Is PDSS a good way to analyze the Raptors defense?

[BorisDK1]

yeah, it's quite funny actually. On top of that Boris has failed to answer a very important question that SS keeps asking. Yes, I read the above, and calling thousands of mins of data "noise" seems like a cop out answer here. Seems like everyone is pushing their own agenda and won't concede anything.

And lastly, if PDSS says Bargs is an above average defender or a top 3 defender amongst rotation guys on last seasons team then anyone with common sense should be questioning the data / formula used.

Reignman, I think you are engaging in misrepresentation about half as bad as Ripp. I have engaged what supersub has said, I've expressed my concerns that it is indirect analysis, that there is considerable noise (which supersub agreed with, but then said the noise somehow goes away when you go further back into the past), that despite admitting that on/off court data does not isolate one player (or even pass judgment on the quality of that player's defense) supersub has tried claiming that that actually happens just by including older data on teams past (which I believe to be untenable).

That aside, I am quite willing to engage that data. Lineups with Bargnani didn't do particularly well on the defensive end than other ones: I think we all get that. The point of PDSS is to further narrow down why that is the case: who gave up how much of what? is what it's trying to show. I am fully willing to consider and evaluate on/off court and APM data; by contrast, supersub and Ripp are 100% unwilling to even consider PDSS as having anything meaningful to say. So my question, is really - who's interacting less with their opponent? IMO, I think it can't be said that it's me...

At the end of the day, I keep coming back to the fact that we need to synthesize some data and at the end of the day, take the questions that the stats bring up (because stats provoke questions, not answers - you have to find the answers on the basketball court) and do business there. It seems supersub and ripp in this conversation up to this point aren't really doing that. And I'd love to get to there eventually, but it might take some time and some doing.

[BorisDK1]

Now, you are accusing me of forcing my stats on you, but you are doing the same with PDSS, no matter how solid you think the stats are. Synergy Sports has compiled a huge database supported by thousands of dollars paid for by the same NBA teams that you are analyzing. I'm not trying to denigrate your work, but given a choice between your analysis and that of Synergy, you know who I'll take 100 time out of 100, especially that you have admitted to a bias against Calderon, the same Calderon you have tried to pin the blame on.

I just Synergy again, and Bargnani seems to do 1 thing really well, P&R Roll Man plays (only 71 plays though), the rest? He's ok on post-ups, but wretched on everything else. Overall, he's ranked 259th among NBA players.

I keep giving you stats other than PDSS, and you keep throwing them back in my face.

And dagger, that's really childish.

LOL - supersub, is it possible that you dismiss me because I don't have a website, a budget and simply because you want to justify to yourself the money you spent on Synergy Sports when you could have had better data for free? Despite your obvious acumen and passion and ability to handle numerous metrics, you for whatever reason are very loathe to accept this. And no, I'm not trying to force PDSS on you or anybody else: I'm merely pointing out that PDSS gives us far more direct data and it should be interpreted and should force questions. I'm not seeing any questioning or any willingness to engage these numbers at all. Is that because you are such a firm believer in the methods you have become comfortable with using? I'd suggest that it's entirely the case, and at some point you'd be best advised to broaden rather than narrow your horizons.

As far as the bias against Calderon goes, now you're lurching into argumentum ad hominem: "you can't be trusted because you admit you don't like Calderon". Yeah, I made a humourous comment about Jose. I've made many in my life, mostly in retaliation for the ridiculousness I've had to endure at raptorspace.com at the hands of various Spaniards insisting Jose is simply a dominant point guard and anybody who dares usurp any of his minutes should be drawn and quartered in Nathan Phillips Square. I've also been critical of Bargnani - odd you don't mention that in your accusation. But that doesn't change the fact that I have invested too much of my time and passion and energy and knowledge into doing with the greatest possible integrity because, at the end of the day, I want a product I can be proud of and I believe I've produced something I can be proud of. You do the same thing with your spreadsheets, and would no doubt chafe if I accused you of hopeless bias in your recording of things. And finally, since this accusation is entirely fallacious as far as argument goes, I'll just point out that for the sake of argument I could have 16 contracts out on Jose's life as we speak and spent every Saturday night of the past 4 years slashing his tires in his parking garage every Saturday and I still might have recorded everything properly. What I think about Jose has nothing to do with how accurate I may have been, and I'm chuckling at the kind of mindset that thinks it might. And I also think you'd blow your stack if I accused you of the same thing with Bargnani.

