Author Topic: Statistics as they related to defense (split from game thread)  (Read 8993 times)

0 Members and 0 Guests are viewing this topic.

Re: Statistics as they related to defense (split from game thread)
« Reply #15 on: February 01, 2009, 03:57:58 PM »

Offline KCattheStripe

  • Danny Ainge
  • **********
  • Posts: 10726
  • Tommy Points: 830
So the top guys in steals this year are Chris Paul, Rondo, Kidd, Wade, Lebron, Chalmers, Gerald Wallace, Iguodala, Ariza, and Caron Butler.

The top blocks guys are D Howard, Camby, Turiaf, Brook Lopez, Bynum, Duncan, Daelembert, and Okafor

So I am to conclude these are the worst defenders in the league and the guys with zero blocks and steals are the best because they are the best positioners who probably always seal off the lane?

I don't think anybody is arguing that steals and blocks are irrelevant to good defense.  However, look at your own list.  Do you consider Jason Kidd to be a good defender at this stage of his career?  Is Turiaf the third best defensive big man in the game?  Is Chris Andersen a top-four defender?  Is Al Jefferson a good defender, despite being #12 in blocks?  Of course not.  They have good statistics, but aren't top-tier defenders.

On the other hand, there are plenty of excellent defenders who do have good stats (Rondo, Kirilenko, Igoudala, etc.)  It's very difficult to quantify defense statistically, which is why observation is so important.

Well whats the argument, "good" or "top-tier"? I would argue that Al Jeffs' numbers are an indicator of being a good defender but not automatically a top-tier one.

Re: Statistics as they related to defense (split from game thread)
« Reply #16 on: February 01, 2009, 04:15:16 PM »

Offline Roy Hobbs

  • In The Rafters
  • The Natural
  • *********************************
  • Posts: 33333
  • Tommy Points: 6430
  • Doc could learn a thing or two from Norman Dale
So the top guys in steals this year are Chris Paul, Rondo, Kidd, Wade, Lebron, Chalmers, Gerald Wallace, Iguodala, Ariza, and Caron Butler.

The top blocks guys are D Howard, Camby, Turiaf, Brook Lopez, Bynum, Duncan, Daelembert, and Okafor

So I am to conclude these are the worst defenders in the league and the guys with zero blocks and steals are the best because they are the best positioners who probably always seal off the lane?

I don't think anybody is arguing that steals and blocks are irrelevant to good defense.  However, look at your own list.  Do you consider Jason Kidd to be a good defender at this stage of his career?  Is Turiaf the third best defensive big man in the game?  Is Chris Andersen a top-four defender?  Is Al Jefferson a good defender, despite being #12 in blocks?  Of course not.  They have good statistics, but aren't top-tier defenders.

On the other hand, there are plenty of excellent defenders who do have good stats (Rondo, Kirilenko, Igoudala, etc.)  It's very difficult to quantify defense statistically, which is why observation is so important.

Well whats the argument, "good" or "top-tier"? I would argue that Al Jeffs' numbers are an indicator of being a good defender but not automatically a top-tier one.

Big Al is not a good defender, or even an average one.  How often do you watch Twolves games?  Routinely, guys blow by Big Al, but there's no statistic that's going to show that (although you can look at Opponent's Production -- as flawed as it is -- and see that opponents are shooting an eFG% of 52.9% against him.)

All the negativity in this town sucks. It sucks, and it stinks, and it sucks. - Rick Pitino

Portland CrotoNats:  2009 CB Draft Champions

Re: Statistics as they related to defense (split from game thread)
« Reply #17 on: February 01, 2009, 04:43:41 PM »

Offline Fafnir

  • Bill Russell
  • ******************************
  • Posts: 30863
  • Tommy Points: 1330
If someone is open for an outside shot it's because someone else didn't keep his man from penetrating and then a third guy was late in rotation.  
This isn't true. Good ball reversal and passing can create an open shot even if every man successfully executes his assignments and rotation.

Now the shot will still be challenged, or maybe the shooter will be run off the 3-point line. But it will be open.

