Scoring: A Volume & Efficiency Analysis

[italianleather]

When it comes to scoring analysis, there are 2 camps, i.e. volume and efficiency. Both sides doesnt seemed to agree. Here is an interesting debate found in this thread, viewtopic.php?f=64&t=1125757&start=135.

To go back to some of the discussions at hand. Why does everyone dismiss Kobe's "volume" scoring like the extra points don't matter at all. Especially when the "volume" is coming at a better efficiency than the majority of Garnett's career (and Hakeem for that matter). When did scoring 35 ppg on 56% TS become chucking?

Do people not realize that even if someone has a 60% TS while scoring 25 points, the difference in efficiency is really not all that great in the grand scheme of things. We are talking about 4%. Over the course of a game that equates to a point or two, and does not come close to matching the extra 10 ppg output of the first.

The dismissal of Kobe's scoring season is quite ridiculous.

You need to consider it from another angle as well though:

Basketball is a game of possessions. Take any given player out of the game, and unless they are an outstanding offensive rebounder, the number of shots per possessions will not be greatly changed. Hence, your direct impact is what you give your team in the shots you take compared to what others could have done.

There's not one clear way to calculate this, but the most obvious way is to simply take the difference in efficiency of the player compared to the league. In '05-06, Kobe averaged 55.9% TS and the league average was 53.6%. The same shots taken with league average efficiency would yield 33.9 points which was 1.5 points less than what Kobe did.

Understand? The direct advantage gained either by bumping up volume or efficiency is not huge.

In the DMJ's example, his formula is as follows:

Points Added = TS(p)%*[2*(FGA(p) + 0.44*FTA(p)] - TS(lavg)%*[2*(FGA(p) + 0.44*FTA(p)]
= [TS(p)% - TS(lavg)%]*[2*(FGA(p) + 0.44*FTA(p)]
= 35.4 - 33.9
= 1.5

*TS(p)% denotes TS% of player
*TS(lavg)% denotes league average TS%

I have a couple of problems with this metric:

1) It focuses entirely on efficiency. Notice the part "2*(FGA(p) + 0.44*FTA(p)]" is a constant throughout the equation
2) It assumes that the player's replacement (and the rest of the team) is able to create shots and draw fouls as effectively

The scoring analysis should include both volume and efficiency. A simple formula that takes both into account is as follows:

Score Metric = Nomarlized TS% * Normalized PPG
= [TS(p)%/TS(lavg)%]*[PPG(p)/PPG(lavg)]

Some numbers from the past 5 yrs scoring leaders:
Durant '11: 3.14 - PPG(p): 27.7 PPG(lavg): 9.4 TS(p): 58.9 TS(lavg)%: 55.3
Durant '10: 3.95 - PPG(p): 30.1 PPG(lavg): 8.3 TS(p): 60.6TS(lavg)%: 55.7
Wade '09 : 3.84 - PPG(p): 30.2 PPG(lavg): 8.3 TS(p): 57.4 TS(lavg)%: 54.4
James '08 : 3.99 - PPG(p): 30.0 PPG(lavg): 7.9 TS(p): 56.8 TS(lavg)%: 54.0
Kobe '07 : 4.12- PPG(p): 31.6 PPG(lavg): 8.2 TS(p): 58.0 TS(lavg)%: 54.2

Data gathered from http://hoopdata.com/advancedstats.aspx.

I cant find data prior to '07 season.

Hopefully we can fine tuned and establish a better metric to evaluate scoring.

[penbeast0]

Fun article I read once that tried to quantify the tradeoff between volume and efficiency:

Usage v. Efficiency for lineups
http://sonicscentral.com/apbrmetrics/vi ... php?t=1679

[italianleather]

"The requested topic does not exist."

