SoTD: Bargnani's defensive numbers over 4 years

[andreafan]

Well calderon pulled his hammy so andrea won't have to cover his arse on defense for a while , nice.

[BorisDK1]

Sure, for the beer, I'll email Adam when I am in the Burlington vicinity... Should be good!

I don't dispute that you've been diligent with your work, but I'll keep it as a reservation for now, because of eye (things can happen quickly, and as you said some of the possessions were not logged), prejudice, system (how do you know what Triano wants?), etc.

I'll bite on the Jose theory:
Team DRTG for the combo - Calderon / Bosh / another big (NO BARGNANI): 104.67
Team DRTG for the combo - Calderon / Bargnani / another big: 113.72

Your turn.

I don't think the eye is any more jaundiced in my case then in the NBA's case with assists, rebounds or steals. It's a simple system: either you're guarding a guy or not, and in the outcome of a play you're either contributing to a stop or allowing a score in whatever way. It's not that hard to do.

When I said not every possession was tracked, I meant with Synergy, not with me. With my system, I got the outcome of every play this past season. I'm basing only on what I've heard, I'm not a subscriber and I'll certainly accept correction if necessary, but it sounds like Synergy is tracking outcome only of primary identifiable and trackable action (ball screens, post ups, dribble drives (?)). Are they tracking what's happening in transition? In ball reversal? In offensive rebounding? I get all that, you see. It's sometimes very difficult, in some offensive rebounding and transition, but I'm not giving ridiculously imbalanced results. I'm not tagging Jose for a FGA in a situation when he's still in the frontcourt when the opponents score on a 2-on-1, for instance. I hate his play, but not so much more than the rest of these softies that I'd go that far out of my way to blame everything on him.

How do I know what Jay wants? You watch enough Raptors basketball with that much attention to detail with a base knowledge of defensive tactics and it's not that hard. The Raptors played a very standard defense this year. The people decrying Jay's "system" IMO know nary a thing about basketball strategy and tactics. There's the occasional play that's just a complete mess-up where they're all completely out to lunch but far more often the issue was just plain poor execution. I think I'm well-versed enough to determine who was the principal cause of breakdown on a diagonal down-screen sequence, for example (something the Raptors could not stop to save their lives this year). It's not that hard to tell when a team is trying to jump-switch vs. contact-switch vs. lock-and-trail vs. jump over vs. jump under vs. just screw things up when you recognize these actions for what they are. It didn't even take me long to recognize that Jay ended up developing an entirely separate set of rules for Jose than for everybody else: switch all ball screens, switch all crossing action (except dribble handoffs), switch all screens when he's guarding a cutter. Nobody else had separate rules just for them. (Not that it did an iota of good.)

So, you're trying to pose the on/off court data (still) as explanatory for cause, but I still feel you're trying to use those data as 1) causal of team's bad defense and 2) judging solely one player at a time. I know you don't believe the latter because you've warned people against that, but I still think you're trying to use a metric which is merely an observation as to what happens as trying to explain why that happened, which I do not think is warranted.

But having said that, let's look at this:

Oct 28 - Dec 4 (Jose starting and healthy) Team DRat = 117.8
Dec 5 - Jan 3 (Games starting when Jose got hurt till the end of his injury) Team DRat = 106.4
Jan 6 - Mar 14 (Jose coming off bench) Team DRat = 112.9
Mar 17 - Apr 14 (Jose starting again) Team DRat = 113.4

Now, if you want to talk about a significant bloc of data, let's talk Dec 5 - Jan 3 when out of 821 minutes of game time and 1583 possessions, the team's defense suddenly improved by 11.4 points / 100 possessions. That's a big bloc of time, a good chunk of possessions. What are the odds that was a complete fluke?

Now, the 80 games in which Andrea played: Team DRat = 113.1
The 2 games he missed (Dec 9 and Jan 22, ironically both against Milwaukee): 115.1 (against a poor offensive team).

I think the on/off court method leaves way too many questions unanswered (against whom? who's responsible?). It indicates that Andrea is a defensive problem, and I would agree with that, but at the absolute worst I think he's a secondary cause of our problems. I'm not arguing he's a good defender, because he's not, but he's not egregiously bad in some regards either. I think the metrics when properly viewed show fairly conclusively that the only race for problem child defensively on this team is a race for second place next to Jose.

[Ripp]

I'm sorry to weigh in on such a controversial topic for my first post here, but I find much of the argumentation here overly facile. The people blindly insisting Bargnani's defense is somehow acceptable are oblivious to the fact that he just played over 2,800 minutes for what is the 21st worst defense in league history - well, since 1978 when we first had the requisite data, at any rate (measured by points allowed / 100 possessions worse than league average) so obviously he has to be lacking something.

I'm not sure I agree....it is possible to be a good defender on a bad defensive team. Or conversely, a bad defender on an elite defensive team. The question is, who is having what effect? Were the Magic an elite defensive team last year because of Hedo? Being able to quantify a bad defender on a good team and good defenders on bad teams is very useful, as the Hedo example indicates...

On the other hand, the discussion of primary cause of the defense from the OP and others is equally facile. I find using on/off court data as indicators of cause to be fallacious (post hoc, ergo propter hoc) and further, probably not using the best data available.

On some level, you are correct...we cannot be 100% sure from statistical data who is causing what. As you say, it would technically be a fallacy to look at +/- data or box scores, or in fact ANY sort of statistics to prove that for example LeBron was the reason the Cavs were an elite offensive team. Or Dwight causing the Magic an elite defensive team. Nevertheless, we will still look at the data and make decisions based on it.

I believe the best data available is through a method called Project Defensive Score Sheet (spelled out in Dean Oliver's Basketball on Paper), in which all the stops produced and scores allowed are tabulated for all the players (and uncontested action as well) and the individual Defensive Rating generated from that is the best way of generating defensive metrics. That data is not currently kept in boxscores by the NBA, so that data has to be kept manually. I did that for this past year - every single possession, all 82 games.

So can you outline his methodology, describe it? Some methodologies make sense, others do not (by "make sense" I mean accurately gauging who is doing a good job on defense and who is not.)

