What is the "best" defensive statistic?

[nonemus]

There are many stats that measure offensive efficiency, but are there any stats that measure defensive tenacity?

I am not talking about rpg/bpg/spg. The only advanced stats I know which seem to fit the bill are the following:

eFGa (Allowed EFG)
dRTG (Points given up per 100 possesions)

Are there any other ones?

[Jimmy76]

Some people like defensive winshare which from I can gather is great at estimating the amount of wins a team got with their defense but questionable in terms of assigning individual responsibility

I'm a big adj +/- and on/off fan but neither should be used in a vacuum

[nonemus]

What is the drawback of eFGa?

[Jimmy76]

What is the drawback of eFGa?

In what context?

In terms of measuring a teams a ability to hold their opponents to low shooting percentages no

[nonemus]

What is the drawback of eFGa?

In what context?

In terms of measuring a teams a ability to hold their opponents to low shooting percentages no

In the context of measuring a player's defensive effectiveness.

If the eFGa of a player is low, doesnt it mean that they are a good defender? Or is it not neccesairly the case?

[Chicago76]

If the eFGa of a player is low, doesnt it mean that they are a good defender? Or is it not neccesairly the case?

It's not necessarily the case. It depends upon who the player is matching up with defensively. Imagine a team with either a PF or C who is a good defender while the other is poor. The good defender matches up against the oppositions top interior talent while the other guy guards the big man who tends to only be a rebounder/defender type. The good defender is being penalized for being good.

Also: hard to indentify if the shot given is the defender's true man. Maybe he needed to switch off and someone else gave up the shot.

Also: eFG is not going to tell us how good a player is at eliminating a shot attempt from the guy he is guarding the first place.

These things are better covered by looking at +/- stats of the player being guarded by someone over at 82games.com

[Nivek]

The question with eFGa is how the data is being gathered and calculated. If it's counterpart data like the stuff at 82games, it has limited value. That's because that data doesn't account for the realities of help defense. If it's hand-tracked data, it's likely better -- although that depends on how send the collection method is, and the skill and knowledge of the collector.

Drtg is okay, but it's only an estimate. It uses steals, rebounds and blocks to estimate stops. It also estimates defensive possessions. But, positional defenders who don't generate steals, blocks or rebounds aren't rated accurately because the estimation doesn't pick up how they defend. The system Dean Oliver created does only a so-so job of estimating defensive possessions -- again not fully reflecting help D or defensive activities other than blocks, steals and defensive rebounds.

Defensive Win Shares are calculated using drtg.

Plus/minus stuff is okay, if you know what you're looking at. I actually don't trust the adjusted +/- data because there are so many variables and I don't think the APMers fully understand them. I do like using the on/off data at 82games to get a sense. But it needs to be combined with other information.

To me, the best information is stuff that comes from hand tracking. But that's a major time suck.

[Shinyhubcaps]

One that is used is Defensive Rating... maybe this is the same as one already mentioned, but as far as validating what you think to be true, this one is hit-or-miss.

Here are the 2009-10 leaders:

Defensive Rating
1. Dwight Howard-ORL 95.4
2. Andrew Bogut-MIL 98.1
3. Gerald Wallace-CHA 99.7
4. Lamar Odom-LAL 100.4
5. Rasheed Wallace-BOS 100.5
6. Marcus Camby-TOT 100.8
7. Tim Duncan-SAS 100.8
8. Kevin Garnett-BOS 101.0
9. Joakim Noah-CHI 101.0
10. Josh Smith-ATL 101.4
11. Kendrick Perkins-BOS 101.5
12. LeBron James-CLE 101.7
13. Anderson Varejao-CLE 101.8
14. Ersan Ilyasova-MIL 101.8
15. Pau Gasol-LAL 101.8
16. Carlos Boozer-UTA 101.8
17. Andrew Bynum-LAL 102.1
18. Carlos Delfino-MIL 102.6
19. Dwyane Wade-MIA 102.6
20. Michael Beasley-MIA 102.6

It starts out innocently enough, but Ersan Ilyasova? Michael Beasley? I don't know about that...

[Jimmy76]

Drtg is heavily based on team drtg so players on good defensive teams can end up with high rankings even if they aren't great defenders ( though it's usually sign that they aren't massive defensive liabilities, in Beasleys case the fact the HEAT played such great defense without a great anchor or lockdown wings probably means Beasleys weak defense is overstated)

Also advanced stats don't try to mimic common perception. Half the point is to pick up data no one else is noticing; if they just replicated common perception they'd be useless not to mention probably tampered with to achieve a particular result.

[tsherkin]

I don't like using just one.

I do think a Four Factors analysis is a really good place to start, though.

[Ripp]

One that is used is Defensive Rating... maybe this is the same as one already mentioned, but as far as validating what you think to be true, this one is hit-or-miss.

Here are the 2009-10 leaders:

Defensive Rating
1. Dwight Howard-ORL 95.4
2. Andrew Bogut-MIL 98.1
3. Gerald Wallace-CHA 99.7
4. Lamar Odom-LAL 100.4
5. Rasheed Wallace-BOS 100.5
6. Marcus Camby-TOT 100.8
7. Tim Duncan-SAS 100.8
8. Kevin Garnett-BOS 101.0
9. Joakim Noah-CHI 101.0
10. Josh Smith-ATL 101.4
11. Kendrick Perkins-BOS 101.5
12. LeBron James-CLE 101.7
13. Anderson Varejao-CLE 101.8
14. Ersan Ilyasova-MIL 101.8
15. Pau Gasol-LAL 101.8
16. Carlos Boozer-UTA 101.8
17. Andrew Bynum-LAL 102.1
18. Carlos Delfino-MIL 102.6
19. Dwyane Wade-MIA 102.6
20. Michael Beasley-MIA 102.6

It starts out innocently enough, but Ersan Ilyasova? Michael Beasley? I don't know about that...