Re: Statistics as they related to defense (split from game thread)
« Reply #18 on: February 01, 2009, 06:36:03 PM »

Offline BballTim

  • Dave Cowens
  • ***********************
  • Posts: 23724
  • Tommy Points: 1123
I'm sure no one looks forward to playing Rondo but the fact is that his teammates many times have to bail him out because of his gambling for steals.  He is not good at keeping his man in front of him and that many times results in someone elses man getting an open shot do to rotations.
This is what absolutely drives me crazy about anyone talking about one guy shutting another guy out or so and so letting his man light us up from outside.  That never EVER happens.  Defense is a team games when played correctly and the Celtics do it very, very well. If someone is open for an outside shot it's because someone else didn't keep his man from penetrating and then a third guy was late in rotation. 

  Rondo is better than average at keeping his man in front of him. Opposing PGs have a lower than average eFG%, take a smaller than average percentage of inside shots, get fewer than average assists and more than average turnovers. Also, there's more to defense than just keeping your man in front of you.

Re: Statistics as they related to defense (split from game thread)
« Reply #19 on: February 01, 2009, 10:53:38 PM »

Offline Fafnir

  • Bill Russell
  • ******************************
  • Posts: 30863
  • Tommy Points: 1330
I'm sure no one looks forward to playing Rondo but the fact is that his teammates many times have to bail him out because of his gambling for steals.  He is not good at keeping his man in front of him and that many times results in someone elses man getting an open shot do to rotations.
This is what absolutely drives me crazy about anyone talking about one guy shutting another guy out or so and so letting his man light us up from outside.  That never EVER happens.  Defense is a team games when played correctly and the Celtics do it very, very well. If someone is open for an outside shot it's because someone else didn't keep his man from penetrating and then a third guy was late in rotation. 

  Rondo is better than average at keeping his man in front of him. Opposing PGs have a lower than average eFG%, take a smaller than average percentage of inside shots, get fewer than average assists and more than average turnovers. Also, there's more to defense than just keeping your man in front of you.
Our scheme is often to have him pressure guards tightly to disrupt the offense. Sometimes this leads to him giving up penetration though. Without hand checking it is hard to most PGs in front of you at all times.

Re: Statistics as they related to defense (split from game thread)
« Reply #20 on: February 02, 2009, 08:13:58 AM »

Offline BudweiserCeltic

  • Bill Sharman
  • *******************
  • Posts: 19003
  • Tommy Points: 1833
I'm sure no one looks forward to playing Rondo but the fact is that his teammates many times have to bail him out because of his gambling for steals.  He is not good at keeping his man in front of him and that many times results in someone elses man getting an open shot do to rotations.
This is what absolutely drives me crazy about anyone talking about one guy shutting another guy out or so and so letting his man light us up from outside.  That never EVER happens.  Defense is a team games when played correctly and the Celtics do it very, very well. If someone is open for an outside shot it's because someone else didn't keep his man from penetrating and then a third guy was late in rotation. 

  Rondo is better than average at keeping his man in front of him. Opposing PGs have a lower than average eFG%, take a smaller than average percentage of inside shots, get fewer than average assists and more than average turnovers. Also, there's more to defense than just keeping your man in front of you.
Our scheme is often to have him pressure guards tightly to disrupt the offense. Sometimes this leads to him giving up penetration though. Without hand checking it is hard to most PGs in front of you at all times.

This is clearly not a scheme problem. He has improved quite a bit this season though, but it's not something that is coming as a result of how he pressures the ball full court. Man to man defense is not his strength defensively by any stretch of the imagination.

Re: Statistics as they related to defense (split from game thread)
« Reply #21 on: February 02, 2009, 09:52:23 AM »

Offline Brendan

  • Jim Loscutoff
  • **
  • Posts: 2990
  • Tommy Points: 72
All things being equal, getting the steal, deflection, or block, is better than not. For every gambler like Rondo who gets a lot of steals, there is a bunch of guys gambling unsuccessfully.

I think PER contrasted with Adjusted +/- gives you a feel for intangibles.

Re: Statistics as they related to defense (split from game thread)
« Reply #22 on: February 02, 2009, 12:04:00 PM »

Offline drza44

  • Jaylen Brown
  • Posts: 749
  • Tommy Points: 187
Defensive stats are definitely a work in progress for basketball, but there are getting better and are definitely enough to get a general view of a player's defense.  As others have pointed out, looking only at the boxscore stats (i.e. rebounds, blocks, steals) can give a bit of a skewed ranking if they are the ONLY stats considered, since they don't look at all at how a player does at preventing the opponent from scoring (which, after all, is the primary goal of defense).  But those stats aren't useless, just limited.

