MAVS 108, PELS 92

[fifteenth]

@"Kammrath" I'm familiar with your thoughts about +/-, after all, you and I have visited and revisited this topic for a decade! 
I just can't recall a discussion about what exactly a coach sees when they get reports from the analytics guys. We've heard talk in the past about metrics developed for teams. I'm curious about whether the coaches just look at the output of the spreadsheets or also the raw data itself. Don't know if I'll get an answer to that, but I'm curious.

[mavsluvr]

It tells you how the TEAM (which includes that player) played/performed when that player was on the court.

No, it doesn't. It tells you what the score differential was during any particular time period. It tells you absolutely nothing about how the team performed during those periods. Doesn't tell you whether they shot well or poorly; whether they were active on the glass or not; not whether they were encumbered by injuries; not whether the officiating was of good or bad quality; not the quality of the opponents on the floor during any time segment. If the team as a whole has a +/- of -5 (meaning they lost the game by five points), it tells you nothing about whether they played well or poorly. If they have a +5, it doesn't tell you anything about whether they played well or poorly. It simply gives you a point differential. 

That is the fact it presents.

The fact it presents is the point differential -- NOT how any particular lineup played, or how well they played. 

For instance, The Mavs have been severely outscored this year when Luka has played. They have on the other hand the Mavs have handily outscored their opponents when Luka hasn't played. That is the fact which +/- is telling us and it won't and cannot "lie."

It actually doesn't tell us that. There have been times when Luka was on the floor when the point differential was positive. There have been other times when it was negative. In the cumulative total over ten games, it is apparently negative. That doesn't tell you that Luka has played worse than the guys who played without him. It doesn't tell you anything about how he played, or how any of the other nine players who were on the court when he was played. The performance of  the various groups of ten on the court when he wasn't there tells us nothing about whether Luka played the right way, the wrong way, was showing progress, was a total dud, or anything of the sort. It certainly doesn't tell us that the players with better +/- numbers played better than he did. 

I think that is where we may be attempting to make a leap that is not logically valid. That stat tells us a factual piece of information -- what the point differential was. We cannot logically extrapolate anything at all about the players' performance from that number. 

It is what it is. Our conclusions beyond that are 100% debatable, but with larger sample sizes our confidence in reaching more conclusions will get higher. 

But we also have lots of other numbers and observations to add to what +/- is telling us. For instance we see Luka shooting 29.5% from three. We see the team as a whole shooting 26.7% on OPEN threes. So for instance we can safely surmise that Luka's -2.5 offensive on/off is partly (mostly?) from Luka shooting poorly himself AND from his teammates missing so many of the open looks he has gotten them.

We can tell what percentage of shots Luka is making from watching him and from looking at the field goal percentage stats. We cannot tell anything about how well he or his teammates are shooting or from looking at his +/- stat. It is possible to shoot very well, score a lot of points, and still have a negative +/-, and vice versa. 

The stat tells us the truth about point differentials, as you point out. It tells us nothing else. Whatever surmises we make based on it are matters of interpreting that data. We may be able to use it as one of a number of tools in our evaluation box. But it's not some Delphi-like oracular stat that constitutes some kind of infallible information about how a person, or a lineup, or a team, plays.

[mvossman]

I could be wrong, but I would bet a lot of money that Kidd thinks that +/- lineup data is in fact reliable and gives useful information that he takes into account as a coach. I cannot imagine any modern coach refusing to look at that data or thinking it doesn't have something to offer. Again, I bet Kidd was protecting his players and also pushing back against knee-jerk interpretations of such numbers. 

One thing I feel like I can be quite certain of: the analytics department guys for the Mavs are not blowing off Luka's +/- numbers so far. I bet they are digging into those numbers trying to make sense of what is going on. Nothing may need to change and things may "correct themselves," but if you care about the team then you don't just watch the Mavs getting outscored by a rate of 10.6 pts when Luka plays and not pause with concern and curiosity.