This is individual Drtg from Dean Oliver's book, right? I've looked at it a lot, and am not sure it is a good way to rank individuals defensively.

[Nivek]

Dang, are folks reading the thread or just posting in it? As I said earlier in this thread --

Drtg is okay, but it's only an estimate. It uses steals, rebounds and blocks to estimate stops. It also estimates defensive possessions. But, positional defenders who don't generate steals, blocks or rebounds aren't rated accurately because the estimation doesn't pick up how they defend. The system Dean Oliver created does only a so-so job of estimating defensive possessions -- again not fully reflecting help D or defensive activities other than blocks, steals and defensive rebounds.

[Dr Positivity]

I think defensive adjusted +/- is the best one. Certainly a lot more trustworthy than its offensive counterpart

[Ripp]

Err, what? Your post isn't the final word on Drtg...there are larger methodological issues with it. For example, it isn't clear that it places enough weight on individual performance versus team. His formula adds team Drtg to an individual term....so if the individual term is sufficiently small, you get Michael Beasleys appearing on the list.

Basically, there are tons of ways to choose MyRating = TeamRating + F, where F is some function of individual stuff, and it isn't clear that Oliver's choice passes certain smell tests.

This is why you see things like Hedo Turkoglu going from 104 to 113 in one year, for example.

[nonemus]

So, do different sites use different formulas for dRTG or what?

Ex, Dwight Howard has a dRTG (points allowed per 100 possession) of 103 at 82games.com and 95 at b-r. Is there no universal formula or what?

[Ripp]

b-r.com uses Dean Oliver's "individual" Drtg, which is a formula defined in his book.

82 calculates what your team's Drtg was while you were on the floor (http://www.82games.com/0910/09ORL12.HTM, 103 for Dwight).

basketballvalue.com looks at your team's Drtg when you are on the floor, off the floor and calculates the difference (http://basketballvalue.com/teamplayers. ... 0&team=ORL, 102.68, 106.05 and -3.37, respectively).

Notice that these are two very different things...Oliver tries to rank your own individual defensive impact indirectly through a formula of some sort, while on/off numbers like from 82games.com and bballvalue.com just measure what the team's defense was when you were on and off.

So in a perfect world, Oliver's formula accurately gauges who is a good defender and who is not.

And in a perfect world, on/off sort of stuff will accurately tell you that player X is hurting/helping the defense by Y points.

[Shinyhubcaps]

Dang, are folks reading the thread or just posting in it? As I said earlier in this thread --

We were actually posting at the same time. I just had all my tabs open and never refreshed the page; no one had previously mentioned Defensive Rating, so I threw it up there, while you were posting concurrently.

Also advanced stats don't try to mimic common perception. Half the point is to pick up data no one else is noticing; if they just replicated common perception they'd be useless not to mention probably tampered with to achieve a particular result.

Obviously that's why I found it a bit fishy but not outright wrong. The formula has value, I'm sure. It's just a check, though. There are a lot of guys, maybe like hustle guys or whatever, who play well in rotations and keep down points, then are bound to stumble into steals and rebounds, but one-on-one or with a different group of guys, they wouldn't do so well. When they show up in Def. Rating, it's borderline. Still, does that accurately portray the top 20 defensive players in the NBA? Probably not.

If guys showed up on Def. Rating, Defensive Win Shares, Adj. +/-, and some other defensive stats, then they're probably good defenders. Each formula has a little value, but if you combine them all equally, you'll probably just get the consensus best defenders.

[Barstool Blues]

When you take into account Drating you have to take into account the players defensive rating to the league and THEIR TEAM. The only reason Beasley is up there is because of how slow the Heat played, but compared to the overall team his defensive rating isn't special at all for a SF/PF. If you play with bad defensive players though you can still be a dominate defensive player, but it won't show up that way unless you look at the team too. Some of Hakeems seasons are like that. His defensive rating isn't as low because the team wasn't that great, but his drating to his team is -8 which is pretty much insane. He's the only player to be as much as 10 points under his team in drating...

[Nivek]

When you take into account Drating you have to take into account the players defensive rating to the league and THEIR TEAM. The only reason Beasley is up there is because of how slow the Heat played, but compared to the overall team his defensive rating isn't special at all for a SF/PF.

Not sure what you're getting at here. Drtg is a possession-based stat. A fast-paced team can have a good defensive rating, just as a slow-paced team can have a bad one. Last season, for example, the league's slowest-paced team was Portland and they were only about average defensively. Indiana, Denver, Utah, Chicago, Oklahoma City, and the Lakers were faster than average to average (the Lakers were right at league average pace), but better than average defensively.

If you play with bad defensive players though you can still be a dominate defensive player, but it won't show up that way unless you look at the team too. Some of Hakeems seasons are like that. His defensive rating isn't as low because the team wasn't that great, but his drating to his team is -8 which is pretty much insane. He's the only player to be as much as 10 points under his team in drating...

This is a good point. Team defense is a big factor in drtg. I understand why DeanO did it that way, but it's still a weakness when it comes to measuring individual defensive contributions.