Defensive Rating and Defensive Win Shares are two pretty popular defensive measures these days, and are kept up to date for every player at basketball-reference.com.  I blogged on DRTG recently as part of a series on advanced basketball stats (http://www.rotosynthesis.com/default.asp?display=569) and I'll be looking at Win Shares probably this week.  But generally speaking, Defensive Rating is the number of points a team gives up per 100 possessions with a certain player on the floor.  It's not a perfect measure because it is so teammate-influenced, but at the same time it tends to be useful.  This year Dwight Howard and KG are fighting for the title, last year KG won outright, and in recent years it has been names like Duncan, Ben Wallace, and Marcus Camby near the top...in other words, it passes the sniff test for a defensive measure.

Defensive Win Shares takes DRTG a step further...it attempts to calculate the defensive contribution (based partially on defensive rating) of the average NBA player in wins, then does a similar calculation for the player in question.  A defensive win share is basically the defensive contribution of an individual player divided by the contribution of the average player, giving a basketball defensive stat similar to VORP (value over replacement player) in baseball.  Again, you can go to basketball-reference.com to look at the numbers, but the players that measure out the best again pass the sniff test for best defenders in the league.

Adjusted +/- is another good way to look at defense, and there are several out there that focus on that side of the ball.  82games.com did one for last season that again passed the sniff test, with Garnett clearly on top of the league http://www.82games.com/ilardi2d.htm#table .

Finally, also on 82games.com, you can look at their calculations of opponent's production (kind of like a PER of each opponent) or the individual statistics (shooting percentage, rebound rate, etc.) of each players' opponents.  This is imperfect as well, because as others have pointed out it is hard to correct for switches and team defense specifically enough to do this perfectly.  Nevertheless, it gives another set of defensive numbers to corroborate with the DRTG, Defensive Win Shares, adjusted +/-, and box score info to give a clearer idea of how an individual plays defense.

Like I said, defensive stats in basketball are still a work in progress.  They aren't perfect.  But they are at least reasonable, and if you look at a variety of stats you can get a pretty solid feel for who the better and worst defenders in the league are.  And you get a MUCH clearer view than you would by simply looking at steals/blocks in box scores.

Re: Statistics as they related to defense (split from game thread)
« Reply #23 on: February 02, 2009, 12:54:50 PM »

Offline cordobes

  • NCE
  • Ray Allen
  • ***
  • Posts: 3556
  • Tommy Points: 576
  • Basketball is like chess, only without the dice
Defensive stats are definitely a work in progress for basketball, but there are getting better and are definitely enough to get a general view of a player's defense.  As others have pointed out, looking only at the boxscore stats (i.e. rebounds, blocks, steals) can give a bit of a skewed ranking if they are the ONLY stats considered, since they don't look at all at how a player does at preventing the opponent from scoring (which, after all, is the primary goal of defense).  But those stats aren't useless, just limited.

Defensive Rating and Defensive Win Shares are two pretty popular defensive measures these days, and are kept up to date for every player at basketball-reference.com.  I blogged on DRTG recently as part of a series on advanced basketball stats (http://www.rotosynthesis.com/default.asp?display=569) and I'll be looking at Win Shares probably this week.  But generally speaking, Defensive Rating is the number of points a team gives up per 100 possessions with a certain player on the floor.  It's not a perfect measure because it is so teammate-influenced, but at the same time it tends to be useful.  This year Dwight Howard and KG are fighting for the title, last year KG won outright, and in recent years it has been names like Duncan, Ben Wallace, and Marcus Camby near the top...in other words, it passes the sniff test for a defensive measure.

Defensive Win Shares takes DRTG a step further...it attempts to calculate the defensive contribution (based partially on defensive rating) of the average NBA player in wins, then does a similar calculation for the player in question.  A defensive win share is basically the defensive contribution of an individual player divided by the contribution of the average player, giving a basketball defensive stat similar to VORP (value over replacement player) in baseball.  Again, you can go to basketball-reference.com to look at the numbers, but the players that measure out the best again pass the sniff test for best defenders in the league.