This past year, Bargnani played ~ 2800 minutes. In those minutes, he forced 553.5 missed field goal attempts (not including blocks), 75.5 turnovers (not including steals) and 64 missed free throws, to go along with his 389 defensive rebounds, 111 blocks and 25 steals. He allowed 404 field goals and 188 made free throws to go along with those. That translates to a stop% of .552 while facing 19.2% of the opponents' possessions defensively, giving him an ultimate Defensive Rating of 110.0 points allowed / 100 possessions faced, which was second-best of anybody on our team playing more than 350 minutes, behind only Amir Johnson (.551 stop% @ 22.8% DPoss%, 109.5 DRat) and tied with Bosh (.545 stop% @ 19.5% DPoss%, 110.0 DRat). (All the team defensive numbers used to generate the DRat only come from the games in which the individual actually play.) Now, that's a far cry from being the hapless yutz that some want to present him as. To be sure, you would hope your center is better than that - and certainly facing more possessions. But it's far better than some want to portray.

So this Stop% stuff sounds kooky to me. Again, it would be helpful if you could explain the methodology so we can understand how it works. However, I do like tracking defensive results, Synergy style. I'd be really happy if you published that sort of stuff on Google docs.

BTW, just to reveal my own biases...I'm pretty pro APM and its variants. Statistical Plus/Minus is a pretty useful tool, and I've worked on some variants of it myself that I think are even better. That is sort of the right direction for analyzing players...using both +/- data and box scores. Honestly, I wish synnergy could publish a sort of expanded end-of-season box score for players that included defensive variables. That I think would be extremely useful.

[BorisDK1]

I'm not sure I agree....it is possible to be a good defender on a bad defensive team. Or conversely, a bad defender on an elite defensive team. The question is, who is having what effect? Were the Magic an elite defensive team last year because of Hedo? Being able to quantify a bad defender on a good team and good defenders on bad teams is very useful, as the Hedo example indicates...

Playing that many minutes on a historically bad team has to raise some serious questions. There are types of defenders - Joe Dumars, for example, who defended very well but who did not produce a lot of defensive statistics - who have played on both great and bad defensive teams. Is Andrea in that category? I'd say no.

On some level, you are correct...we cannot be 100% sure from statistical data who is causing what. As you say, it would technically be a fallacy to look at +/- data or box scores, or in fact ANY sort of statistics to prove that for example LeBron was the reason the Cavs were an elite offensive team. Or Dwight causing the Magic an elite defensive team. Nevertheless, we will still look at the data and make decisions based on it.

That was a comment specifically on the on/off court data, and trying to use it to establish cause.

So can you outline his methodology, describe it? Some methodologies make sense, others do not (by "make sense" I mean accurately gauging who is doing a good job on defense and who is not.)

It tracks who produces what defensively by not only the blocks, steals and defensive rebounds we already have, but by tracking:

Forced misses: A player is guarding a player, who shoots and misses from the floor. That is a forced miss. Credit can be shared between teammates (player gets beat on the dribble, but stays in the play and helping teammate forces a missed attempt. They both get .5 FM). Blocked shots are separate.
Field goals allowed: same deal, except the try goes in.
Forced turnover: a player is guarding another player, who turns the ball over (not including steals). A forced turnover is credited to that player.
Forced free throw: a player fouls an opponent, who makes the resulting free throw(s). These are forced free throws.
Forced free throws missed: same deal, except free throws are missed.

There are some peculiarities for special situations, but suffice it to say uncontested action is attributed to category "team", not an individual player.

So this Stop% stuff sounds kooky to me. Again, it would be helpful if you could explain the methodology so we can understand how it works. However, I do like tracking defensive results, Synergy style. I'd be really happy if you published that sort of stuff on Google docs.

Meaning, "it sounds new to me".

Stop% = Total stops / Total individual defensive possessions.

You can read more on this method in Google Books, just look up "Dean Oliver Basketball on Paper".

BTW, just to reveal my own biases...I'm pretty pro APM and its variants. Statistical Plus/Minus is a pretty useful tool, and I've worked on some variants of it myself that I think are even better. That is sort of the right direction for analyzing players...using both +/- data and box scores. Honestly, I wish synnergy could publish a sort of expanded end-of-season box score for players that included defensive variables. That I think would be extremely useful.

I have very little use for these magical one-number metrics (APM, SPM, aSPM, PER, etc.).

I have done what Synnergy could do - except more accurately, and involving more than direct action.

[Ripp]

Oh, I forgot to say..welcome BorisDK1! Always nice to have other voices here, especially when they can reason and think. Helps to counter the usual large number of...people...that feel obliged to post and comment on things they don't understand (I'm looking at you, ZefSyde.)

[Ripp]

I'm not sure I agree....it is possible to be a good defender on a bad defensive team. Or conversely, a bad defender on an elite defensive team. The question is, who is having what effect? Were the Magic an elite defensive team last year because of Hedo? Being able to quantify a bad defender on a good team and good defenders on bad teams is very useful, as the Hedo example indicates...

Playing that many minutes on a historically bad team has to raise some serious questions. There are types of defenders - Joe Dumars, for example, who defended very well but who did not produce a lot of defensive statistics - who have played on both great and bad defensive teams. Is Andrea in that category? I'd say no.

I dunno..you might be right. I've not looked at enough bad defensive teams to know for sure. But I know several elite defensive teams that have a few bad defenders mixed in, average defensive teams with elite/mediocre/terrible defenders, etc. Anyway, it isn't a big deal either way...I agree that Bargs is a bad defender, but don't think all the guys who played major minutes on this team are bad.

On some level, you are correct...we cannot be 100% sure from statistical data who is causing what. As you say, it would technically be a fallacy to look at +/- data or box scores, or in fact ANY sort of statistics to prove that for example LeBron was the reason the Cavs were an elite offensive team. Or Dwight causing the Magic an elite defensive team. Nevertheless, we will still look at the data and make decisions based on it.

That was a comment specifically on the on/off court data, and trying to use it to establish cause.

Ah, but which technique in the world can establish causality? How does one prove that X caused Y in basketball, except in non-trivial cases? I think most of us agree that LeBron is the reason the Cavs were an elite offensive team. How then does one prove this? Most statistical techniques will do a good job of revealing this, but estimates/inferences implicitly have confidence intervals or whatever associated....in layman's terms, some level of uncertainty in their answers. So there is always uncertainty, no matter what you do.

So can you outline his methodology, describe it? Some methodologies make sense, others do not (by "make sense" I mean accurately gauging who is doing a good job on defense and who is not.)