[hasslinghoff]

"The requested topic does not exist."

gladly, there's still a backup of the topic:
http://sonicscentral.com/apbrmetrics/vi ... .php?t=132

[Ryoga Hibiki]

I've been out of stat community for years, so I'm not sure this analysis has been actually done.
I believe that a key parameter would be when in the shot clock those shots are taken. A .5 ts% shot is horrile if taken after 5 seconds, but can be a good one if taken after 22.
Is there anything that goes into this topic?

[Paydro70]

I simply don't think it's possible to quantify the "trade off" between efficiency and volume. To answer that question, you would have to know something you cannot: how likely the alternative to that player's shot (a hypothetical shot that did not take place) would be to going in.

Kobe taking a 35% 2pt shot is certainly "inefficient" from a global perspective, but the alternative to that shot is likely not a league average shot, or a team average shot, or some specific teammate's average shot. It might be a turnover, or an even lower likelihood shot. Even looking at shot clock time doesn't really tell you very much, because each play would have to be individually analyzed to see which decisions were made to lead to that shot being taken at that particular time.

[Fencer reregistered]

Let's consider, for a moment, basic zero-sum game theory. If your opponent's best strategy gets him $10/minute from you, and the second best strategy gets him $8/minute, you'll tweak yours to bring that $10 figure down, even if it takes the $8 figure up. Whether they meet at $8.50, $9, or $9.75, you're still better off than you were when he could get $10, and also than when he could get $9.77.

In basketball terms, this means (to a first approximation): If the defense is doing its job right, the efficiency of the offense's several top options will be equal.

In particular, even the best scorer should draw so much defensive attention that he's no more effective at scoring than his teammates.

But if he does draw a lot of attention, his teammates should be more efficient than they otherwise would have been.

Obviously, that's only approximate. Given a choice, you'd rather optimize points per possession over scoring efficiency. (Big difference: TO%) Beyond that, you optimize not just offensive success, but also positioning for the change to other phases of the game (defense and rebounding, although the latter seems to be more in theory than practice these days). Etc.

But anyhow, that's the framework on which to hang a "volume scoring is important" argument. The best stat is one that measures a team's overall effectiveness with a guy on or off the court, and actually succeeds in controlling for other variables. (Hah.) An interesting proxy could be to look at other individuals' efficiency with the guy on or off the court, although I'm not immediately seeing why the proxy would be any easier to compute than the real thing.

[Indeed]

Let's consider, for a moment, basic zero-sum game theory. If your opponent's best strategy gets him $10/minute from you, and the second best strategy gets him $8/minute, you'll tweak yours to bring that $10 figure down, even if it takes the $8 figure up. Whether they meet at $8.50, $9, or $9.75, you're still better off than you were when he could get $10, and also than when he could get $9.77.

In basketball terms, this means (to a first approximation): If the defense is doing its job right, the efficiency of the offense's several top options will be equal.

In particular, even the best scorer should draw so much defensive attention that he's no more effective at scoring than his teammates.

But if he does draw a lot of attention, his teammates should be more efficient than they otherwise would have been.

Obviously, that's only approximate. Given a choice, you'd rather optimize points per possession over scoring efficiency. (Big difference: TO%) Beyond that, you optimize not just offensive success, but also positioning for the change to other phases of the game (defense and rebounding, although the latter seems to be more in theory than practice these days). Etc.

But anyhow, that's the framework on which to hang a "volume scoring is important" argument. The best stat is one that measures a team's overall effectiveness with a guy on or off the court, and actually succeeds in controlling for other variables. (Hah.) An interesting proxy could be to look at other individuals' efficiency with the guy on or off the court, although I'm not immediately seeing why the proxy would be any easier to compute than the real thing.

