KeonClark wrote:
View Post
Announcement
Collapse
No announcement yet.
Measuring the D...
Collapse
X
-
-
Primer wrote: View PostFair enough, you go with whatever random shit Chandler says (a guy who has never played with Love), and I'll go with empirical stats over several seasons. Some of you are so entrenched in what you wanna believe that I realize there is no changing your mind, I just wanted to lay the objective stat based argument on the table for anyone reading that wants to see it.9 time first team all-RR, First Ballot Hall of Forum
Comment
-
KeonClark wrote: View PostLOL! Chandler PLAYS. Matched up against love. Many, many times. Folks this isn't what happens when analytics go too far.
Comment
-
Primer wrote: View PostWhy would Love be defending Chandler? Love isn't a C. What's many many times? 2 times per year since they spent the majority of their careers in the opposite conference.9 time first team all-RR, First Ballot Hall of Forum
Comment
-
Primer wrote: View PostDraymond Green is a 99, Tony Parker is a 98, so basically the same. San Antonio plays excellent defense, so not sure how you could say Parker is a bad defender (I'm guessing you aren't). Excellent team defense can absolutely move a player up the metric, but that follows logic doesn't it (if a player plays heavy minutes on a team with excellent defense you can logically say he's playing good defense, the proof is in the pudding)?
1. Player DRTG is only good within the team itself. It's not a good metric at all to compare with other players on other teams.
2. Parker is not a good defender. I really hope you don't think that Draymond Green and Tony Parker are on equal footing defensively.
3. As seen with the low minute filter I placed, you don't even need to play heavy minutes to gain a boost in the rating.
Barolt wrote: View PostThe Spurs might have the best defense ever. So it's not surprising that their players have good Drtg.
Primer wrote: View Post100 minutes over half a season is way way way too small a sample size. Try 800 or 900 minutes. 800 minutes is only 19 minutes per game. Anyone who plays less than that isn't playing enough for you to get a reliable number, and that goes for all advanced stats, you can't look at guys who barely play.
As noted by Barolt, the Spurs have one of the best defenses in the history of the NBA right now, so to say anyone on that team is playing bad defense is just wrong.
Comment
-
-
tDotted wrote: View PostCouple of things..
1. Player DRTG is only good within the team itself. It's not a good metric at all to compare with other players on other teams.
2. Parker is not a good defender. I really hope you don't think that Draymond Green and Tony Parker are on equal footing defensively.
3. As seen with the low minute filter I placed, you don't even need to play heavy minutes to gain a boost in the rating.
So would plugging any one of those guys on another team result in the same individual rating right? No.. It would drastically change. That's the point I'm making.. you can't use something that relies heavily on team defense to evaluate individuals.
Again, is their whole roster suddenly top 25 defensive talents?
Comment
-
tDotted wrote: View PostAnd just like that, I take back what I said.
Comment
-
I would love to see the look on David Blatts face, or any other coach or player for that matter, if a fan walked up to them and said "kevin love is a great defender has been for years, here's his drtg and his dprm and his r2d2!!"9 time first team all-RR, First Ballot Hall of Forum
Comment
-
In general, I take a multi-stat approach to defence and even then you have to be realistic about how limited even the combination of those stats are.
First, individual DRTG. This absolutely does have to be judged relative to teammates. It's hard though, you can't just subtract the team rating because if your team is made up entirely (or mostly) of above average defenders then some of your defenders will end up with an unfair negative rating. Also, this number applies almost exclusively to individual defence - playing your man one on one. Team defence situations are complex and it is very hard to apply ownership over success or failure in these cases statistically. The individual DRTG uses a "stop rate" to determine how far off your team rating you are so it is useful enough, though help from your teammates is always a factor. And individual defence is just one small aspect obviously.
Team impact is a huge one, but is even more muddied. Simple on-off splits are pretty great, but are VERY teammate and role dependent, especially depending on rotations. For example, last season Tyler Hansbrough had a magnificent on-off split, but he ONLY played against bench units and ONLY played as a part of the Raptors' above average bench. So of course RPM tries to filter that out, but looking at the year to year variations in performance... It doesn't seem like the least bit reliable, not even as reliable as previous versions of the stat (APM, RAPM) that were either less convoluted (APM is a simpler, more direct regression) or less prone to fluctuation (as RAPM is a two year regression that flattens out the results). Still, it might be another useful tool, but again certainly not one to use alone.