Adjusted +/- is another good way to look at defense, and there are several out there that focus on that side of the ball.  82games.com did one for last season that again passed the sniff test, with Garnett clearly on top of the league http://www.82games.com/ilardi2d.htm#table .

Finally, also on 82games.com, you can look at their calculations of opponent's production (kind of like a PER of each opponent) or the individual statistics (shooting percentage, rebound rate, etc.) of each players' opponents.  This is imperfect as well, because as others have pointed out it is hard to correct for switches and team defense specifically enough to do this perfectly.  Nevertheless, it gives another set of defensive numbers to corroborate with the DRTG, Defensive Win Shares, adjusted +/-, and box score info to give a clearer idea of how an individual plays defense.

Like I said, defensive stats in basketball are still a work in progress.  They aren't perfect.  But they are at least reasonable, and if you look at a variety of stats you can get a pretty solid feel for who the better and worst defenders in the league are.  And you get a MUCH clearer view than you would by simply looking at steals/blocks in box scores.

Agreed. However, looking at the players playing is what gives you the best chance of assessing their defence. The problem with metrics like DRtg, as you explain, is that they're heavily team dependent (buy not only). The way the DRtg of a player can vary from season to season shows it can be an extremely misleading metric, even to evaluate who are the best defensive players in a team. Adjusted plus/minus has lots of noise and it's only significant with a big sample of several seasons (and players improve/regress)and for players who have changed teams (teammates) and played plenty of minutes. Rosenbaum published the same study a couple of years ago and some results were still suspect. Counterpart production is moderately useless at best and extremely misleading at worst.  I think that measuring statistically the defensive prowess of a player is a lost cause (and, to be honest, I don't think it's a very needed thing, as it's relatively easy to assess to defensive qualities of a player).

Re: Statistics as they related to defense (split from game thread)
« Reply #24 on: February 02, 2009, 02:04:30 PM »

Offline reggie35

  • Jrue Holiday
  • Posts: 361
  • Tommy Points: 51

Agreed. However, looking at the players playing is what gives you the best chance of assessing their defence. The problem with metrics like DRtg, as you explain, is that they're heavily team dependent (buy not only). The way the DRtg of a player can vary from season to season shows it can be an extremely misleading metric, even to evaluate who are the best defensive players in a team. Adjusted plus/minus has lots of noise and it's only significant with a big sample of several seasons (and players improve/regress)and for players who have changed teams (teammates) and played plenty of minutes. Rosenbaum published the same study a couple of years ago and some results were still suspect. Counterpart production is moderately useless at best and extremely misleading at worst.  I think that measuring statistically the defensive prowess of a player is a lost cause (and, to be honest, I don't think it's a very needed thing, as it's relatively easy to assess to defensive qualities of a player).

I agree that defensive qualities are easy to assess if you can watch a player for a couple of games. It would still be useful to have some way to know who out of the hundreds of options (overseas and here) is worth watching. Good stats would help with that.

Re: Statistics as they related to defense (split from game thread)
« Reply #25 on: February 02, 2009, 02:57:14 PM »

Offline drza44

  • Jaylen Brown
  • Posts: 749
  • Tommy Points: 187
Defensive stats are definitely a work in progress for basketball, but there are getting better and are definitely enough to get a general view of a player's defense.  As others have pointed out, looking only at the boxscore stats (i.e. rebounds, blocks, steals) can give a bit of a skewed ranking if they are the ONLY stats considered, since they don't look at all at how a player does at preventing the opponent from scoring (which, after all, is the primary goal of defense).  But those stats aren't useless, just limited.

Defensive Rating and Defensive Win Shares are two pretty popular defensive measures these days, and are kept up to date for every player at basketball-reference.com.  I blogged on DRTG recently as part of a series on advanced basketball stats (http://www.rotosynthesis.com/default.asp?display=569) and I'll be looking at Win Shares probably this week.  But generally speaking, Defensive Rating is the number of points a team gives up per 100 possessions with a certain player on the floor.  It's not a perfect measure because it is so teammate-influenced, but at the same time it tends to be useful.  This year Dwight Howard and KG are fighting for the title, last year KG won outright, and in recent years it has been names like Duncan, Ben Wallace, and Marcus Camby near the top...in other words, it passes the sniff test for a defensive measure.