As Mavslvr pointed out, I think you are looking at +/- data differently than teams generally do.  I don't think they generally look at it to assess an individuals impact on the court, especially over a small sample size.  I think teams are more interested in lineups, and what combinations work and don't work.  There is still a lot of noise in small sample lineup +/- due to who they played against and the general high variance of 3 point shooting on both sides.  I would not be surprised if they use something more advanced that looks at things like open or wide open threes and derives an estimated +/- on small samples.

[Kammrath]

It tells you absolutely nothing about how the team performed during those periods.

So the NBA has ONE statistic that they have tracked their entire existence: points scored. 

It is the only stat that the NBA makes its decisions on as it compares the points scored of each team every game and gives a win to the team that had the positive differential and a loss to the team that had the negative differential. The NBA has chosen to say that the team that scores more points than their opponent has won and therefore out-performed their opponent. So when I say how a team "performs" that's what I mean. It is the only performance the NBA cares about. NBA has never given a team a playoff spot for shooting a better percentage or having less turnovers, only for outscoring their opponent over the course of a certain number of games. 

Therefore +/- absolutely tells you how a team is "performing" in the only way the NBA actually cares about at the end of the day. Because +/- tells you if you have outscored your opponent or not: did you win or not. Last night the Mavs were +16 as a team, therefore the NBA gives them one of those treasured wins.

[Kammrath]

That stat tells us a factual piece of information -- what the point differential was. We cannot logically extrapolate anything at all about the players' performance from that number. 

Like I have repeatedly said: It tells us about the TEAM's performance while that player has played. Is the team performing well (again defined by "outscoring the opponent," the only objective in basketball that matters) over the course of a single game or many games while Luka has played? 

Last night? The team was outscored by 2 points when Luka played. The team outscored NOP by 18 when Luka sat. 

For the season? The team has been outscored by a rate of 10.6 pts (per 100 possessions) when Luka has played. The team has outscored their opponents by rate of 18.3 pts while Luka has sat. 

And again, Luka is PART of the ten players on the court when he plays, so he is PART of what is happening. How much is on him is totally debatable but he is part of the equation and therefore bears some of the responsibility.

[fifteenth]

Is the disagreement here about symantics, or no?

ML's point, if I can be so bold as to speculate on that, is that a team can get outscored while playing well, and can outscore it's oponent while playing poorly. Therefore, the numbers only tell you who scored the most, not how they performed. 

It looks like Kam is possibly saying that "outscoring your oponnent", or not, suggests something about your performance often enough that it should be looked at, even though it doesn't give you the details about why you outscored or got outscored. 

Is there a place to land between those two takes? Or is it a "never the twain shall meet" situation?

[mavsluvr]

So the NBA has ONE statistic that they have tracked their entire existence: points scored. 

It is the only stat that the NBA makes its decisions on as it compares the points scored of each team every game and gives a win to the team that had the positive differential and a loss to the team that had the negative differential. The NBA has chosen to say that the team that scores more points than their opponent has won and therefore out-performed their opponent. So when I say how a team "performs" that's what I mean. It is the only performance the NBA cares about. NBA has never given a team a playoff spot for shooting a better percentage or having less turnovers, only for outscoring their opponent over the course of a certain number of games. 

Therefore +/- absolutely tells you how a team is "performing" in the only way the NBA actually cares about at the end of the day. Because +/- tells you if you have outscored your opponent or not: did you win or not. Last night the Mavs were +16 as a team, therefore the NBA gives them one of those treasured wins.

Maybe we are getting down to the ambiguities in the use of language. 

I understand you to be saying that, when you use the word "performance," what you really mean is "point differential." I believe the word "performance," as used by most sports analysts, covers a wide variety of aspects of play, generally adding up to whether the player executed well or poorly in the various elements comprising a game. 