It tracks who produces what defensively by not only the blocks, steals and defensive rebounds we already have, but by tracking:

Forced misses: A player is guarding a player, who shoots and misses from the floor. That is a forced miss. Credit can be shared between teammates (player gets beat on the dribble, but stays in the play and helping teammate forces a missed attempt. They both get .5 FM). Blocked shots are separate.
Field goals allowed: same deal, except the try goes in.
Forced turnover: a player is guarding another player, who turns the ball over (not including steals). A forced turnover is credited to that player.
Forced free throw: a player fouls an opponent, who makes the resulting free throw(s). These are forced free throws.
Forced free throws missed: same deal, except free throws are missed.

There are some peculiarities for special situations, but suffice it to say uncontested action is attributed to category "team", not an individual player.

This is a great idea. I'd love to see box scores of this type of stuff published. You have them for game by games for this past season, right? Would be nice if you could put it on Google docs.

So this Stop% stuff sounds kooky to me. Again, it would be helpful if you could explain the methodology so we can understand how it works. However, I do like tracking defensive results, Synergy style. I'd be really happy if you published that sort of stuff on Google docs.

Meaning, "it sounds new to me".

Stop% = Total stops / Total individual defensive possessions.

You can read more on this method in Google Books, just look up "Dean Oliver Basketball on Paper".

By "kooky", I really mean that "some things in this formula you posted seem weird...can you provide a reference explaining the decisions." I'm happy to see new things...but as a courtesy, one usually explains a methodology rather than just saying, "here, take and accept this."

I'll do some research on this method and then post my findings for everyone else too.

BTW, just to reveal my own biases...I'm pretty pro APM and its variants. Statistical Plus/Minus is a pretty useful tool, and I've worked on some variants of it myself that I think are even better. That is sort of the right direction for analyzing players...using both +/- data and box scores. Honestly, I wish synnergy could publish a sort of expanded end-of-season box score for players that included defensive variables. That I think would be extremely useful.

I have very little use for these magical one-number metrics (APM, SPM, aSPM, PER, etc.).

I have done what Synnergy could do - except more accurately, and involving more than direct action.

I'm not particularly a one-number metric sort of guy...but in experiments I've done, variants of SPM seem to do the best in actually predicting performance in games. It makes sense to look at techniques that accurately predict what will happen rather than those which do not.

Regarding your comment...it would probably be best to put your stuff online. That way nobody has to take anyone's word for anything...we can see how accurate you really were, etc.

[Indeed]

I think the answer is significantly more complex than "blame Andrea". I think the people who want to do that, unfortunately, are using guesswork generated by on/off court analysis from 82games.com instead of direct analysis. I suppose that's becasue there is a dearth of popularly published data for defensive metrics, and because many people struggle with defensive analysis. This tends to be expressed by assigning the vast majority of blame/credit for defensive performance to the big men, which I personally believe is an extremely naive view of defense. And then we have the blaming of Triano's defensive tactical beliefs for the team's defensive woes, which further is untenable.

I think the problems lay in having so many players (Andrea, Hedo, Jose) on the court who are at considerable disadvantage of speed and quickness, and consequently the team had absolutely no pressure on the basketball at any time. Pressure on the ball is the absolute foundation of your defense, and without it nothing you try will work.

Thank you for the analysis and insight, sorry, didn't read this earlier.

I completely agree with your analysis, and definitely we lacked pressuring our opponent as one of the league worst in steals, as Bargnani is no Howard nor anything close in help defense/toughness.

I am wondering if you have the Stop% for DeRozan as well.
- Can he guard SF?
- Is he quick enough to defend PG? (Weems is quick enough)

[Crazy-Canuck]

Boris, what do your numbers tell you about JJ?

Most the stats guys here show that defensively, Jose and JJ seem comparable. The problem with these stats are that they just dont pass the eye test. Maybe Im stubborn, but I refuse to believe JJ is as bad as Jose.

[timdunkit]

Personally, I find that the Raptors defense last year was a bit of a statistically anomaly. How many times has a team had one of the worst defenses of all time in consecutive years? Let say the Raptors improve there defense by 2-3 points in terms of efficiency, does that mean the defense actually got better? Well statistically, it would ... but in reality, you still rank near the worst in the league.

Personally, I find it easier to rank defense in tiers rather then using specific stats. If your in a certain range in terms of defensive impact then thats where you rank. Taking a range, gives a better idea and it accounts more for errors. For example, Bosh and Bargs DRTG was 113 and 116 respectively. We were the worst defensive team with a DRTG of 113. Is Bosh's 113 really that better then Bargnanis 116 taking into account that the worst defensive team had a DRTG of 113? When there both about in line the worst defense in the league. It shouldn't matter whether the DRTG is 114,115,112,111 etc ... in context, there absolutely terrible DRTG so why compare them? Can you really say one is better then the other, when it could just be due to an error (actually, it would be interesting to see if you could possibly formulate an error value for calculating DRTG).

As for the Raptors defense, I think its a joke if anyone argues that Bargnani is a good help defender ... likewise, I think its a joke if you single him out as the single cause of our defensive woes. Aside from D12, there isn't a single centre or anchor in the league, that would have help this team reach top 20 in defense.

We were absolutely terrible at defending in all aspects. We were terrible at defending the 3pt line, gave away way too much dribble penetration and terrible at containing the penetration, terrible in transition, terrible at forcing turnovers, terrible at guarding 1 on 1 players on the perimeter and terrible on the defensive glass. This isn't just one players fault, it was a problem for the entire team.

[Ripp]

Alright, that was tremendously annoying. Would be so nice if Google Books didn't snip out so many useful pages from that book. Anyway, so from what I can piece together from that and some useful sites I Googled (linked at the bottom of this post), here is the motivation behind this.

Dean Oliver wants to give a individual Drtg to each player (note that this isn't the same as Drtg from say basketballvalue.com, which measures what actually happened when the player was on the floor...a bit unfortunate and confusing that these two things have the same name.)

So here is how you calculate Oliver's Drtg:

Code: Select all

OliverDrtg = TeamDrtgOne + NetPerformanceModifier
NetPerformanceModifier=.2 x [TeamDrtgTwo x (1 - Stop%) - TeamDrtgOne]


Note that the formula is similar to the one Boris posted...I've absorbed the "100 x (Opponent Points / Scoring Possession)" term that Boris wrote into a new term called TeamDrtgTwo. We'll discuss later how these two terms are calculated, but just keep in mind that they should be somewhat related.