Defensive Win Shares takes DRTG a step further...it attempts to calculate the defensive contribution (based partially on defensive rating) of the average NBA player in wins, then does a similar calculation for the player in question.  A defensive win share is basically the defensive contribution of an individual player divided by the contribution of the average player, giving a basketball defensive stat similar to VORP (value over replacement player) in baseball.  Again, you can go to basketball-reference.com to look at the numbers, but the players that measure out the best again pass the sniff test for best defenders in the league.

Adjusted +/- is another good way to look at defense, and there are several out there that focus on that side of the ball.  82games.com did one for last season that again passed the sniff test, with Garnett clearly on top of the league http://www.82games.com/ilardi2d.htm#table .

Finally, also on 82games.com, you can look at their calculations of opponent's production (kind of like a PER of each opponent) or the individual statistics (shooting percentage, rebound rate, etc.) of each players' opponents.  This is imperfect as well, because as others have pointed out it is hard to correct for switches and team defense specifically enough to do this perfectly.  Nevertheless, it gives another set of defensive numbers to corroborate with the DRTG, Defensive Win Shares, adjusted +/-, and box score info to give a clearer idea of how an individual plays defense.

Like I said, defensive stats in basketball are still a work in progress.  They aren't perfect.  But they are at least reasonable, and if you look at a variety of stats you can get a pretty solid feel for who the better and worst defenders in the league are.  And you get a MUCH clearer view than you would by simply looking at steals/blocks in box scores.

Agreed. However, looking at the players playing is what gives you the best chance of assessing their defence. The problem with metrics like DRtg, as you explain, is that they're heavily team dependent (buy not only). The way the DRtg of a player can vary from season to season shows it can be an extremely misleading metric, even to evaluate who are the best defensive players in a team. Adjusted plus/minus has lots of noise and it's only significant with a big sample of several seasons (and players improve/regress)and for players who have changed teams (teammates) and played plenty of minutes. Rosenbaum published the same study a couple of years ago and some results were still suspect. Counterpart production is moderately useless at best and extremely misleading at worst.  I think that measuring statistically the defensive prowess of a player is a lost cause (and, to be honest, I don't think it's a very needed thing, as it's relatively easy to assess to defensive qualities of a player).

Like I said, none of them are perfect.  And obviously personal observation is great as well.  But personal observation isn't necessarily a standalone either for many reasons.  For one, it is impractical for most people to watch every player play enough times to really judge them objectively.  Secondly, like in all sports, half of the fun is debating who is the best...and everyone's opinion is going to be different.

It's nice to be able to use stats in these situations, as they apply for every player in all 82 games (as opposed to the 2 times I might get to watch Andrei Kirilenko or Andris Biedrins, for example).  They also apply relatively even-handedly, where everyone is in the same boat.  Unlike opinions, where personal preferences color all analysis, at least individual stats tell their story and stick to them.  You have to know the strengths and weaknesses of each stat, and obviously you have to fit them into the umbrella with personal observations as well as other stats and other opinions in order for them to have any real meaning.  But I definitely think they're useful for many reasons, even if they are imperfect.

Re: Statistics as they related to defense (split from game thread)
« Reply #26 on: February 02, 2009, 03:07:13 PM »

Offline Brendan

  • Jim Loscutoff
  • **
  • Posts: 2990
  • Tommy Points: 72
Human observation is highly flawed too. You not only have to watch the players regularly - you could just catch a bad game. You have to understand what the team is trying to do and asking the player to do. You also have to understand the opponents, etc.

I liken it to baseball guys that think the player who "runs everything out" is the better player. Even when the stats say the lazy (Manny-type) guy who jogs is better. Typically the hustle guy is injured all the time. Maybe I should call it the Delonte effect.

As a human with biases, your eyes are not that reliable and its why developing and promoting good scouts is so important, you cannot trust the layman's eye.

Re: Statistics as they related to defense (split from game thread)
« Reply #27 on: February 02, 2009, 03:30:10 PM »

Offline Jaycelt

  • Jaylen Brown
  • Posts: 653
  • Tommy Points: 7
Quote
But generally speaking, Defensive Rating is the number of points a team gives up per 100 possessions with a certain player on the floor.  It's not a perfect measure because it is so teammate-influenced, but at the same time it tends to be useful.  This year Dwight Howard and KG are fighting for the title, last year KG won outright, and in recent years it has been names like Duncan, Ben Wallace, and Marcus Camby near the top...in other words, it passes the sniff test for a defensive measure.