Defensive FG% is a decent one, but again only for individual defence, and there's context aplenty - for example, a player who fouls like crazy can have a very low DFG%, but be a terrible defender, giving up free points all over the place. Or a player who gambles a lot and is a ball hawk who generates a ton of steals and other turnovers but gives up the occasional layup will see his DFG% hurt, even though he may be well above average.
Same goes for more specific defensive play types, like ISO defence or post defence, which are fairly isolated and can probably be taken at face value. Of course, with all individual numbers, like individual DRTG, ISO defence or post defence, the quality of player being defended must be taken into account. DeMar will almost always take the lesser offensive wing player, so his numbers there will be inflated. Same goes for Love, who will likely take the lesser big man threat in any matchup (although with big men it is often more about size than skill, so less so than DeMar).
For big men there is rim protection, but that's extremely volatile - it seems like the sample size of shots faced in one season is not really enough to average out the opposition and the opportunities they are taking advantage of to get to the rim in the first place, which affects timing of help. Also falls victim to the same trap as defensive FG%, as fouls called there are not tracked, meaning a guy could cause a low FG% at the rim due to very physical play that more often than not gives his opponents a trip to the line for two freebies.
Things like defensive WS are basically useless (for describing defence, it has some value in terms of rebounding production), as it's essentially a box score stat counter, which we all know is not a great judge of defence, though there is a little more to it than that - but in the end the largest contributor to DWS is just the individual DRTG anyway, which as discussed has it's raw value heavily impacted by the team structure.
There's not a lot in way of catch all defensive stats beyond that. You can do a deep dive WOWY analysis for every player on a team and see if a player helps or hurts most of his teammates. You can do a lineup comparison of the heaviest used lineups and see if subbing out just that one player for another helps or hurts. These are all things that take a huge amount of work though, so are not helpful for judging the league in general.
Obviously you also have to take into account rebounding, because that is a contributing factor as well (a good defensive rebounder is valuable). But it is not as valuable as once believed - many defensive rebounds would end up falling into a teammate's hands anyway. But still, there is some value to be considered there.
All in all, there simply is no really solid option for judging defensive players. Usually I'll start with a glance at DRPM and DRTG, then their on-off DRTG split, and defensive rebounding rate. If they are looking fairly average at that point they probably are. Anything above or below average by a significant amount I like to dig deeper, look at some of their historical DAPM, DRAPM, then look at some lineups and the equivalents with their substitutes. Sometimes I'll glance at defensive FG%, but not often (way too many really good defenders show badly in that stat), and mostly out of curiosity.
It's the one area where I'm really not confident in relying on statistics when there is limited viewing available. Offensively you can get away with judging guys from their stats most of the time (although obviously it is best to deal with all the info available, including watching the game). Defensively? There's numbers aplenty, but I have basically zero faith in their veracity or relevance. Really the best we can do is describe whether the player is really hurting their team with on-off splits and WOWY analysis.
Comment
-
Measuring the D...
There you go guys, I moved all 41 posts related to your discussion on different defensive metrics to a brand new thread.
The title was inspired from a comment in last night's game threadHeir, Prince of Cambridge
If you see KeonClark in the wasteland, please share your food and water with him.
Comment
-
The basic problem with measuring defense is that an outsider doesnt really know what the scheme is or what a guy is being asked to do until he's watched dozens and dozens of games. SVG made this point at one of the analytics conferences. He went on to say that, for example, JJ Reddick was a good defensive player for him cause he always did what he was asked, never missed rotations, etc. That helped stabilize his defense cause it held things together even if JJ wasn't a great one-on-one defender.
The current defensive models aren't very good and a lot of the tracking stats are immaterial or even irrelevant cause they don't have any of the requisite context.
Comment
-
slaw wrote: View PostThe basic problem with measuring defense is that an outsider doesnt really know what the scheme is or what a guy is being asked to do until he's watched dozens and dozens of games. SVG made this point at one of the analytics conferences. He went on to say that, for example, JJ Reddick was a good defensive player for him cause he always did what he was asked, never missed rotations, etc. That helped stabilize his defense cause it held things together even if JJ wasn't a great one-on-one defender.
The current defensive models aren't very good and a lot of the tracking stats are immaterial or even irrelevant cause they don't have any of the requisite context.Heir, Prince of Cambridge
If you see KeonClark in the wasteland, please share your food and water with him.
Comment
-
As a defensive question, given Casey's love for using JV/Biz offense/defense switches late game, why don't we do the same with Ross/DeMar?
While DeMar is an adequate defender, Ross can be a true lockdown guard. Given our difficulties defending the 3pt line this season, that would appear to have value in late game defensive situations, right?twitter.com/anthonysmdoyle
Comment
Comment