As far as my ability to interpret the game, I don't have anything to prove. I've spent hours in seminars, reading, discussion with other coaches and officials about things and I do coach myself (at a very small college). I was coaching one of my high school's teams while I was still in high school. If you want to think that some people you don't know who in all likelihood do not have the technical or theoretical basis in the game that I do and who only have bothered to track a portion of the data that I have somehow help you out more than I do, go ahead. If I suggested that you wouldn't know the difference between a UCLA cut and a flex cut if your life depended on it, does it make my point any stronger? No. (For the record, I'm not suggesting that. I use that only as a for-instance of a ridiculous argument I'm not going to make - yet for some reason, you feel comfortable doing.)

And I'm not "throwing those stats back in your face" any more than you're dismissing data which happens to contradict your opinions out-of-hand. Let's be honest about that.

[Ripp]

BorisDK1:

1) supersub15 included the data of previous years to show that the on/off results for this year also held in previous years. In other words, he is saying that there is some consistency in the trends, and not a crazy bias that is somehow making certain players look worse than they are.
2) I'm willing to consider PDSS, but it needs to be shown to be correct in some way. Remember the RippRatings I brought up earlier? I ran them on my computer. Here are both the PDSS ratings and the RippRatings stacked against each other:

Code: Select all

   PDSS Drat   Ripp Drat
Mensah-Bonsu   107.1000     113.43
Evans   107.9000      112.66
Banks   108.0000      112.28
Johnson   109.5000      113.18
Bargnani   110.0000      112.56
Bosh   110.0000      113.48
Belinelli   111.4000      112.28
Jack   111.7000      113.15
DeRozan   111.9000      112.76
Turkoglu   112.3000      113.48
Weems   112.3000      112.32
Wright   112.8000      113.09
O'Bryant   113.5000      112.18
Nesterovic   113.6000      113.89
Calderon   118.3000      113.37
“Team”   115.2000      111.82


Now, your PDSS ratings:

[1] are just as useless in predicting lineup performance as mine.
[2] Your ratings are just as useless in predicting 4th quarter performance as mine.
[3] However, mine will predict the final season team defensive rating much more precisely than yours.

So my rating system fails on just the same tests that yours does (lineups, 4th quarter), but does better on this third "test" (total season Drtg, i.e., regurgitating the data I provided to both algorithms.)
So if we want to use your rating system as a basis for evaluating players, then mine is equally valid. But like I said, my rating system is a complete crackpot. In fact, do you know how I generated it? I typed the following command into a program on my computer:

Code: Select all

octave:7> 113.2+.5*randn(16,1)


In other words, I took the team Drtg for the season, and then added random noise to it! So in other words, my RippRatings don't give ANY insight into what is going on, who is good defensively and who isn't. There is absolutely no insight it offers...they just take the team Drtg I was given, and then add literally random noise to it, and then spits that out as the answers.

If your rating system is inferior to mine, and mine is literally spewing out random garbage, why should I listen to yours at all? There has to be some sort of test that your system passes so that we can distinguish it from a algorithm that produces just random noise.

Again, the tracking of additional stats you've done is great, wonderful stuff. The Stop% stuff might also be decent too. However, going from there to your DRat seems to be flawed, if I can come up with a "RippRating" system that is clearly useless but just as good in explaining what actually happened in the real world.

[BorisDK1]

You don't understand statistics, is the problem....one of the major themes of statistics is that with enough noisy samples, the underlying behavior of the generative process can be understood to exceedingly accurate accuracy by the appropriate algorithm/estimator/technique.

Let me illustrate with a small example. Suppose someone is flipping a coin. One side of the coin is marked 0, the other is marked 1. The coin has some probability "p" of coming up 1. Suppose he flips the coin N times. Assuming independence between coin flips, then you can estimate the number "p" to within accuracy on the order of 1/sqrt{N} (using the obvious estimator that just sets the estimate to the average number of 1s that were in the sequence of flips.)
So in other words, if he flips the coin 1000 times, your estimate of "p" is good to within a tolerance on the order of roughly .0316. If he flips it 1 million times, your estimate of "p" is good to within something on the order of 10^{-3}.