So let's get the gist of what this formula is trying to do. Well, you take some sort of team Drtg (TeamDrtgOne), then add to that some sort of modifier NetPerformanceModifier that measures whether a player is having a net positive or negative affect on the team's defense.

What is the gist of this NetPerformanceModifier term? You take some other sort of team Drtg (TeamDrtgTwo), multiply by (1-Stop%) to get how much a player has improved the team defense by, subtract off another team Drtg, and then multiply by the fraction of defensive possessions a player was involved in (the term %DPoss.)
So if you are some sort of amazing defender, your Stop% will be really high. Like, a guy with a Stop% of 100% (e.g., prime KG on crack) will have a massively negative NetPerformanceModifier term. We might have another guy, Jose Pyloneron who has a Stop% of 0, and thus will give a large positive NetPerformaceModifier, indicating a very bad defender (Actually this isn't quite accurate...I'm assuming that TeamDrtgTwo is always larger than TeamDrtgOne, something that wasn't obvious to me from the formulas for the two terms. But under this assumption that is the case, what I said is correct..)

So hopefully this gives you a sense of how this formula works. Seems like magic, right? If you can figure out this Stop% stuff, then you win in a most epic manner.

So let's think about this magical Stop% number. Here is the formula for it:

Code: Select all

Stop% = [Stops/PlayerMin] x [TmOppMin]/[TmOppPoss]


In other words, we calculate some quantity called "Stops", divide by the number of minutes I've played, then adjust for the total number of possessions our opposing teams were involved in. The "feel" of this formula is, "what fraction of opposing team possessions did I individually get stops on?"

So, the final ingredient is, how exactly are "Stops" defined? Here is the formula:

Code: Select all

Stop = FTO + FFTA/10 + FM x  FMwt x (1 - DOR%) + DREB x (1 - FMwt)


Let's define each term:
FTO = Forced Turnover
FFTA = Forced Free Throw Attempt, I guess?
FMwt = A weighting term between 0 and 1 that figures out how we should weight Defensive Rebounds.
DREB = Defensive Rebounds


Alright, this covers most of the stuff in the formula. I didn't discuss how TeamDrtg1 and TeamDrtg2 are calculated...there are different ways to calculate it and frankly it really is not going to be important for my criticism in my subsequent post (and you can click through and read through for the reference in my next post.)

So my next post will have a link to my references used. After that, I'll detail my criticism of Oliver's Drtg.

[Ripp]

Here are the resources I used, and a short summary of each:

1. http://sonicscentral.com/apbrmetrics/vi ... f44a8fbf12
(Oliver's book via Google docs. Unfortunately many of the relevant pages cannot be seen...so kinda annoying)
2. http://www.rawbw.com/~deano/articles/JordanvsOlaj.html
(a VERY nice article by Oliver in which he motivates and describes his Drtg calculation. It offers a slightly different perspective than what I said, and should help you understand it even better than what I wrote.)
3. http://sonicscentral.com/apbrmetrics/vi ... f44a8fbf12
Discussed how Oliver's Drtg is calculated differently from the play-by-play based stuff on basketballvalue.com
4. http://sonicscentral.com/apbrmetrics/vi ... 42fad1ee78
Basically a post by a guy who found some counterintuitive results from Oliver's Ortg and Drtg. I'll talk about this in my next post.
5. http://sonicscentral.com/apbrmetrics/vi ... 0bbf8f83e7
A bit of controversy on how exactly to calculate TeamOrtg1 and TeamOrtg2.
6. http://hoyaprospectus.blogspot.com/2008 ... tings.html
A guy who compared Oliver Ortg and Drtgs to play-by-play based ones, ala basketballvalue.com. He observes that Oliver's Ortg matchs up fairly well with the actual play-by-plays, but Oliver's Drtg (and thus Stop Percentage) does not.


Alrightly, hope some of you guys found this useful. Took my an assload of time to do w/o the book, but I feel that I understand this formula, its motivations and limitations pretty well now.

[Ripp]

Alright, so once you understand the formula, its limitations become pretty clear. Here are the main assumptions the formula makes that are at least arguable:

1) A player's defensive value is primarily a function of their Stop Percentage.
2) How exactly does one trade off defensive rebounds and other stops? I.e., what is the right value for this Fmwt term? Dean Oliver has a formula in his book, and an ad hoc weighting of 1/2 in that article written by him that he wrote. Both to me seem very subjective.
3) Why are players getting credit for FTs that the guy they fouled missed? Seems perverse for me to benefit from sending someone to the line....
4) Why divide by 10 in this Free Throws Missed part of the formula?
5) Why the weighting of .2 for NetPerformanceModifier? As a commenter in link #1 points out, this doesn't really make any sense.

So the above problems are big deal. But really the beef I have with this formula is that it is the following:
1) This formula is very separated from actual play. It doesn't seem to pass basic smell tests. For example, if I take a time-weighted average of the Oliver Drtgs of the players on a team, does this match up closely with the actual team Drtg? Reference #4 Explores this issue for the Phillys. Highly recommend that you read this and understand why it is very concerning. Like, of what use is a Drtg formula and estimate of a player's defensive performance that doesn't even pass a basic smell test like this? It means it'll have no predictive value for what is actually going to happen for the game played tomorrow, for example...
2) The key property of this stat is to boil down a player's defensive value into this Stop Percentage variable. However...notice that there is a tremendous amount of subjectively and room for error in assigning credit here. Again, I'm ignored questionable DESIGN decisions like giving somebody credit for dudes they fouled who missed FTs, or the 1/10 factor there. Like, even if those two choices are the CORRECT ones, depending on how you assign credit for defensive mistakes, it entirely changes the Stop% quantity....and thus the validity of this stat.

This is fundamentally why I had so much beef with the way you posted this....it is extremely bad form to advocate a statistic, tool, or methodology and then not discuss major issues like this. When you introduce a tool, you need to explain very carefully how it works, its flaws and limitations. Not trying to rag on you much, but this is why people beef with statisticians and sometimes do not trust them.

