Minnesota Sports Forum

Q-was-here wrote:

Camden wrote:D'Angelo Russell had 13.0 potential assists last night, according to NBA.com, which was tied for fourth-best around the league last night out of the 14 teams that played. It sure seemed to me like he was moving the ball, creating shots for himself and his teammates, and playing with good energy overall forcing turnovers and making the right play in general. But hey, none of that matters if the plus-minus says otherwise -- despite it essentially being a lineup stat with moving parts that individual players can't always affect. But hey, we see things differently. There's no point in trying to go any deeper now.

The problem Cam is that the guy has been a negative in that stat over the course hundreds of games and thousands of minutes and this season has been no different. Last year was the one time in his career he had a positive net rating and net on/off rating. At some point, shouldn't the player get some blame if after 8 seasons his Net rating is negative AND his Net On/Off rating is negative? At that point, "context" doesn't matter because so many different contexts are applicable. If all the other variables are changing around you, but you still are a negative, then that seems to point to the one constant! But unfortunately it's a point you'll never concede when it comes to DLO.

Sure, however, that ignores the other stats that paint him in a more positive light, which is why I've always argued that the assessment of any player is about much more than any one stat -- if I have a handful of stats that say this player is good, I won't ignore those because this other particular stat(s) says he's a negative. And that's exactly what happens here among a few constant posters.

The larger issue I have with traditional plus-minus and on/off stats is that if we realize they're fundamentally flawed or misleading at the single-game level, due to the lack of context (which always matters), then why is a bigger sample of all these same single-game values any more meaningful? Collecting more flawed data doesn't make the data any less flawed. It just means you have more of it.

You and others hail plus-minus when making your assessments. That's fine. Others like VORP, BPM, RAPTOR, and whatever else is available that matches what their eyes tell them as well as more traditional numbers. And that's also fine. I've long said that the best analysis involves all of them, though. I don't see how that can be argued. But somehow we always come back to the plus-minus talks around here as if it's THE stat, and it's just not -- unless we're talking about the performance of a specific lineup combination.

Camden wrote:

Q-was-here wrote:

Camden wrote:D'Angelo Russell had 13.0 potential assists last night, according to NBA.com, which was tied for fourth-best around the league last night out of the 14 teams that played. It sure seemed to me like he was moving the ball, creating shots for himself and his teammates, and playing with good energy overall forcing turnovers and making the right play in general. But hey, none of that matters if the plus-minus says otherwise -- despite it essentially being a lineup stat with moving parts that individual players can't always affect. But hey, we see things differently. There's no point in trying to go any deeper now.

The problem Cam is that the guy has been a negative in that stat over the course hundreds of games and thousands of minutes and this season has been no different. Last year was the one time in his career he had a positive net rating and net on/off rating. At some point, shouldn't the player get some blame if after 8 seasons his Net rating is negative AND his Net On/Off rating is negative? At that point, "context" doesn't matter because so many different contexts are applicable. If all the other variables are changing around you, but you still are a negative, then that seems to point to the one constant! But unfortunately it's a point you'll never concede when it comes to DLO.

Sure, however, that ignores the other stats that paint him in a more positive light, which is why I've always argued that the assessment of any player is about much more than any one stat -- if I have a handful of stats that say this player is good, I won't ignore those because this other particular stat(s) says he's a negative. And that's exactly what happens here among a few constant posters.

The larger issue I have with traditional plus-minus and on/off stats is that if we realize they're fundamentally flawed or misleading at the single-game level, due to the lack of context (which always matters), then why is a bigger sample of all these same single-game values any more meaningful? Collecting more flawed data doesn't make the data any less flawed. It just means you have more of it.

You and others hail plus-minus when making your assessments. That's fine. Others like VORP, BPM, RAPTOR, and whatever else is available that matches what their eyes tell them as well as more traditional numbers. And that's also fine. I've long said that the best analysis involves all of them, though. I don't see how that can be argued. But somehow we always come back to the plus-minus talks around here as if it's THE stat, and it's just not -- unless we're talking about the performance of a specific lineup combination.

Cam, like any statistical analysis, sample size matters. The reason why plus/minus is more meaningful with a bigger sample is for the same reason other stats require more than one game. It smooths the random variation inherent with any result that includes multiple, complex variables. So when DLO shoots 3-11 in a single game, I don't declare him suddenly a sub-30% shooter. Yet if he were to shoot sub-30% for an entire 82-game season, I think most of us would agree he is not a good shooter. Period.