Look guys...when you use your cell phone and talk to your buddy, your voice gets converted to bits, and those are sent through the air. Well, the air is NOISY and random, and corrupts those bits sent! So if people (in this case, those good electrical engineers, statisticians and mathematicians who developed communication theory and information theory) did not know how to extract information from noisy information, lots of cool things you and I take for granted would not work.

So the noise argument is not going to fly, given that SS15 has 9000+ minutes worth of data. A bias argument might (bias is a bit more insidious, and I can illustrate this with a coin-flipping example too), but this certainly will not.

EDIT: And the independence assumption I made above can be dramatically weakened, so long as the correlation between coin flips isn't huge. So don't think it requires an assumption that doesn't show up in the real world.

Look, when it comes right down to it I can be as much of a pedant as the next guy, but you are taking pedantry to new and unheard-of heights.

Now, explain to me this: why is it, when Andrea Bargnani is facing a possession defensively, 55.2% of the coin flips (~1030 samples) comes up "heads" (stops) as opposed to "tails" (scores)? Andrea's doing something to cause that. Maybe you, like supersub, have to pay $30/year to me to be able to trust my data: if that's the case, I'll certainly give you my mailing address and I'll be happy to bask in my new-found "trusted" status.

At this point, it might behoove you to remind yourself that basketball is not an engineering application: it's a basketball operation solely, and please forgive the tautology. You are attempting to divorce some stats entirely from the actual game of basketball. And at that point, you're missing a lot. I also notice you're throwing out a whole lot of accusations about my ignorance of higher maths (information which I volunteered), yet I think pigs will fly the day you'd actually try to take on Dr. Oliver himself about this method because you know that the outcome wouldn't be good for you. (Call that an appeal to authority if you will, but it's an authority within the relevant field.)

I understand what the stats that have been quoted tell me: that the team has done better with Bargnani than without him. I have no doubt that that is the case: that's not the "noise" I'm concerned with. The "noise" comes in from making an indirect observation about performance ("the team with Bargnani vs. the team without Bargnani") and trying to take that to an unwarranted place, that it speaks to Bargnani's individual performance to in any way that we can identify. The reason PDSS is here is to develop that ERA in basketball, so we know exactly what as individuals they were / were not responsible for how and how that fit in with the team.

At this point, you have stepped over that line and become one of those people who try to do basketball analysis solely by stats, and I'm sorry - that's not how basketball analysis is done and you don't have much to say that means much. And what you have said, in the main, about PDSS - including your original post - has been entirely baseless as well.

[Fenris-77]

... Will you still trust this algorithm that someone is trying to sell you? It is telling you that your team will be better defensively in the 4th than earlier, and this is in direct conflict with what actually happened, and you measured and saw in games. Personally, I would not trust such an algorithm...it would be evidence that there is something very wrong with the person's methodology.[edited for length and a certain piquante something]

Wow. So now the primary tool I should be using to gauge advanced basketball metrics is whether or not those metrics agree with what Ripp saw on the court? Pardon me while I snicker into my hand.

The difference between PDSS and your recollection is that PDSS tracks every single posession, while you (and everyone else watching any particlular game or games) only remember those plays that, for whatever reason, stand out in your memory. That's not to suggest that you know nothing about basketball mind you, the selective funtion of memory is common to everyone, which is why coaches use game tape for analysis, and not just their recollections. If any kind of bias exists here I submit that it's far more likely to consist of you watching games and looking for what you want to find, rather than some sort of algorhythmic conspiracy.

I also find it intensely humourous that you seem to be flailing around in vain looking for a strawman argument that will let you defame DRat as an idea. DRAt and ORat are commonly used advanced metrics, and both enjoy widespread approval and attention from the basketball stats community. PDSS is just like the common DRat, except that tracks everything that it's originator would have liked to include, rather than just what the NBA box score provides. Where does that leave you? Oh yes, not liking the stat because you don't agree with the results. Fortunately, and here I'll borrow a phrase from you, the PDSS results are based on 'what actually happened', not on your memory or interpetation of what happened. Given the choice between those two options, I think I'll stick with with what actually happened.

