APBRmetrics Forum Index APBRmetrics
The statistical revolution will not be televised.
 
 FAQFAQ   SearchSearch   MemberlistMemberlist   UsergroupsUsergroups   RegisterRegister 
 ProfileProfile   Log in to check your private messagesLog in to check your private messages   Log inLog in 

Steals Redux

 
Post new topic   Reply to topic    APBRmetrics Forum Index -> General discussion
View previous topic :: View next topic  
Author Message
Ed Küpfer



Joined: 30 Dec 2004
Posts: 616
Location: Toronto

PostPosted: Thu Mar 17, 2005 4:45 am    Post subject: Steals Redux Reply with quote

To estimate the value of a steal, we need to look at the cost of a failed steal attempt. Obviously a successful steal is valuable -- equal not only to the expected number of points allowed per possession, but also to the increase in offensive efficiency following a steal. What is an unsuccessful steal worth? And more to the point, how do we determine whether a steal attempt is worthwhile?

The expected number of points in a possession cycle following a steal attempt can be calculated by the following equation:

Code:
ExpPts =p(STL) * ORTG * MarSTLVal + [1 - p(STL)] * (-MarSTLCost * DRTG + ORTG)


Lemmee define those terms: p(STL) is the probability of a steal attempt being successful. ORTG is the expected number of points to be scored in a possession. MarSTLVal is the marginal steal value, the increase in ORTG following a steal, expresses as a percentage: I think someone said offensive efficiency following a steal goes up 30%, so MarSTLVal would equal 130%. DRTG is the expected number of points to be allowed in a possession. MarSTLCost is the increase in the expected number of points allowed following a failed steal attempt, expressed as a percentage: if a failed steal attempt increases offensive efficiency by 20%, then MarSTLCost would be 120%.

A sample calculation: two teams of equal offensive and defensive ability, ORTG and DRTG both equal to 1.0 (or 100 points per 100 possessions). Imagine that a successful steal leads to an increase in offensive efficiency of 30%, and that a failed steal attempt gives the offense a boost of 20%. TeamA, on defense, has a steal success rate of 30%. What is the expected outcome should they decide to attempt a steal?

Code:
ExpPts =p(STL) * ORTG * MarSTLVal + [1 - p(STL)] * (-MarSTLCost * DRTG + ORTG)
       = 0.3 * 1.0 * 1.3 + [1 - 0.3 ] * (-1.2 * 1.0 + 1.0)
       = 0.25


Meaning TeamA will outscore their opponents by 25 points per 100 possessions.

It can be shown that using the Marginal value and costs numbers above, and assuming equal strength teams, that the "break even" point -- the point above which the defense experiences a net benefit by attempting a steal, and below which they will be outscored -- the break even point for steal success rate ( the p(STL) term above) is 13%. If the marginal cost is lowered to 110% -- that is, if the cost of attempting a steal and failing is that the offense increases its efficiency by 10% -- the break even point becomes only 7%.

Now, hold the steal success rate steady at 30%, and the marginal value at 130%. Where is the break even point for marginal costs? A failed steal attempt would have to cost the defense a 56% increase in opponents offensive efficiency for a steal to be not worth the attempt -- an unlikely scenario.

You can plug the equation into a spreadsheet to come up with your own scenarios. From fooling around with it for a few minutes, I get the feeling that steals are probably very worthwhile.
_________________
ed
Back to top
View user's profile Send private message
kjb



Joined: 03 Jan 2005
Posts: 671
Location: Washington, DC

PostPosted: Thu Mar 17, 2005 9:31 am    Post subject: Reply with quote

I've been tracking "steal attempts" for the Wizards over the past 15-20 games. I have not compiled the data yet, but eyeballing the individual scoresheets, my sense is that the success rate on "steal attempts" is less than 10% -- BUT for most steal attempts there isn't much of a negative. There is some negative that comes from steal attempts that leaves a defensive player out of position. I've been tracking that too, but over a shorter period. Here again, I haven't added up all the data yet.

One thing I quickly found when trying to track steals is that there are different kinds of steal attempts. I've been tracking 2 broad categories -- "reaches" and "attempts to steal passes". Reach attempts are far more plentiful, and generally less downside. The passing lane attempts have a greater negative because the defender tends to get out of position more often. Neither appears to have more value in translating to the offensive end.

Another important point is that a lot of steals "just happen." In other words, the defender isn't actively trying to steal the ball -- the offensive player just makes a mistake.
_________________
My blog
Back to top
View user's profile Send private message Visit poster's website AIM Address Yahoo Messenger
S.K.



Joined: 18 Feb 2005
Posts: 52
Location: Toronto

PostPosted: Thu Mar 17, 2005 11:23 am    Post subject: Reply with quote

This is actually similar to the way steals were sabrmetrically analyzed as an offensive strategy in baseball. Just a funny little coincidence.

Anyway, I'd find it very interesting to track steal% for players on an individual basis (I know, add it to the list!). I'm thinking that a guy like Jason Williams, who has lots of steals but not much defensive value otherwise, is going to have a much lower % of successful steals than someone like Andrei Kirilenko. A combination of total steals and steal% would be exponentially better as an indicator of defensive skill than steals alone.
_________________
No books - no articles - no website.
Just opinions.