I can assure you that point differential is not the only stat the league makes decisions on or cares about. The NBA awards the team scoring the most points a win, but does not purport to declare that the winner has outperformed the loser in any other respect. 

If you are indeed using performance to equal "point differential," then use of +/- as a measure is simply stating a truism. It then only means that whichever group has the highest point differential has scored more than their opponent for the segment in question. Therefore, it can never be "wrong," because all it is doing is defining itself. 

I do not think these terms mean what you think they mean.

[dirkfansince1998]

Would love to see what @"dirkfansince1998" thinks about raw +/- and how useful it is in the hands of us message board folk, or in the hands of a coach. 

I think the following is a legit quesiton, but am willing to be coached up if I'm wrong: Do coaches look at raw +/- or do they wait for the analytics department to gather sufficient data, process the data, and issue reports to the coach and others?

Very high on all kinds of lineup, raw +/- data or some of the +/- based advanced metrics.
Agree with @"Kammrath". Just need to know what we are looking at. In case of +/-. The scoring margin when a player was on the floor. The important part for me. Sample size. +/- comes with a lot of noice and I don´t think single game +/- is telling us a lot. But in most cases the numbers normalize as the season progresses. Luka´s numbers would be concerning if they don´t improve over the season. But his +/- was just as bad in december 2020. Same for Powell. And both managed to turn the numbers around.

In bigger sample sizes they are showing clear trends. As @"Kammrath" told us all year long. Richardson was the weak spot in the lineup. +/- confirmed it. Not sure if the coaching staff needed the numbers to make that decision but they adjusted the lineup prior to the playoffs and benched Richardson.

I don´t really think any stat can be good/bad (well I hate PER). It always comes down to the interpretation. What exactly is the +/- data telling us. To use a recent example from pre Luka Mavs seasons. For the most part the starters used to have terrible +/- numbers. Bench players on the other hand looked great. Doesn´t mean that the mentioned bench players are better than the starters. Just that they performed better in their given role. Would need to test those players in a different role to find out.

I think that is exactly what Kidd is doing right now. Kidd/staff might have a prefered starting five but it is hard to keep track of all the different lineups Kidd is trying right now. We are seeing so many different combinations in small sample sizes that it is really hard to come up with a trend. My only halfway confident opinions right now. Single big lineups are outperforming twin tower lineups. Defense is a lot better when Luka is on the bench. I guess we could argue about eyetest vs confirmation bias but in both cases the +/- numbers match the eyetest.


Side note: Talking about more advanced +/- metrics. As far as I know Jeremias Engelmann (creator of ESPNs RPM, before they changed the prior/formula multiple times) is still one of the senior analysts. My best guess is that he is doing similar things for the Mavs.

[Kammrath]

Is the disagreement here about symantics, or no?

Yes absolutely to some extent. I think it is why my position has been consistently misrepresented (I am assuming it is not trolling). 

For me I am a wins/losses guy (point differential guy). It is really the only metric I really care about. I love following Luka and triple doubles as a side thing, but at the end of the day I want a player who impacts winning. For instance, I would rather have Caruso than Westbrook because of that. Westbrook can "outperform" Caruso in a lot of ways, but Caruso will consistently kick Westbrook's butt in impact on winning for his team on both ends. 

Now that doesn't mean other stats and measures don't matter....because they DO and they can be a helpful look under the hood. But at the end of the day point differential stuff tells me about the total state of the car and that is what the NBA cares most about. 



One other historical point: 

When +/- first became a thing it was too quickly hailed as the "Holy Grail" of stats and a LOT of clumsy and hasty misinterpretations ensued. People then started pushing back against that. However, I think many threw out the baby with the bathwater when they pushed back and started acting like it doesn't tell us anything. But the reality is that +/- is indeed the Holy Grail of stats, but that Holy Grail is buried within the context and finding it is a fleeting and contingent thing. You have to keep digging and exploring and examining to get the goods from +/- and it will never tell you anything "for certain" but it will lead you to many conclusions you can count on. So I 100% get the hesitancy with +/- numbers, but that hesitancy should cause us to treat the +/- numbers with more respect and time, not just throw them out or dismiss them. There is a reason that basically all the best advanced metrics incorporate +/- in such a large part...they know how important it is.