I will agree that Plus/minus and On/Off requires a larger sample size than other stats that are more individual, like rebounding, shooting percentage, etc. With DLO, we now have almost 8 yearsof data! I think we've reached the minimum threshold of sample size at this point!

And On/Off is not my "preferred" stat. It's one thing I look at among many. But the problem is when it says the same thing pretty much every year, I can't just toss it aside as random noise like you do. Something is going on. And we know those other stats don't account all that well for defensive impact.

I think DLO has some very above average NBA skills (his passing is elite), but his weaknesses make it super hard to find the right combination of guys to play with him. Last year we managed to find the magic combination both in terms of personnel and scheme. It was good while it lasted.....

Russell had a very solid game overall. He showed more fight than normal defensively on the ball. And he shot very well (10 - 16 fg). He was instrumental in that game being close. Granted, Russell had THREE inexplicably bad unforced TOs during the big Thunder run that simply made the uphill battle that much more difficult. They came out of nowhere so it clouded what was a really good game for him.

But the Thunder shot 15 - 24 fg in the 4th quarter which hurt the most. That's not on Russell, no matter what his +/- says.

The Timberwolves should be ecstatic to get THAT Russell every game. Turns out, they get him only once out of every 5 or 6 games.



[Note: The loss is on Gobert. His ejection was stupid... and deserved. The loss is on him. Good thing we're good fans and probably won't boo him for directly costing the Wolves a win.]

Of course sample size matters, no matter the statistic. I wouldn't disagree with that, and that also wasn't my argument. However, unlike shooting efficiency, which is a simple make or miss, traditional plus-minus has a number of variables that cannot be controlled by the individual -- no matter who it is. That's where that comparison absolutely falls flat for me.

Traditional plus-minus -- in the manner that it's often referenced here -- is the entire lineup's performance during which said player is on the court. That individual's score is affected by not just his performance, but his teammate's performances as well as the opposition's. There's a lot of moving parts involved that create variance, which in turn make it noisy at best and unreliable at worst when used to grade a single player.

I think most people would agree that single-game plus-minus values can be misleading. They don't always reflect the true impact an individual player made in the game -- good, bad, or neutral. With that understanding in mind, and knowing that it's a stat that absolutely lacks context for an individual, why would we continue to value it highly by simply having more of the same flawed data? The large sample size is essentially made up of these smaller sample sizes that can't always be trusted at face value, which means the large sample size shouldn't be either, at least on its own. That's my point. That's what my argument is -- not that sample sizes don't matter, respectfully.

I also trust my eyes being a fan and follower of the game over a long period of time. That's not to say that I think I'm better at evaluating, or whatever. I just think most people would agree that what they see comes first when evaluating a player, team, game, etc. First and foremost, my eyes tell me when a player is making a positive, negative, or neutral impact in a game -- not plus-minus values. They may confirm it or they may not, but my eyes come first.

Specifically for last night's game, which is what this discussion originated from, I do not see how anyone watched that game -- on both ends of the floor -- and came away from it thinking, "Yep, D'Angelo Russell played poorly. Minnesota was much worse with him on the floor." Sorry, but no. We will not see eye to eye on that, and I would not be able to agree.

Q-was-here wrote:

Camden wrote:

Q-was-here wrote:

Camden wrote:D'Angelo Russell had 13.0 potential assists last night, according to NBA.com, which was tied for fourth-best around the league last night out of the 14 teams that played. It sure seemed to me like he was moving the ball, creating shots for himself and his teammates, and playing with good energy overall forcing turnovers and making the right play in general. But hey, none of that matters if the plus-minus says otherwise -- despite it essentially being a lineup stat with moving parts that individual players can't always affect. But hey, we see things differently. There's no point in trying to go any deeper now.

The problem Cam is that the guy has been a negative in that stat over the course hundreds of games and thousands of minutes and this season has been no different. Last year was the one time in his career he had a positive net rating and net on/off rating. At some point, shouldn't the player get some blame if after 8 seasons his Net rating is negative AND his Net On/Off rating is negative? At that point, "context" doesn't matter because so many different contexts are applicable. If all the other variables are changing around you, but you still are a negative, then that seems to point to the one constant! But unfortunately it's a point you'll never concede when it comes to DLO.

Sure, however, that ignores the other stats that paint him in a more positive light, which is why I've always argued that the assessment of any player is about much more than any one stat -- if I have a handful of stats that say this player is good, I won't ignore those because this other particular stat(s) says he's a negative. And that's exactly what happens here among a few constant posters.