Ill-informed opinions.
Back to top
View user's profile Send private message Send e-mail MSN Messenger
kjb



Joined: 03 Jan 2005
Posts: 671
Location: Washington, DC

PostPosted: Thu Mar 17, 2005 11:37 am    Post subject: Reply with quote

S.K. wrote:
This is actually similar to the way steals were sabrmetrically analyzed as an offensive strategy in baseball. Just a funny little coincidence.

Anyway, I'd find it very interesting to track steal% for players on an individual basis (I know, add it to the list!). I'm thinking that a guy like Jason Williams, who has lots of steals but not much defensive value otherwise, is going to have a much lower % of successful steals than someone like Andrei Kirilenko. A combination of total steals and steal% would be exponentially better as an indicator of defensive skill than steals alone.


This is exactly what I'm doing for the Wizards -- tracking it individually. I'll know more when I can add up all the data (which means deciphering my notations), but my impression is that Arenas tends to get out of position more often than does Hughes. Jamison seems to get most of his steals just by standing in the right place. Smile
_________________
My blog
Back to top
View user's profile Send private message Visit poster's website AIM Address Yahoo Messenger
S.K.



Joined: 18 Feb 2005
Posts: 52
Location: Toronto

PostPosted: Thu Mar 17, 2005 11:57 am    Post subject: Reply with quote

WizardsKev wrote:
S.K. wrote:
This is actually similar to the way steals were sabrmetrically analyzed as an offensive strategy in baseball. Just a funny little coincidence.

Anyway, I'd find it very interesting to track steal% for players on an individual basis (I know, add it to the list!). I'm thinking that a guy like Jason Williams, who has lots of steals but not much defensive value otherwise, is going to have a much lower % of successful steals than someone like Andrei Kirilenko. A combination of total steals and steal% would be exponentially better as an indicator of defensive skill than steals alone.


This is exactly what I'm doing for the Wizards -- tracking it individually. I'll know more when I can add up all the data (which means deciphering my notations), but my impression is that Arenas tends to get out of position more often than does Hughes. Jamison seems to get most of his steals just by standing in the right place. Smile


Right, but I think we'd have to have a lot more data than you yourself are capable of gathering in order to make judgements about whether it's a specific skill or just something that happens (ie Arenas might just be getting unlucky lately and Jamison lucky).
_________________
No books - no articles - no website.
Just opinions.

Ill-informed opinions.
Back to top
View user's profile Send private message Send e-mail MSN Messenger
kjb



Joined: 03 Jan 2005
Posts: 671
Location: Washington, DC

PostPosted: Thu Mar 17, 2005 12:10 pm    Post subject: Reply with quote

S.K. wrote:
WizardsKev wrote:
S.K. wrote:
This is actually similar to the way steals were sabrmetrically analyzed as an offensive strategy in baseball. Just a funny little coincidence.

Anyway, I'd find it very interesting to track steal% for players on an individual basis (I know, add it to the list!). I'm thinking that a guy like Jason Williams, who has lots of steals but not much defensive value otherwise, is going to have a much lower % of successful steals than someone like Andrei Kirilenko. A combination of total steals and steal% would be exponentially better as an indicator of defensive skill than steals alone.


This is exactly what I'm doing for the Wizards -- tracking it individually. I'll know more when I can add up all the data (which means deciphering my notations), but my impression is that Arenas tends to get out of position more often than does Hughes. Jamison seems to get most of his steals just by standing in the right place. Smile


Right, but I think we'd have to have a lot more data than you yourself are capable of gathering in order to make judgements about whether it's a specific skill or just something that happens (ie Arenas might just be getting unlucky lately and Jamison lucky).


I agree. Ideally, this would be tracked for every player in the league. Smile
_________________
My blog
Back to top
View user's profile Send private message Visit poster's website AIM Address Yahoo Messenger
Ed Küpfer



Joined: 30 Dec 2004
Posts: 616
Location: Toronto

PostPosted: Fri Mar 18, 2005 9:46 pm    Post subject: Reply with quote

Kevin, I don't have to tell you how much I want to see your numbers. Please post them whenever you get a chance. The steal "types" are categories I never even thought of, but should have.
_________________
ed
Back to top
View user's profile Send private message
kjb



Joined: 03 Jan 2005
Posts: 671
Location: Washington, DC

PostPosted: Sat Mar 19, 2005 6:52 am    Post subject: Reply with quote

Ed Küpfer wrote:
Kevin, I don't have to tell you how much I want to see your numbers. Please post them whenever you get a chance. The steal "types" are categories I never even thought of, but should have.


I just need to get the time to add it all up. Work has been a bear this year and I've had far less time to devote to hoops-related stuff.
_________________
My blog
Back to top
View user's profile Send private message Visit poster's website AIM Address Yahoo Messenger
Display posts from previous:   
Post new topic   Reply to topic    APBRmetrics Forum Index -> General discussion All times are GMT - 5 Hours
Page 1 of 1

 
Jump to:  
You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot vote in polls in this forum


Powered by phpBB © 2001, 2005 phpBB Group