[Arioch]

I could be wrong, but I would bet a lot of money that Kidd thinks that +/- lineup data is in fact reliable and gives useful information that he takes into account as a coach. I cannot imagine any modern coach refusing to look at that data or thinking it doesn't have something to offer. Again, I bet Kidd was protecting his players and also pushing back against knee-jerk interpretations of such numbers. 

Mavsluvr's summary says Kidd said "he doesn’t look at stat sheets on the bench". I don't know what the actual quote was, but just based on that summary, Kidd didn't completely dismiss +/- or say he never looked at it. He said he doesn't look at it 'in-game' which is a good thing, and I expect was true of Carlisle as well. Small sample-size is the bane of any statistic and just looking at the last 12 minutes of +/- would be the very definition of a tiny sample size creating an extremely noisy (and useless) stat. +/- only becomes truly useful when you are looking at it across multiple games--something the coaching staff probably does during prep sessions well away from the actual games.

[Kammrath]

I believe the word "performance," as used by most sports analysts, covers a wide variety of aspects of play, generally adding up to whether the player executed well or poorly in the various elements comprising a game.

Of course people will use performance in a lot of ways in basketball. 

Do you really not believe that the ultimate "performance" metric in basketball is winning?

[fifteenth]

I think it is why my position has been consistently misrepresented (I am assuming it is not trolling). 

Take this comment or leave it, but you might find it helpful. When you say ML is misrperesenting you, I often see him as just repeating back to you what he thinks you're saying. 

Communicaiton is hard. All of us married folks should be able to affirm that!  :-)

[Kammrath]

Take this comment or leave it, but you might find it helpful. When you say ML is misrperesenting you, I often see him as just repeating back to you what he thinks you're saying. 

Communicaiton is hard. All of us married folks should be able to affirm that!  :-)

It IS hard. And yes marriage highlights that in spades!!

Right, I think he is trying to understand my position, but I think in my case fails pretty consistently (especially on this topic). I don't just assume that is on him, so I am trying my best to reword stuff. 

On the other hand, I think you have consistently understood my position on this as has @"dirkfansince1998". Even if y'all disagree I feel like you see my position for what it actually is.

P.S. And on here, just like in my marriage, I am not trying to convince anyone to adopt my view points, but I do want to be understood so that when you reject my viewpoints you are rejecting my ACTUAL position and not a misunderstood one.

[fifteenth]

Of course people will use performance in a lot of ways in basketball. 

Do you really not believe that the ultimate "performance" metric in basketball is winning?

This is a great question! I won't answer because then I'd have to admit to being Mark Followill. 

Here's a question back to you: If ALL you know is that a team got outscored for 48 minutes, do you konw anything about their performance?

And @"mavsluvr" , if you have your ears on, if a team got outscored for 48 X 82 minutes, would that say anything about their performance, even if the thing it says is very general?

[mavsluvr]

Of course people will use performance in a lot of ways in basketball. 

Do you really not believe that the ultimate "performance" metric in basketball is winning?

I think "performance" is almost always used as a more generalized concept than "point differential," and that when people mean only "point differential" or "wins and losses," that's what they say. 

No, I don't. Winning can be affected by many things not having to do with performance -- injuries, scheduling, conditions, poor officiating, heck, unlucky bounces. 

Winning also does not equate to point differentials except on a one-game basis. If point differentials were all that mattered, we would determine titles and playoff seedings by point differentials, not win-loss record. We could also determine the best players in the season, 1-450, by sorting them by +/-. 

There are stars who end up with poor +/-s, because they play on bad teams. Yet, good teams then hire them because of their excellent overall performance. They don't dismiss them as players who can't impact winning because of their +/-, or because their team didn't have a winning record. 