Defensive Win Shares takes DRTG a step further...it attempts to calculate the defensive contribution (based partially on defensive rating) of the average NBA player in wins, then does a similar calculation for the player in question.  A defensive win share is basically the defensive contribution of an individual player divided by the contribution of the average player, giving a basketball defensive stat similar to VORP (value over replacement player) in baseball.  Again, you can go to basketball-reference.com to look at the numbers, but the players that measure out the best again pass the sniff test for best defenders in the league.

 I think it would be only fair to point out that unlike PP and RA, KG is never the only starter on the court. Neither are Rondo or Perkins but that's beside this particular point.
 That has to be a fairly big advantage when it comes to any defensive rating system as well as +/-.
Not trying to take anything from KG, especially defensivley, but it is true.

Re: Statistics as they related to defense (split from game thread)
« Reply #28 on: February 02, 2009, 05:27:10 PM »

Offline drza44

  • Jaylen Brown
  • Posts: 749
  • Tommy Points: 187
Quote
But generally speaking, Defensive Rating is the number of points a team gives up per 100 possessions with a certain player on the floor.  It's not a perfect measure because it is so teammate-influenced, but at the same time it tends to be useful.  This year Dwight Howard and KG are fighting for the title, last year KG won outright, and in recent years it has been names like Duncan, Ben Wallace, and Marcus Camby near the top...in other words, it passes the sniff test for a defensive measure.

Defensive Win Shares takes DRTG a step further...it attempts to calculate the defensive contribution (based partially on defensive rating) of the average NBA player in wins, then does a similar calculation for the player in question.  A defensive win share is basically the defensive contribution of an individual player divided by the contribution of the average player, giving a basketball defensive stat similar to VORP (value over replacement player) in baseball.  Again, you can go to basketball-reference.com to look at the numbers, but the players that measure out the best again pass the sniff test for best defenders in the league.

 I think it would be only fair to point out that unlike PP and RA, KG is never the only starter on the court. Neither are Rondo or Perkins but that's beside this particular point.
 That has to be a fairly big advantage when it comes to any defensive rating system as well as +/-.
Not trying to take anything from KG, especially defensivley, but it is true.

Is it really?  Because the times when there are no starters on the court for the Celtics generally correspond to when there are no opposing starters out there as well.  I dunno, that seems like a pretty arbitrary statement to make as fact without a bit more support.

Re: Statistics as they related to defense (split from game thread)
« Reply #29 on: February 03, 2009, 11:32:15 AM »

Offline Eja117

  • NCE
  • Bill Sharman
  • *******************
  • Posts: 19274
  • Tommy Points: 1254
So the top guys in steals this year are Chris Paul, Rondo, Kidd, Wade, Lebron, Chalmers, Gerald Wallace, Iguodala, Ariza, and Caron Butler.

The top blocks guys are D Howard, Camby, Turiaf, Brook Lopez, Bynum, Duncan, Daelembert, and Okafor

So I am to conclude these are the worst defenders in the league and the guys with zero blocks and steals are the best because they are the best positioners who probably always seal off the lane?

Maybe not, but I might say Kidd is a better defender than your average 34 year old NBA pg (or however old he is) and that Turiaf is a decent defender for a backup rotational guy

I might not say Ben Wallace is a good defender any more but when he had all the blocks he had I might.

A tougher question for me is maybe whether a Shawn Bradley or Manute Bol was a decent defender. I didn't see them play that much, especially Manite

I don't think anybody is arguing that steals and blocks are irrelevant to good defense.  However, look at your own list.  Do you consider Jason Kidd to be a good defender at this stage of his career?  Is Turiaf the third best defensive big man in the game?  Is Chris Andersen a top-four defender?  Is Al Jefferson a good defender, despite being #12 in blocks?  Of course not.  They have good statistics, but aren't top-tier defenders.

On the other hand, there are plenty of excellent defenders who do have good stats (Rondo, Kirilenko, Igoudala, etc.)  It's very difficult to quantify defense statistically, which is why observation is so important.