If that's the case, would team average makes more sense than league average? So this justified a player on a good team vs bad team?
Score Metric = Nomarlized TS% * Normalized PPG (per 40 minutes)
= [TS(p)%/TS(tavg)%]*[PPG(p)/PPG(tavg)]


And should we put squares to it? It seems like it is harder to improve further.
Score Metric = (Normarlized TS% * Normalized PPG) ^ 2 = (Normalized TS%) ^ 2 * (Normalized PPG) ^ 2

[grimballer]

most ppl agree that scoring 25 ppg on 40% isnt that good, however YOU STILL NEED TO SCORE POINTS TO WIN.

so while 25 ppg on 40% isnt that good its still better than lets say 10 ppg on 65%.

someone like tyson chandler couldnt keep up his scoring efficiency of 70% ts, if given 10 or 15 fga instead of 5.

likewise someone like kevin martin wouldnt shoot 43% if he wasnt expected to score over 20 ppg n take 15 fga.

how to determine which one is better?

how about

fgm x fg% or ppg x ts%?

so in martin vs chandler example

martin = 6.9 x .436 = 3.01, 23.5 x .601 = 14.12
chandler = 3.6 x .654 = 2.35, 10.1 x .697 = 7.04

in this example martins volume scoring > chandlers efficient scoring.

how about an example where there isnt much difference in volume?

lets say rose vs howard?

8.8 x .445 = 3.9, 25.0 x .550 = 13.75
7.9 x .593 = 4.68, 22.9 x .616 = 14.11


so even if rose scores more ppg, howard scoring is better.

[EvanZ]

I've come up with a marginal rating system. This post describes the basics and gives ratings for inside scoring. Overall (meaning inside+mid-range+3pt), here are the ratings of all players who played 40+ games and 25+ mpg:

Code: Select all

TRNK   NAME   TOT
1   LeBron James   2.90
2   Dirk Nowitzki   2.68
3   Stephen Curry   2.47
4   Dwight Howard   2.00
5   Paul Pierce   1.95
6   Ray Allen   1.92
7   Tony Parker   1.87
8   Kevin Durant   1.72
9   Steve Nash   1.67
10   Shawn Marion   1.65
11   Paul Millsap   1.61
12   Beno Udrih   1.57
13   Kevin Garnett   1.56
14   Dwyane Wade   1.45
15   Thaddeus Young   1.42
16   Pau Gasol   1.39
17   Luke Ridnour   1.31
18   Mike Dunleavy   1.30
19   Al Horford   1.30
20   Wilson Chandler   1.29
21   Marcin Gortat   1.29
22   Reggie Williams   1.26
23   Grant Hill   1.24
24   Rudy Gay   1.14
25   Amare Stoudemire   1.13
26   Anthony Morrow   1.10
27   David West   1.08
28   Elton Brand   1.00
29   Luis Scola   0.97
30   Ty Lawson   0.95
31   Carlos Boozer   0.95
32   Luol Deng   0.91
33   David Lee   0.91
34   Jared Dudley   0.89
35   Tayshaun Prince   0.86
36   Nene Hilario   0.85
37   Richard Jefferson   0.85
38   Lamar Odom   0.85
39   LaMarcus Aldridge   0.84
40   Deron Williams   0.82
41   Brandon Bass   0.82
42   Blake Griffin   0.81
43   Zach Randolph   0.80
44   Chauncey Billups   0.80
45   Jason Terry   0.76
46   Jameer Nelson   0.74
47   Chris Paul   0.73
48   Chris Bosh   0.66
49   Hedo Turkoglu   0.63
50   Arron Afflalo   0.62
51   Kevin Martin   0.62
52   Andrew Bynum   0.58
53   Serge Ibaka   0.57
54   Ben Gordon   0.54
55   Josh Smith   0.48
56   Danny Granger   0.43
57   Nick Young   0.41
58   Mike Bibby   0.39
59   Carl Landry   0.33
60   Monta Ellis   0.32
61   Marvin Williams   0.32
62   Eric Gordon   0.31
63   Amir Johnson   0.30
64   Nicolas Batum   0.26
65   George Hill   0.26
66   Tyson Chandler   0.24
67   Jrue Holiday   0.21
68   Kobe Bryant   0.20
69   Dorell Wright   0.18
70   Derrick Rose   0.15
71   Kirk Hinrich   0.09
72   Jeff Green   0.08
73   Mike Conley   0.08
74   Darren Collison   0.07
75   Vince Carter   0.07
76   Sasha Vujacic   0.07
77   Manu Ginobili   0.04
78   Carmelo Anthony   0.02
79   Wesley Matthews   0.01
80      0.00
81   Wesley Johnson   -0.00
82   Michael Beasley   -0.02
83   Kevin Love   -0.02
84   Jose Calderon   -0.03
85   Marc Gasol   -0.03
86   Kris Humphries   -0.04
87   J.J. Redick   -0.04
88   Raymond Felton   -0.10
89   Vladimir Radmanovic   -0.14
90   Jason Richardson   -0.14
91   Jamal Crawford   -0.14
92   Joe Johnson   -0.16
93   Tim Duncan   -0.16
94   Daniel Gibson   -0.19
95   Shane Battier   -0.21
96   Kenyon Martin   -0.23
97   Boris Diaw   -0.25
98   Andre Miller   -0.28
99   DeMar DeRozan   -0.28
100   Landry Fields   -0.28
101   Brook Lopez   -0.30
102   C.J. Miles   -0.31
103   Richard Hamilton   -0.31
104   Channing Frye   -0.31
105   Antawn Jamison   -0.33
106   JaVale McGee   -0.34
107   Andrei Kirilenko   -0.34
108   DeAndre Jordan   -0.36
109   Kyle Lowry   -0.37
110   Emeka Okafor   -0.41
111   Rajon Rondo   -0.45
112   Al Jefferson   -0.47
113   Ramon Sessions   -0.50
114   Andrew Bogut   -0.54
115   Andre Iguodala   -0.56
116   Danilo Gallinari   -0.57
117   Brandon Rush   -0.60
118   O.J. Mayo   -0.62
119   Carlos Delfino   -0.63
120   Ryan Gomes   -0.64
121   Jodie Meeks   -0.67
122   Russell Westbrook   -0.68
123   Andrea Bargnani   -0.72
124   Rodney Stuckey   -0.76
125   Gerald Wallace   -0.80
126   Glen Davis   -0.84
127   Devin Harris   -0.85
128   John Salmons   -0.87
129   Baron Davis   -0.89
130   J.J. Hickson   -0.91
131   Anthony Parker   -0.92
132   Acie Law   -0.93
133   Ekpe Udoh   -0.94
134   Greg Monroe   -0.96
135   Raja Bell   -0.96
136   Stephen Jackson   -0.98
137   James Harden   -1.02
138   Joakim Noah   -1.07
139   Luc Mbah a Moute   -1.07
140   Derek Fisher   -1.10
141   Travis Outlaw   -1.12
142   Andray Blatche   -1.22
143   Andris Biedrins   -1.25
144   D.J. Augustin   -1.33
145   Ersan Ilyasova   -1.33
146   Ron Artest   -1.45
147   Roy Hibbert   -1.45
148   Kwame Brown   -1.49
149   Brandon Jennings   -1.49
150   John Wall   -1.57
151   Brandon Roy   -1.60
152   Tyreke Evans   -1.65
153   Trevor Ariza   -1.66
154   DeMarcus Cousins   -1.70
155   Thabo Sefolosha   -1.74
156   Chuck Hayes   -1.75
157   Jason Kidd   -1.78
158   Lou Amundson   -2.08
159   Marcus Camby   -2.96

[jambalaya]

I've been out of stat community for years, so I'm not sure this analysis has been actually done.
I believe that a key parameter would be when in the shot clock those shots are taken. A .5 ts% shot is horrible if taken after 5 seconds, but can be a good one if taken after 22.
Is there anything that goes into this topic?

Yes several studies have addressed this issue recently. One at the last Sloan Conference
http://www.sloansportsconference.com/?p=652

and this one:
http://arxiv.org/abs/1107.5793

They agree with your basic observation and look at the issue in greater specificity.