

12.8%

0.89

69

Ellis

Isolation

23.9%

0.83

116

Williams

Isolation

10.8%

0.82

125

Radmanovic

Isolation

9.5%

0.73

175

Amundson

Isolation

4.5%

0.64

-

Law

Isolation

11.7%

0.54

248

Biedrins

Isolation

1.2%

0.50

-

 

Post-Up

Ugh. League average was 0.87 PPP. The Warriors were the second to worst team in the league with a 0.76 PPP,  ahead of only Sacramento. It might surprise you to learn that Dallas led the league with a 1.02 PPP. The Warriors, indeed, have quite a ways to improve here. Biedrins and Udoh were woeful. Lee was a little better than woeful. Miserable? Anyway, when your best post-up player by far is Lou Amundson, you know some things are broke. Again, to give some comps, Nene Hilario had a 1.01 PPP (ranked #24) is very good here. Marc Gasol (0.87 PPP, ranked #81) is about average (but better than what we have), and DeAndre Jordan (0.72 PPP, #150) is probably not going to help much.

NAME

PLAY

RATE

PPP

LEAGUE RK

Williams

Post-Up

1.0%

0.86

-

Amundson

Post-Up

8.7%

0.86

-

Wright

Post-Up

2.0%

0.85

93

Ellis

Post-Up

2.7%

0.83

101

Lee

Post-Up

15.6%

0.78

119

Biedrins

Post-Up

34.5%

0.67

163

Udoh

Post-Up

30.0%

0.64

169

Law

Post-Up

1.8%

0.50

-

Radmanovic

Post-Up

1.5%

0.33

-


Summary

So, I hoped this helped you get a picture of where we stand on offense. Some things really need to improve, while other things are actually pretty good. Likewise, some things may be easier to fix than others. Finding a guy who can really help us with post-ups may be really difficult, but reducing the number of Ellis isolation plays is very easy. Just find a coach who has the cajones and wisdom to do it. Next time, I'll discuss defense. Remember, we're actually ok on offense, overall. I promise that looking at the defense is going to be a truly eviscerating experience for y'all!

X
Log In Sign Up

forgot?
Log In Sign Up

Forgot password?

We'll email you a reset link.

If you signed up using a 3rd party account like Facebook or Twitter, please login with it instead.

Forgot password?

Try another email?

Almost done,

Join Golden State Of Mind

You must be a member of Golden State Of Mind to participate.

We have our own Community Guidelines at Golden State Of Mind. You should read them.

Join Golden State Of Mind

You must be a member of Golden State Of Mind to participate.

We have our own Community Guidelines at Golden State Of Mind. You should read them.

Spinner.vc97ec6e

Authenticating

Great!

Choose an available username to complete sign up.

In order to provide our users with a better overall experience, we ask for more information from Facebook when using it to login so that we can learn more about our audience and provide you with the best possible experience. We do not store specific user data and the sharing of it is not required to login with Facebook.

tracking_pixel_9347_tracker