[supersub15]

When I said not every possession was tracked, I meant with Synergy, not with me. With my system, I got the outcome of every play this past season. I'm basing only on what I've heard, I'm not a subscriber and I'll certainly accept correction if necessary, but it sounds like Synergy is tracking outcome only of primary identifiable and trackable action (ball screens, post ups, dribble drives (?)). Are they tracking what's happening in transition? In ball reversal? In offensive rebounding? I get all that, you see. It's sometimes very difficult, in some offensive rebounding and transition, but I'm not giving ridiculously imbalanced results. I'm not tagging Jose for a FGA in a situation when he's still in the frontcourt when the opponents score on a 2-on-1, for instance. I hate his play, but not so much more than the rest of these softies that I'd go that far out of my way to blame everything on him.

For the Synergy thing, they do track every play, but there's a number of plays that are assigned to team defence instead of individual defence. Things like cuts are very hard to assign blame to individual players, loose balls ending in a bucket where the entire defence is out of shape, zone defence where the shooter ends up having no one guarding him, transition baskets, etc. I'm not sure how you were able to assign blame/credit with 100% certainty with your system.

How do I know what Jay wants? You watch enough Raptors basketball with that much attention to detail with a base knowledge of defensive tactics and it's not that hard. The Raptors played a very standard defense this year. The people decrying Jay's "system" IMO know nary a thing about basketball strategy and tactics. There's the occasional play that's just a complete mess-up where they're all completely out to lunch but far more often the issue was just plain poor execution. I think I'm well-versed enough to determine who was the principal cause of breakdown on a diagonal down-screen sequence, for example (something the Raptors could not stop to save their lives this year). It's not that hard to tell when a team is trying to jump-switch vs. contact-switch vs. lock-and-trail vs. jump over vs. jump under vs. just screw things up when you recognize these actions for what they are. It didn't even take me long to recognize that Jay ended up developing an entirely separate set of rules for Jose than for everybody else: switch all ball screens, switch all crossing action (except dribble handoffs), switch all screens when he's guarding a cutter. Nobody else had separate rules just for them. (Not that it did an iota of good.)

I'm impressed with your knowledge of basketball plays, and I'm confident that you know your stuff, but again, on some plays, there is no way for you or anyone else to assign individual blame. It goes back to team defence breaking down.

So, you're trying to pose the on/off court data (still) as explanatory for cause, but I still feel you're trying to use those data as 1) causal of team's bad defense and 2) judging solely one player at a time.

As Ripp has said, basketball is a fluid game and there's really few statistical data available to measure a player's full impact on the court. I can surmise from using 4 years of data and 4000+ minutes of play time that Bargnani has a negative overall impact on team defence. I mean when, year-over-year, team defence suddenly improves when he's off the court, it sure means one thing and one thing only. If it happened in only one year, then you have a point. As to judging one player at a time, I've tried isolating different players with and without Bargnani on the court, and in ALL cases, the player I've tried to isolate ended up having a better influence on team defence when Bargnani is off the court. It's weird. Try explaining that.

But having said that, let's look at this:

Oct 28 - Dec 4 (Jose starting and healthy) Team DRat = 117.8
Dec 5 - Jan 3 (Games starting when Jose got hurt till the end of his injury) Team DRat = 106.4
Jan 6 - Mar 14 (Jose coming off bench) Team DRat = 112.9
Mar 17 - Apr 14 (Jose starting again) Team DRat = 113.4

Now, if you want to talk about a significant bloc of data, let's talk Dec 5 - Jan 3 when out of 821 minutes of game time and 1583 possessions, the team's defense suddenly improved by 11.4 points / 100 possessions. That's a big bloc of time, a good chunk of possessions. What are the odds that was a complete fluke?

Now, the 80 games in which Andrea played: Team DRat = 113.1
The 2 games he missed (Dec 9 and Jan 22, ironically both against Milwaukee): 115.1 (against a poor offensive team).

I think the on/off court method leaves way too many questions unanswered (against whom? who's responsible?). It indicates that Andrea is a defensive problem, and I would agree with that, but at the absolute worst I think he's a secondary cause of our problems. I'm not arguing he's a good defender, because he's not, but he's not egregiously bad in some regards either. I think the metrics when properly viewed show fairly conclusively that the only race for problem child defensively on this team is a race for second place next to Jose.

First, let me say that I agree that Jose Calderon is a very poor defender. But, as I've said on numerous occasions before, when Jose gets beat at the point of attack, someone usually has to rotate to stop the drive. When it's Bargnani as the last line of defence, it's an almost-guaranteed layup. When it's Johnson, Bosh, Humphries, Rasho, Reggie, O'Neal, etc., there's at least some resistance, and the basket isn't guaranteed anymore. As I've shown you earlier, when Calderon isn't on the court with Bargnani (4000+ minutes worth of data, over a 4-year period), the team DRTG is 104.67. But when he plays with Bargnani, the team DRTG jumps to 113.72. It's mind-boggling. Pick a player, any player, and I can run the numbers for you with/without Bargnani, and it'll be the same result, every time.

Now that this is out of the way, let's look at the stats you posted.

The chunks of games that you used include Bargnani playing 35+ minutes, so that means that the team's DRTG is already inflated with or without Calderon. What you need to do is isolate court time in this fashion: Calderon/Bargnani and Calderon/No Bargnani. But then, you run into small sample size, and the whole exercise becines moot. Which brings me to the 2-game Milwaukee section you used (Bargnani injured). Do you truly believe that 96 minutes of data is a worthy chunk of data when I've been using 4000+?

Finally, you'll notice that I haven't put last year's disastrous defence entirely on Bargnani's shoulder. There were too many holes to fix on the team. For instance, you can't keep bailing out Jose with help defence. And some of the decisions he makes are just baffling: why are you guarding Rondo chest to chest? Why are you laying off Jameer Nelson? Similar questions should be asked about DeRozan and Turkoglu. It was just a bad combination overall. But, what I did do, was - I hope - prove that Bargnani has a negative overall influence on team defence. Why else does the team defence improve by 8+ points when he's not playing?

I guess we just disagree on just how bad is Bargnani. I think he's egregiously bad, you think he's just bad. Different shades of pale.

And Ripp, thanks for the refresher. I enjoyed reading your posts.