The larger issue I have with traditional plus-minus and on/off stats is that if we realize they're fundamentally flawed or misleading at the single-game level, due to the lack of context (which always matters), then why is a bigger sample of all these same single-game values any more meaningful? Collecting more flawed data doesn't make the data any less flawed. It just means you have more of it.

You and others hail plus-minus when making your assessments. That's fine. Others like VORP, BPM, RAPTOR, and whatever else is available that matches what their eyes tell them as well as more traditional numbers. And that's also fine. I've long said that the best analysis involves all of them, though. I don't see how that can be argued. But somehow we always come back to the plus-minus talks around here as if it's THE stat, and it's just not -- unless we're talking about the performance of a specific lineup combination.

Cam, like any statistical analysis, sample size matters. The reason why plus/minus is more meaningful with a bigger sample is for the same reason other stats require more than one game. It smooths the random variation inherent with any result that includes multiple, complex variables. So when DLO shoots 3-11 in a single game, I don't declare him suddenly a sub-30% shooter. Yet if he were to shoot sub-30% for an entire 82-game season, I think most of us would agree he is not a good shooter. Period.

I will agree that Plus/minus and On/Off requires a larger sample size than other stats that are more individual, like rebounding, shooting percentage, etc. With DLO, we now have almost 8 yearsof data! I think we've reached the minimum threshold of sample size at this point!

And On/Off is not my "preferred" stat. It's one thing I look at among many. But the problem is when it says the same thing pretty much every year, I can't just toss it aside as random noise like you do. Something is going on. And we know those other stats don't account all that well for defensive impact.

I think DLO has some very above average NBA skills (his passing is elite), but his weaknesses make it super hard to find the right combination of guys to play with him. Last year we managed to find the magic combination both in terms of personnel and scheme. It was good while it lasted.....

I agree that when you consider a player over many years over many teams playing with a multitude of players the stat takes out the noise.

Also some of the things he does well or should do well are not that important to this team. We have two elite scorers in Ant and KAT (possibly Nowell off the bench is nearly elite at scoring/getting his own shot). We would be much better served with a strong defender at the point who hustles and boards. We don't need another shot jacker who doesn't defend or board. That is the opposite of what would be most useful.

Cam, you are like the democrats. You call people out for doing the exact thing you are guilty of doing. You are the absolute king of finding a single stat and building an argument around it. I almost fell out of my chair reading what you wrote above about "always using more than one stat". I'll let you rant and rave about DLO 95% of the time, but I'm not going to let you use revisionist history when you get backed into a corner. Someone has to keep it real here.

CoolBreeze44 wrote:Cam, you are like the democrats. You call people out for doing the exact thing you are guilty of doing. You are the absolute king of finding a single stat and building an argument around it. I almost fell out of my chair reading what you wrote above about "always using more than one stat". I'll let you rant and rave about DLO 95% of the time, but I'm not going to let you use revisionist history when you get backed into a corner. Someone has to keep it real here.

The bolded text is laughable, especially from you, and we can prove it right here since you're so resolute about it. What single statistic would that have been? When have I ever built an entire assessment around a single statistic? Provide me a real example of what you're accusing. That should be easy to do considering you presented it like it's a reoccurrence.

Or do I typically rely on multiple measures to come to my conclusions? How else would I maintain any of my positions for anything longer than one debate? Think about that for a moment.

You can disagree with me all you'd like, but at least come correct. Someone does have to keep it real and it's rarely you.

Camden wrote:Of course sample size matters, no matter the statistic. I wouldn't disagree with that, and that also wasn't my argument. However, unlike shooting efficiency, which is a simple make or miss, traditional plus-minus has a number of variables that cannot be controlled by the individual -- no matter who it is. That's where that comparison absolutely falls flat for me.

Traditional plus-minus -- in the manner that it's often referenced here -- is the entire lineup's performance during which said player is on the court. That individual's score is affected by not just his performance, but his teammate's performances as well as the opposition's. There's a lot of moving parts involved that create variance, which in turn make it noisy at best and unreliable at worst when used to grade a single player.