+/- is one of a number of available statistical tools. Statistical tools help evaluate performance. They don't define performance. +/- is not a number that tells you anything about how, or how well, someone plays. Even a stat that does tell you something about how the player's performance (FT%, for example) only gives you a small piece of the story. 

If stats told us everything, we wouldn't even need to watch the games. We could just look at the stat sheet and know everything about what happened. There's a reason they play the games.

[burekemde]

+/- is a real number of course, but interpreting it, is where you can go wrong. You can also interpret it correctly and then it's valuable. How to interpret it correctly? One way is to watch the plays and let them describe the reasons behind the +/-. Another one is to use multitude of other stats and combine those together with the +/-, in order to interpret a players performance and contribution. But also this in the end involves watching the plays and judging if the combined statistical interpretation is correct.

Bottom line is this, each stat or combination of stats needs to be verified by watching.

There is no way going around by watching and evaluation. But stat can help.

[fifteenth]

Winning also does not equate to point differentials except on a one-game basis.

Above is the mavsluvr way to state the sentence. 

Here's Kam's: Point differentials equate to winning on a one-game basis.

[haveitall]

Above is the mavsluvr way to state the sentence. 

Here's Kam's: Point differentials equate to winning on a one-game basis.

+- aside I think everyone can agree that the Mav's haven't played perfect games.  Yet they are 7-3.  Luka looks really fat, and he isn't scoring 40 points nor playing 48mpg, yet the team is 7-3.  The point is that this team hasn't played its best basketball, the shots aren't falling, yet they are winning.  Their ceiling is much higher than any recent season.  The reason for that is better defense and that deep bench with Bullock, Frank and JB playing awesome basketball.

[KillerLeft]

Their ceiling is much higher than any recent season.  The reason for that is that deep bench with Bullock, Frank and JB playing awesome basketball.

That's a factor, sure. I'd say that if they get the KP from the last two games for, idk, 60 games, including the playoffs, that would raise their ceiling more than anything. I'm still skeptical about the likelihood of that, but at least I'm believing in the possibility of it again.

[mavsluvr]

This is a great question! I won't answer because then I'd have to admit to being Mark Followill. 

Here's a question back to you: If ALL you know is that a team got outscored for 48 minutes, do you konw anything about their performance?

And @"mavsluvr" , if you have your ears on, if a team got outscored for 48 X 82 minutes, would that say anything about their performance, even if the thing it says is very general?

Are you talking about losing every game, or accumulating a negative point differential over a season? 

If we are talking about losing every game, sure. I would think they were either tanking, which involves intentionally performing badly, or just performing badly enough as a group to lose every game. 

If we are talking about having a negative point differential over a season, that wouldn't necessarily make me think they are a bad team. You can have a positive win-loss record with a negative point differential. (Look at the Mavs' ten-game stats, for example.) Teams with a negative point differential have made the playoffs. Or, for example, they could be a well-oiled machine that had two of their stars incur season-ending injuries fifteen games before the season ends. 

Also, performance, in its usual sense, takes account of expectations. A team that came up from last in the league to having a -1 total point differential over the season might be considered to have performed very well, and vice versa.  

However, in either case I would not at all think that a losing record or a negative point differential meant that every player on the team played badly. It is pretty common for young bad teams to have outstanding young players, and no one dismisses them because of their team's season-long point differential or record.

Sports inherently involve an element of randomness, which means that the better team, or even the team that plays better that night, doesn't always win the game. It's part of what makes sports exciting. If we wanted to see a competition where randomness isn't a factor, we could watch a chess game. 

I have never taken the position that +/- is a worthless statistical tool. In fact, I think it is a very valuable tool when it is used properly. I also think it can be misleading, and it is easy for a fan to think it proves more than it does. I think it can be related to performance, but doesn't tell you anything specific about performance, unless you simply define performance to mean having a good +/-.