[I_Like_Dirt]

I don't think the eye is any more jaundiced in my case then in the NBA's case with assists, rebounds or steals. It's a simple system: either you're guarding a guy or not, and in the outcome of a play you're either contributing to a stop or allowing a score in whatever way. It's not that hard to do.

See, by suggesting it's simple is where I very quickly start to disagree. Individual analysis on a play-by-play basis when your man has the ball is definitely a useful tool, but a person needs to be aware of its limitations in order to actually get any value from it. One particular limitation is that it doesn't take team defense into account whatsoever.

In Bargnani's case, this is particularly relevant, because while he is close to passable when it comes to man defense, he does so because his teammates support him a lot better than he does. Bargnani's man makes a motion to drive by him and his teammates react behind him, cutting off the bigger problem angles. A guard's man makes a motion to attack, and Bargnani is quite often caught looking the other way entirely and no reaction leads to an easy score. It doesn't happen all the time. Heck, the percentage of times it happens is relatively small, but it doesn't take many times per game to add up to 8 points. The end result is a bunch of players on the floor constantly second guessing who is going to be helping who at any time Bargnani is on the floor. That's where Bargnani's lack of value defensively costs the team more than any other player. He can harass his man all he wants, but so long as he doesn't support his teammates, he's just plain awful defensively.

[MrBojangelz71]

I don't think the eye is any more jaundiced in my case then in the NBA's case with assists, rebounds or steals. It's a simple system: either you're guarding a guy or not, and in the outcome of a play you're either contributing to a stop or allowing a score in whatever way. It's not that hard to do.

See, by suggesting it's simple is where I very quickly start to disagree. Individual analysis on a play-by-play basis when your man has the ball is definitely a useful tool, but a person needs to be aware of its limitations in order to actually get any value from it. One particular limitation is that it doesn't take team defense into account whatsoever.

In Bargnani's case, this is particularly relevant, because while he is close to passable when it comes to man defense, he does so because his teammates support him a lot better than he does. Bargnani's man makes a motion to drive by him and his teammates react behind him, cutting off the bigger problem angles. A guard's man makes a motion to attack, and Bargnani is quite often caught looking the other way entirely and no reaction leads to an easy score. It doesn't happen all the time. Heck, the percentage of times it happens is relatively small, but it doesn't take many times per game to add up to 8 points. The end result is a bunch of players on the floor constantly second guessing who is going to be helping who at any time Bargnani is on the floor. That's where Bargnani's lack of value defensively costs the team more than any other player. He can harass his man all he wants, but so long as he doesn't support his teammates, he's just plain awful defensively.

Suggesting that Andrea receives better support on help defense than he provides is somewhat comical to me.

According to your theory then, Andrea relies on help defense a lot less than our wings and power forwards, therefore he is required to provide help more often than he needs it to correct his man defense. So then it would be safe to say he will be more prone to more mistakes with it because our man defense sucks and it exploits our center to correct it.

Allowing your wing or guard to blow by you and then lay blame for the center not rotating quick enough to hide your mistake should fall onto the teams responsibility as much as it seems to fall onto Andrea's responsibility.

[YogiStewart]

I am sorry....
posts like these do not belong on our board during the offseason.

can you guys please talk about our jerseys or getting JR Smith or about the Dance Pak? you're kind of freaking me out a bit.

[timdunkit]

I don't think the eye is any more jaundiced in my case then in the NBA's case with assists, rebounds or steals. It's a simple system: either you're guarding a guy or not, and in the outcome of a play you're either contributing to a stop or allowing a score in whatever way. It's not that hard to do.

See, by suggesting it's simple is where I very quickly start to disagree. Individual analysis on a play-by-play basis when your man has the ball is definitely a useful tool, but a person needs to be aware of its limitations in order to actually get any value from it. One particular limitation is that it doesn't take team defense into account whatsoever.

In Bargnani's case, this is particularly relevant, because while he is close to passable when it comes to man defense, he does so because his teammates support him a lot better than he does. Bargnani's man makes a motion to drive by him and his teammates react behind him, cutting off the bigger problem angles. A guard's man makes a motion to attack, and Bargnani is quite often caught looking the other way entirely and no reaction leads to an easy score. It doesn't happen all the time. Heck, the percentage of times it happens is relatively small, but it doesn't take many times per game to add up to 8 points. The end result is a bunch of players on the floor constantly second guessing who is going to be helping who at any time Bargnani is on the floor. That's where Bargnani's lack of value defensively costs the team more than any other player. He can harass his man all he wants, but so long as he doesn't support his teammates, he's just plain awful defensively.

Suggesting that Andrea receives better support on help defense than he provides is somewhat comical to me.

According to your theory then, Andrea relies on help defense a lot less than our wings and power forwards, therefore he is required to provide help more often than he needs it to correct his man defense. So then it would be safe to say he will be more prone to more mistakes with it because our man defense sucks and it exploits our center to correct it.

Allowing your wing or guard to blow by you and then lay blame for the center not rotating quick enough to hide your mistake should fall onto the teams responsibility as much as it seems to fall onto Andrea's responsibility.

Did you forget that we allowed team to shoot a high percetages of 3's and a lot of them too ... Bargnani should be out there contesting those SHATS FOLKS!

Like I said, trying to make Bargnani a scapegoat for our defensive lapses, is a joke because there was too many dam wholes in our defense. Our bigs sucked, our wings sucked ... arguing who sucked more is nice but it doesn't change the fact that they ALL SUCKED!

The Raptors played there best defense during the stretch Jose was hurt. Why? Because the back-up PG was Marcus Banks who I was more then willing to go to as the backup for the rest of the season and trad Jose. Why? Banks picked up his man full-court, he always applied pressure and always fought and created resistance when his man went to the basket ... Hence, he was probably our best defender. Maybe Boris can give us a more detail report, but its no wonder Banks had the highest DRTG on the team a 104.

Basketball is a game of relationships ... You can't put the blame on one thing and say its not working because its a team effort. But just like in relationships, you change one thing, and it could slightly change the dynamics or completely change the dynamics of the team. Bank being the polar opposite of Jose in every way possible, change the dynamic defensively by a large margin. Him fighting through screens and applying pressure, gave everyone else more time to spend guarding there man and more importantly, gave a slow guy defensively like Bargnani to react to whats happening.