[Ripp]

... Will you still trust this algorithm that someone is trying to sell you? It is telling you that your team will be better defensively in the 4th than earlier, and this is in direct conflict with what actually happened, and you measured and saw in games. Personally, I would not trust such an algorithm...it would be evidence that there is something very wrong with the person's methodology.[edited for length and a certain piquante something]

Wow. So now the primary tool I should be using to gauge advanced basketball metrics is whether or not those metrics agree with what Ripp saw on the court? Pardon me while I snicker into my hand.

I'm not talking about recollection, dude. Look, we can measure how many points were scored in the 4th quarter exactly, right? We can measure how many possessions were used in the 4th quarter exactly right? And same for quarters 1 through 3. You agree here, right, that we can measure these things exactly? The scorekeepers measure exactly how many points were scored in the 4th quarter. From reading a play-by-play, you can count exactly how many possessions happened in the 4th quarter.

So if I have an "advanced basketball metric" that tells me that 80 points per 100 possessions in the 4th quarter, but instead they give up 150 points per 100 possessions, which should I trust? That advanced basketball metric, or the actual number of points per possession that were actually scored over in the 4th? Will I trust the metric, or what actually happened on the floor?

[BorisDK1]

BorisDK1:

1) supersub15 included the data of previous years to show that the on/off results for this year also held in previous years. In other words, he is saying that there is some consistency in the trends, and not a crazy bias that is somehow making certain players look worse than they are.

What part of "I understand the team has been better off without Bargnani than with him, but you cannot use that to be somehow authoritatively indicative of Bargnani's individual performance" do you not understand?

The team has changed since Bargnani came into the league (only two players, as of today, were on this team when he was a rookie). The league has changed. Andrea has changed. Everything has changed, and despite those non-constants it's pretended as though nothing else changed.

No basketball analysis has been done here at all. Why not?

2) I'm willing to consider PDSS, but it needs to be shown to be correct in some way. Remember the RippRatings I brought up earlier? I ran them on my computer. Here are both the PDSS ratings and the RippRatings stacked against each other:

Code: Select all

   PDSS Drat   Ripp Drat
Mensah-Bonsu   107.1000     113.43
Evans   107.9000      112.66
Banks   108.0000      112.28
Johnson   109.5000      113.18
Bargnani   110.0000      112.56
Bosh   110.0000      113.48
Belinelli   111.4000      112.28
Jack   111.7000      113.15
DeRozan   111.9000      112.76
Turkoglu   112.3000      113.48
Weems   112.3000      112.32
Wright   112.8000      113.09
O'Bryant   113.5000      112.18
Nesterovic   113.6000      113.89
Calderon   118.3000      113.37
“Team”   115.2000      111.82


Now, your PDSS ratings:

[1] are just as useless in predicting lineup performance as mine.
[2] Your ratings are just as useless in predicting 4th quarter performance as mine.
[3] However, mine will predict the final season team defensive rating much more precisely than yours.

I don't think they are useless in predicting lineup performance. I think as a coach if I had this data at my command, I'd do a pretty good job putting together the better lineups. In fact, on my team it is the tool we use to analyze defensive performance and since we want to promote a team culture that the defensive end of the floor is where we build our identity, we made a commitment to use it liberally.

Again, at some point you have to actually do business with basketball theory and the game. You seem completely unwilling to do so. Why?

So my rating system fails on just the same tests that yours does (lineups, 4th quarter), but does better on this third "test" (total season Drtg, i.e., regurgitating the data I provided to both algorithms.)
So if we want to use your rating system as a basis for evaluating players, then mine is equally valid. But like I said, my rating system is a complete crackpot. In fact, do you know how I generated it? I typed the following command into a program on my computer:

Code: Select all

octave:7> 113.2+.5*randn(16,1)


In other words, I took the team Drtg for the season, and then added random noise to it! So in other words, my RippRatings don't give ANY insight into what is going on, who is good defensively and who isn't. There is absolutely no insight it offers...they just take the team Drtg I was given, and then add literally random noise to it, and then spits that out as the answers.

So that proves you can waste your time (which I don't care about - but you managed to waste my time, which I do care about). How does this further the conversation?

If your rating system is inferior to mine, and mine is literally spewing out random garbage, why should I listen to yours at all? There has to be some sort of test that your system passes so that we can distinguish it from a algorithm that produces just random noise.