I think most people would agree that single-game plus-minus values can be misleading. They don't always reflect the true impact an individual player made in the game -- good, bad, or neutral. With that understanding in mind, and knowing that it's a stat that absolutely lacks context for an individual, why would we continue to value it highly by simply having more of the same flawed data? The large sample size is essentially made up of these smaller sample sizes that can't always be trusted at face value, which means the large sample size shouldn't be either, at least on its own. That's my point. That's what my argument is -- not that sample sizes don't matter, respectfully.

I also trust my eyes being a fan and follower of the game over a long period of time. That's not to say that I think I'm better at evaluating, or whatever. I just think most people would agree that what they see comes first when evaluating a player, team, game, etc. First and foremost, my eyes tell me when a player is making a positive, negative, or neutral impact in a game -- not plus-minus values. They may confirm it or they may not, but my eyes come first.

Specifically for last night's game, which is what this discussion originated from, I do not see how anyone watched that game -- on both ends of the floor -- and came away from it thinking, "Yep, D'Angelo Russell played poorly. Minnesota was much worse with him on the floor." Sorry, but no. We will not see eye to eye on that, and I would not be able to agree.

Cam, I agree it can be taken out of context and that in last night's game it should not be used to indicate DLO's performance. My argument is against your complete dismissal of the stat as it relates to an individual over a large sample size.

If we take what you say about it being totally noise, then in DLO's case, it's an incredible streak of bad luck that for 7 out of 8 seasons on three different teams he's had a) crappy team mates starting next to him (thus the negative net rating - after all, it can't be HIS fault his team always scores less points that the opponents when he's on the floor!), and b) really good backups since, amazingly, his team usually outscores their opponents when he's off the floor.

Bottom line...I predict DLO will be a backup PG or 6th man combo guard with someone within the next 1-3 years. It might even be the Wolves!

Camden wrote:

CoolBreeze44 wrote:Cam, you are like the democrats. You call people out for doing the exact thing you are guilty of doing. You are the absolute king of finding a single stat and building an argument around it. I almost fell out of my chair reading what you wrote above about "always using more than one stat". I'll let you rant and rave about DLO 95% of the time, but I'm not going to let you use revisionist history when you get backed into a corner. Someone has to keep it real here.

The bolded text is laughable, especially from you, and we can prove it right here since you're so resolute about it. What single statistic would that have been? When have I ever built an entire assessment around a single statistic? Provide me a real example of what you're accusing. That should be easy to do considering you presented it like it's a reoccurrence.

Or do I typically rely on multiple measures to come to my conclusions? How else would I maintain any of my positions for anything longer than one debate? Think about that for a moment.

You can disagree with me all you'd like, but at least come correct. Someone does have to keep it real and it's rarely you.

I'm keeping it real because I have had these arguments with you dozens of times. I've called you out on it time and time again. You are famous for finding a stat and having that be your argument. Are you kidding me? You are your own propaganda machine. Just because you say something more often and louder, doesn't make you right. If you want to play the "find me a specific example" card, I am going to call you out every time you say something stupid. I don't generally like to post that much, but if that's what you want, that's what I'm going to do.

Q-was-here wrote:Cam, I agree it can be taken out of context and that in last night's game it should not be used to indicate DLO's performance. My argument is against your complete dismissal of the stat as it relates to an individual over a large sample size.

If we take what you say about it being totally noise, then in DLO's case, it's an incredible streak of bad luck that for 7 out of 8 seasons on three different teams he's had a) crappy team mates starting next to him (thus the negative net rating - after all, it can't be HIS fault his team always scores less points that the opponents when he's on the floor!), and b) really good backups since, amazingly, his team usually outscores their opponents when he's off the floor.

Bottom line...I predict DLO will be a backup PG or 6th man combo guard with someone within the next 1-3 years. It might even be the Wolves!

Q, my stance, as explained above, on plus-minus predates D'Angelo Russell even entering the league. We've often talked about that stat dating back to the days of Ricky Rubio, Kevin Love, and Andrew Wiggins. My feelings on it and how it's used here have largely remained the same. It's definitely been a reoccurring debate here in recent seasons because of Russell, but I would (and have) argue against plus-minus being THE indicator of a player's impact no matter who it is.

I also haven't said that Russell is blameless in anything, but I respect your sarcasm. I just don't think he's the negative impact player he often gets painted as primary because of that stat -- similar to DeMar DeRozan, Zach LaVine, and others. I've explained why here almost to a fault.

Russell played well last night, plus-minus be damned. I think you would agree with that, which only speaks to my feelings on the stat in general. Anyways, on to the next game and the one after that... and the one after that... where we can do this time and time again. Can't wait.