I don't think defense is about a guy just sucking ... like Boris said earlier, its about execution. The raptors fail as a whole to execute defensively. Bargnani was terrible at executing at providing his help to his wing players but at the same time, the wing players were terrible at giving him the time by providing resistance to let him react.

You can say Bargnani sucks as a help defender ... or you could say Bargnani needs a little more time then the average help defender to react. Something our wing players couldn't provide him. Compile that with the inability of the wing players to provide any sort of resistance or contain there man and you get the **** defense of all time. I'd be a hypocrite to say that it just rests on the wing players. Bargnani was a terrible help defender so he couldn't erase his wing players mistakes ... in the end it was just one terrible relationship.

[roundhead0]

Did you forget that we allowed team to shoot a high percetages of 3's and a lot of them too ... Bargnani should be out there contesting those SHATS FOLKS!

Like I said, trying to make Bargnani a scapegoat for our defensive lapses, is a joke because there was too many dam wholes in our defense. Our bigs sucked, our wings sucked ... arguing who sucked more is nice but it doesn't change the fact that they ALL SUCKED!

Actually, that IS partly Bargnani's fault. It looks like Triano was trying to mask some of the weakness defensively (including Bargs) by collapsing the defense to prevent penetration ("protecting the house") but it was a colossal failure because individuals could not get enough on the ball pressure or recognize and react to ball movement because they were such crappy defenders, and so it was just open threes all night long.

Once they got away from that stupid defense they eventually simplified things and stopped a lot of the switching as well, and the players were able to be more aggressive and looked much better in the simpler system. But then after the all-star break the coaches seemed to put all the switching back in and defensively things just fell apart.

[OvertimeNO]

I agree with Yogi, this thread has taken a turn for the impressive, especially when Boris chimed in.

I know I've been hard on SS and others for using the stats posted in the OP as near-unassailable evidence to support their conclusions. I just want to make clear that I don't question the veracity of the data they've referenced - only the amount of influence observational bias has on the interpretations presented here. As others have pointed out, and as Boris demonstrated, the wealth of quantifiable data available for tracking in every single play - let alone game - is mind-boggling. It's bad science to make any sort of conclusion using just the formula of your choice without first analyzing every possible source of data. And it's terrible, terrible science to have a "conclusion" in mind from the beginning, and then cherry-picking data to "support" it.

I suppose it wrecks my e-cred to have an open mind about things, but whatever. FYI, I'll confess my biases now:

Do I think Bargnani is a good defender? I think he's passable man-to-man situations, and in certain matchups, surprisingly good. That being said, he certainly does look lost out there sometimes too, and he often loses focus. But he leads the team in blocked shots, which sort of runs counter to the "he's terrible at everything" opinion some stand behind. And while I don't think his decent man defence makes up for his inadequacy on help defence, I also don't think that any big in the league could've benefited statistically from exposure to the season-long layup line that our perimeter guys hosted.

Here's a somewhat-related question I have you stat-crunchers, actually: last season, Bargnani played a significant chunk of minutes with Bosh on the bench, presumably because a Jonhnson-Evans front court is offensive suicide. So is there a decrease in offensive efficiency? Is there an increase in turnovers? Is there an increase in the opponent's defensive rebounding and fast break points?

[BorisDK1]

For the Synergy thing, they do track every play, but there's a number of plays that are assigned to team defence instead of individual defence. Things like cuts are very hard to assign blame to individual players, loose balls ending in a bucket where the entire defence is out of shape, zone defence where the shooter ends up having no one guarding him, transition baskets, etc. I'm not sure how you were able to assign blame/credit with 100% certainty with your system.

Cuts are exceptionally easy to assign blame to. Maybe some people struggle just watching the basketball and the guy guarding it, but I've refereed and coached the game at sufficiently high levels so I'm used to watching the stuff away from the basketball more than otherwise. When a guy gets beat on a cut, it's usually pretty obvious who's responsible.

There are certainly times where I cannot assign individual blame. In that case I use the category "team" (meaning, nobody in particular) to credit with field goals allowed. I also do that in help-and-recover situations where the defense is obviously surrendering something (usually a jump shot) in order to focus on something else (say, a double-team to the post). When the player doubling down is asked to recover out to the perimeter from the post, if there is no rotation on the perimeter to that player, the result goes to "team" (read: "thin air"). Offensive rebounding is ascribed to team, sometimes, as are some (but not all) transition scores (and some stops). About 8% of the team's possessions this year were attributed to "team". I think Synergy had far much more variance than I did.

I'm impressed with your knowledge of basketball plays, and I'm confident that you know your stuff, but again, on some plays, there is no way for you or anyone else to assign individual blame. It goes back to team defence breaking down.

On some plays? Sure. On the vast majority? Fairly straightforward. And when a team defense does break down and there are rotations, I get those. Sometimes there is a complete breakdown and blame cannot be assigned, but that is not the case in about 92% of the situations and in the vast majority of the cases of the other 8% it's clear that no single individual or two individuals can be blamed. There's usually one play per game that really make me pause-rewind-replay-pause-rewind-replay on the old PVR and end up with a result that's only partially satisfying. At the end of the day, there is about as much subjectivity and difficulty scoring these games with the Project Defensive Score Sheet methodology as there is for NBA scorers attempting to assign assists (which nobody has a problem with accepting as gospel, despite the subjectivity in that regard).

As Ripp has said, basketball is a fluid game and there's really few statistical data available to measure a player's full impact on the court. I can surmise from using 4 years of data and 4000+ minutes of play time that Bargnani has a negative overall impact on team defence. I mean when, year-over-year, team defence suddenly improves when he's off the court, it sure means one thing and one thing only. If it happened in only one year, then you have a point. As to judging one player at a time, I've tried isolating different players with and without Bargnani on the court, and in ALL cases, the player I've tried to isolate ended up having a better influence on team defence when Bargnani is off the court. It's weird. Try explaining that.

But there are variables in the on/off court data that you can't account for. Against whom are those minutes on, and against whom are those minutes off? You don't know. 82games.com doesn't know. How do you account for the fact that at the end of his injury, Jose's numbers were almost 12 pts/100 poss worse and Bargnani was (IIRC) either a neutral or slightly positive factor in that metric? Jose's numbers looked better after playing extended time against bench players. You're ignoring the blind spot there and pretending that the causal link is clear: it's not. And, mostly, Bargnani played against first units, more or less fresh.