Maybe because the extent to which you think my system is flawed isn't necessarily accurate? You seem to have a very high opinion of your ability to handle statistical data concerning defensive basketball, yet have tripped over your own feet (it's hard to see the ground with your nose up in the air) embarrassingly and repeatedly.

Again, the tracking of additional stats you've done is great, wonderful stuff. The Stop% stuff might also be decent too. However, going from there to your DRat seems to be flawed, if I can come up with a "RippRating" system that is clearly useless but just as good in explaining what actually happened in the real world.

I disagree entirely with your accusations, I think there's a lot of sound and fury here but not a lot of substance in your arguments and frankly it's entirely divorced from basketball reality. You haven't really done business with why this method has been developed the way that it has, what it's showing and why, and what it means. You have dismissed it out of hand (and using recycled and refuted argumentation), and pardon me for guessing that it's because you don't like the outcome, and then hunted rationale to (attempt to) excoriate the method. That doesn't really change anything, you might be right regardless of your a priori bias, but the fact is you have about as much standing to dismiss this method as I do: none.

[BorisDK1]

I'm not talking about recollection, dude. Look, we can measure how many points were scored in the 4th quarter exactly, right? We can measure how many possessions were used in the 4th quarter exactly right? And same for quarters 1 through 3. You agree here, right, that we can measure these things exactly? The scorekeepers measure exactly how many points were scored in the 4th quarter. From reading a play-by-play, you can count exactly how many possessions happened in the 4th quarter.

So if I have an "advanced basketball metric" that tells me that 80 points per 100 possessions in the 4th quarter, but instead they give up 150 points per 100 possessions, which should I trust? That advanced basketball metric, or the actual number of points per possession that were actually scored over in the 4th? Will I trust the metric, or what actually happened on the floor?

But that's not the dilemma at all!

The dilemma is the difference between the arguments, "The team gave up 30 points in 25 possessions with Player X on the floor, but only 25 points in 25 possessions when he was off the floor (which nobody has a problem with pointing out or not believing that it indicates something) - therefore that must have been player X's fault." It's the last part we are objecting to. It's the attribution of cause to an individual from that data that has to be objected to. To establish cause, is it not much more helpful to say, "In those 25 possessions, Player X faced allowed 7 of those points while facing 5 possessions, Player Y faced 4 possessions and gave up 5 points, Player Z gave up 8 points while facing 8 possessions, Player W faced 4 possessions and gave up 7 points and Player Ray-Jay Johnson faced 4 possessions and allowed 3 points" - and then we know who caused what? And you can call him Ray, or you can him Jay, just don't you dare call him Johnson - but that is a great individual performance but was incapable of making his team good overall.

The dilemma isn't in admitting that there is helpful data in on/off court numbers: it's the interpretation of it that's the problem. PDSS gives us far more information for cause that we can use to ask questions of the basketball floor. I feel like a broken record, but anybody using stats as answers and not opportunities for questions has fallen completely off the wagon.

[Ripp]

I'm not talking about recollection, dude. Look, we can measure how many points were scored in the 4th quarter exactly, right? We can measure how many possessions were used in the 4th quarter exactly right? And same for quarters 1 through 3. You agree here, right, that we can measure these things exactly? The scorekeepers measure exactly how many points were scored in the 4th quarter. From reading a play-by-play, you can count exactly how many possessions happened in the 4th quarter.

So if I have an "advanced basketball metric" that tells me that 80 points per 100 possessions in the 4th quarter, but instead they give up 150 points per 100 possessions, which should I trust? That advanced basketball metric, or the actual number of points per possession that were actually scored over in the 4th? Will I trust the metric, or what actually happened on the floor?

But that's not the dilemma at all!

No, that is exactly the point of this thread. An algorithm has been proposed (PDSS) that is inconsistent with what actually happened on the floor, in the way I just described in the above example.

The dilemma is the difference between the arguments, "The team gave up 30 points in 25 possessions with Player X on the floor, but only 25 points in 25 possessions when he was off the floor (which nobody has a problem with pointing out or not believing that it indicates something) - therefore that must have been player X's fault." It's the last part we are objecting to.