I'm not pretending that Bargnani isn't a big concern: he clearly is. But there is an information gap here that I think should leave on/off court data in "suggestive" as far as "definitive" goes. Ultimately, you're measuring peripheries as opposed to direct contribution anyway, and I think trying to use peripheral measures as a trump over direct measures is not where it's at.

First, let me say that I agree that Jose Calderon is a very poor defender. But, as I've said on numerous occasions before, when Jose gets beat at the point of attack, someone usually has to rotate to stop the drive. When it's Bargnani as the last line of defence, it's an almost-guaranteed layup. When it's Johnson, Bosh, Humphries, Rasho, Reggie, O'Neal, etc., there's at least some resistance, and the basket isn't guaranteed anymore. As I've shown you earlier, when Calderon isn't on the court with Bargnani (4000+ minutes worth of data, over a 4-year period), the team DRTG is 104.67. But when he plays with Bargnani, the team DRTG jumps to 113.72. It's mind-boggling. Pick a player, any player, and I can run the numbers for you with/without Bargnani, and it'll be the same result, every time.

You know, I think people's expectations of help defense are mostly inflated anyway, and I really don't think people appreciate the importance of guarding the basketball in defensive basketball. Dean Oliver wrote in Basketball on Paper that the impact of a good defensive big man seems to be greater than the impact of a good defensive wing player, but people have missed his more cautious tone in that statement and taken it to a place where it was not warranted. The entire foundation of any team defense isn't the post players, it's the ability to pressure the basketball without getting beaten. And, more specifically, to not get beaten in a blow-by while pressuring the ball. If one doesn't appreciate all that goes wrong when there is both no pressure on the ball and xballhandler is allowing a straight-line drive, then one will be left with massive difficulties interpreting the game of basketball. There is no such thing as effective help on a straight-line drive to the rim: the only thing you can do is foul, or choose to give up an open layup or jump shot (and probably a three). Those are your only three options.

I understand what the stats say. I'm simply arguing that to blame Bargnani exclusively for the team's defense, or to somehow believe that a team headlined by Jose and Bosh would end up with a DRat of around 104.7 (which you seem to be suggesting, showing that your use of the data isn't quite sound) is taking a few leaps. If only Bargnani were the cause as you are suggesting, why was the team's defense in the bottom 20 of the league last year with Bargnani coming off the bench and playing fewer minutes? Why is there no recognition of the fact that the starting lineup was doomed to be a cluster defensively anyway, by violating the John Wooden Rule (the team with a speed and quickness advantage at at least 3 positions is going to win, in the majority of cases)? The on/off court data, again, isn't telling you against whom these teams are playing, it isn't telling you that the majority of Bargnani's minutes come alongside both Turkoglu and Calderon this year. There's an epistemological leap you have to make, here, and the data just isn't there to support it. And apparently the fact that Jose has been playing against bench players for a good bulk of the past 4 years has been ignored.

Now, if you want to suggest that Bargnani is a poor defender and needs to be moved if we want to be a great defensive team, I don't contest that (although I want to see what happens this year, first)

Now that this is out of the way, let's look at the stats you posted.

The chunks of games that you used include Bargnani playing 35+ minutes, so that means that the team's DRTG is already inflated with or without Calderon. What you need to do is isolate court time in this fashion: Calderon/Bargnani and Calderon/No Bargnani. But then, you run into small sample size, and the whole exercise becines moot. Which brings me to the 2-game Milwaukee section you used (Bargnani injured). Do you truly believe that 96 minutes of data is a worthy chunk of data when I've been using 4000+?

I don't need to do that: I'm presenting data that tell a different story. You want to keep the focus only on Bargnani, I'm presenting data that shows something different that defends my thesis.

I've admitted that the Milwaukee games are a small data size. The entire month Jose was off was not, though. The Raptors were about average without Jose and with Bargnani. And, at that time in the season, Andrea's on/off court defensive numbers looked acceptable. They fell precipitously at the end of the season, as he was particularly poor after the All-Star break, which PDSS data also shows. In fact, in that month of Jose's injuries, Andrea ended up with a stop% of .561 with a DPoss% of .191, giving him a DRat of 102.5 in those ~460 minutes he played. That's...quite decent. Why did the defense get that much better when Jose wasn't playing at all? Why did Andrea's individual defense improve when Jose didn't play? And DeMar's (.561, .181, 104.0) did as well. Odd.

Finally, you'll notice that I haven't put last year's disastrous defence entirely on Bargnani's shoulder. There were too many holes to fix on the team. For instance, you can't keep bailing out Jose with help defence. And some of the decisions he makes are just baffling: why are you guarding Rondo chest to chest? Why are you laying off Jameer Nelson? Similar questions should be asked about DeRozan and Turkoglu. It was just a bad combination overall. But, what I did do, was - I hope - prove that Bargnani has a negative overall influence on team defence. Why else does the team defence improve by 8+ points when he's not playing?

I guess we just disagree on just how bad is Bargnani. I think he's egregiously bad, you think he's just bad. Different shades of pale.

I think there's far more problems with Jose's defense than individual choices with certain players. The fact that he may have shown the occasional instant of stepping into Rondo at the midcourt line (only to immediately drop back 6' and abandon pressure after one turn) doesn't really change the fact that he cannot put pressure on the ball. When there is no pressure, your defense is going to get picked apart. It doesn't matter who your big men are: you're toast, IMO.

I agree with your overall assessment of Bargnani, that he's not a good defender. But because he can guard the perimeter on a switch very well for a big man and guard the post extremely well one-on-one, he's not an entire disaster individually. I'm a big believer in speed and quickness, I think it's the primary physical attribute in basketball that leads to winning, and to put so many slow players on the court together was just going to be a gong show. Even DeRozan, wonderful leaper though he is, isn't particularly quick (and sure didn't test like it at the combine). I'm guessing that will improve with time in the weight room, which he doesn't seem to be neglecting (thankfully). Our disagreement is primary cause of this team, which I pin to be on Calderon. All in all, without Calderon, I think this team would have ended up being around a 109 DRat team.

Anyway, to ripp (?) who asked, I tried uploading the spreadsheet to Google docs, and it was rejected due to size (it's ~ 18 MB). So I made it available for download here. You will need Microsoft Excel, or something that can read .xls files.