But who here in this thread is discussing that, or blaming anyone? You are attacking a straw man...that is not even the topic of this thread. Even if some here in the thread are doing this, screw them...that is not the focus of this thread, and I do not want it derailed into a blame game.

[Ripp]

hitting the grocery store, back in 30-40 min..

[S.W.A.N]

Is PDSS a good way to analyze the Raptors defense.....

Yes I think that this is a pretty useful tool for analyzing defensive plays. Its not the only tool but certainly a valid and useful one.

The biggest problem I see is that the OP seems to know a little about stats but not enough to understand what value it has.

And the reality is that no defensive stats are going to be perfect when it comes to determining a players effect on the defense. If you were part of the team in question and knew all of each players assignments for each situation you could better analysis who creates the most breakdowns in the defense, who is best at preventing breakdowns etc. but without knowing the assignments there is going to be subjective analysis and on top of that how do you track the plays that don't happen. For example a superior help defender may anticipate a play that barnani doesn't see coming and slides over and helps prevent an easy basket. Barny wouldn't be at fault for that play because it did not involve him but it doesn't mean that it shouldn't have involved him.

Regarldless PDSS can provide insight into a player defends and I look forward to anymore analysis that boris provides. Just like I look forward to Supersubs data as well. OP needs to provide some now instead of bashing other peoples work.

[Crazy-Canuck]

I think I got it.

Ripps number come out to xyz, but when adding the individual stats used by Boris, it comes out to xyzz.

So, Ripp is questioning the PDSS because xyz does not equal xyz.

[Ripp]

I think I got it.

Ripps number come out to xyz, but when adding the individual stats used by Boris, it comes out to xyzz.

So, Ripp is questioning the PDSS because xyz does not equal xyz.

It isn't my number, it is the team Drtg from say basketball-reference.com or basketballvalue.com. Certainly not my property or invention.
Also, this was a piece of infomation that is given to PDSS...yet it doesn't even regurgitate the number it was given.
It also doesn't know what the team Drtg for different lineups, nor does it know it for say the 4th quarter.

[supersub15]

LOL - supersub, is it possible that you dismiss me because I don't have a website, a budget and simply because you want to justify to yourself the money you spent on Synergy Sports when you could have had better data for free? Despite your obvious acumen and passion and ability to handle numerous metrics, you for whatever reason are very loathe to accept this. And no, I'm not trying to force PDSS on you or anybody else: I'm merely pointing out that PDSS gives us far more direct data and it should be interpreted and should force questions. I'm not seeing any questioning or any willingness to engage these numbers at all. Is that because you are such a firm believer in the methods you have become comfortable with using? I'd suggest that it's entirely the case, and at some point you'd be best advised to broaden rather than narrow your horizons.

As far as the bias against Calderon goes, now you're lurching into argumentum ad hominem: "you can't be trusted because you admit you don't like Calderon". Yeah, I made a humourous comment about Jose. I've made many in my life, mostly in retaliation for the ridiculousness I've had to endure at raptorspace.com at the hands of various Spaniards insisting Jose is simply a dominant point guard and anybody who dares usurp any of his minutes should be drawn and quartered in Nathan Phillips Square. I've also been critical of Bargnani - odd you don't mention that in your accusation. But that doesn't change the fact that I have invested too much of my time and passion and energy and knowledge into doing with the greatest possible integrity because, at the end of the day, I want a product I can be proud of and I believe I've produced something I can be proud of. You do the same thing with your spreadsheets, and would no doubt chafe if I accused you of hopeless bias in your recording of things. And finally, since this accusation is entirely fallacious as far as argument goes, I'll just point out that for the sake of argument I could have 16 contracts out on Jose's life as we speak and spent every Saturday night of the past 4 years slashing his tires in his parking garage every Saturday and I still might have recorded everything properly. What I think about Jose has nothing to do with how accurate I may have been, and I'm chuckling at the kind of mindset that thinks it might. And I also think you'd blow your stack if I accused you of the same thing with Bargnani.

As far as my ability to interpret the game, I don't have anything to prove. I've spent hours in seminars, reading, discussion with other coaches and officials about things and I do coach myself (at a very small college). I was coaching one of my high school's teams while I was still in high school. If you want to think that some people you don't know who in all likelihood do not have the technical or theoretical basis in the game that I do and who only have bothered to track a portion of the data that I have somehow help you out more than I do, go ahead. If I suggested that you wouldn't know the difference between a UCLA cut and a flex cut if your life depended on it, does it make my point any stronger? No. (For the record, I'm not suggesting that. I use that only as a for-instance of a ridiculous argument I'm not going to make - yet for some reason, you feel comfortable doing.)

And I'm not "throwing those stats back in your face" any more than you're dismissing data which happens to contradict your opinions out-of-hand. Let's be honest about that.

Sorry if I came away as condescending. Not trying to be, but to tell you the truth, I got pissed that you are vehemently dismissing other work (not just ON/OFF DRTG), just because they don't jive with your work.

I respect what you have done. It takes a lot of patience, effort and dedication. And your knowledge of actual basketball plays goes way beyond what I do know.

I said I will stop already once, but couldn't, but that's it, I promise. It was a pretty good discussion, enjoyed it immensely. We will continue to disagree on this, but don't want to get into a pissing match. I'm glad you've decided to finally join RealGM. Looking forward to further contributions.

Cheers.

[andreafan]

Now we have two stat geeks, ss and boris, just great.

[[SJJ]]

Now we have two stat geeks, ss and boris, just great.

Numbers don't lie - Numbers are awesome!!!

[disoblige]

supersub mentions bargs scores around 200th defensively in synergy yet fail to mention calderon is ranked around 400th there.

[Courtside]

So thinking this through a little more, I think I've come up with a way to bridge the 2 sides here...

First - there is some merit in all the stat methods shown. Personally, I find a PDSS/Synergy type breakdown extremely useful in looking at individual guys since team DRats and on/off does leave a lot to be desired. What on/off numbers do is tell you there are problems and points you in certain directions, it just doesn't tell you conclusively what they are. PDSS is able to look closer at the problems, but there is something lacking.... and here is what it is.

Boris - is it possible to add columns Help Given (say HG) and Help Missed (say HM)? If you're breaking down the plays as much as you are, surely you can also determine if a particular guy provided help or was standing 10 feet away with his thumb up his ass, right? There's some nuance in terms of whether he should or could have, but simply getting a 0.5 credit in the defensive stop made isn't enough to quantify this.

Further, I think a column for Defensive Rebounds Made (DRM?) and Offensive Rebounds Surrendered (ORS) would also be instructive, as many defensive possessions end in a Dreb.

I guess you would HG/HM and DRM/ORS and come up with some multiplier that is applied towards your current DRat - let's call it Team Defense Quotient (TDQ). If a guy is an active helper or good rebounder, his DRat would be adjusted by up or down by his DQ, whereas if he is a weak rebounder or help defender, his DQ would adjust his DRat the other way.

This TDQ adjusted DRat would ideally correspond more closely with on/off stats, for example.

Does this make sense to the statistical squareheads?

[MrBojangelz71]

Is PDSS a good way to analyze the Raptors defense.....

Yes I think that this is a pretty useful tool for analyzing defensive plays. Its not the only tool but certainly a valid and useful one.

The biggest problem I see is that the OP seems to know a little about stats but not enough to understand what value it has.

And the reality is that no defensive stats are going to be perfect when it comes to determining a players effect on the defense. If you were part of the team in question and knew all of each players assignments for each situation you could better analysis who creates the most breakdowns in the defense, who is best at preventing breakdowns etc. but without knowing the assignments there is going to be subjective analysis and on top of that how do you track the plays that don't happen. For example a superior help defender may anticipate a play that barnani doesn't see coming and slides over and helps prevent an easy basket. Barny wouldn't be at fault for that play because it did not involve him but it doesn't mean that it shouldn't have involved him.

Regarldless PDSS can provide insight into a player defends and I look forward to anymore analysis that boris provides. Just like I look forward to Supersubs data as well. OP needs to provide some now instead of bashing other peoples work.

+1

Stats are like the perfect woman in that there isn't one. Each has their flaws, and in life, you have to accept the good with the bad. Now if you could take 5 of them and somehow combine only their good traits you may come up with the absolute perfect stat, or woman, that totally defines what you are looking for.

[theonlyeastcoastrapsfan]

"You can't triple stamp a double stamp, lloyd.