Doctor MJ wrote:Bob8 wrote:Doctor MJ wrote:
So, the big picture thing for me is this:
The traditional box score simply doesn't capture the entirety of what's going on on the court, and so we need another family of stats to point us in the direction of the rest. That family has to be based on the team box score because basketball is a team game.
Now, this is not the same as saying +/- stats are fundamentally superior to the box score. +/- gives us a stat with strong validity to go with the box score's strong reliability and exploitability. As an analyst I use all the data I can, both of these families, plus what I see with my eyes and what great basketball scouts can point to that I miss.
And this is all I really advocate for for everyone else too. It may seem like I advocate only for +/- data, but that's because of the context of this discussion where Luka has a long track record of not-so-impressive +/-. To me the critical thing there is not "So Luka's not very good", but "So what explains this?". And of course you and others feel like you've answered this based on specific oddities of this particular season, but these do not explain the years prior.
Last thing I'll say: If you're looking to completely ignore a family of data simply because it's noisy, you're not acting like a scientist. Trying to infer causation from correlation is what scientists do, and that means grappling with confounding variables without agenda rather than throwing out the data that doesn't align with the data you're most comfortable with.
1. Few days ago I gave my view on Luka's career +/-, (which is btw. better than SGA's), when I compared him and Steph. This year, when Steph +/- dramatically dipped, and their prior years. You didn't have any comments. Why not?
2. Basic problem with +/- is not that it's noisy. Problem is what +/- is measuring, and it's not impact of a single player. +/- is good for finding lineups inside the roster that work, but when you try to measure impact of a single player and then compare that impact to a different player in different club, you just got too many variables to get consistent results. There is correlation between good results and good +/-, but for good results 1 player is not enough and then you have take in count team dynamics too. And that's why results are noisy, 5 players together on the court are getting the same +/-, no matter what is their real contribution. You can have game of your life and your teammate playing the worst game of his life and +/- is the same. And then he's substituted, because he's bad and opponents make 9:0 run with 3 made 3s in a minute and he's suddenly + 9 in comparison to you. You can't find correlation between non comparable numbers. Or better yet, you can find correlation, but your interpretation is wrong. You should talk about impact of Luka's lineups and then analyse and compare his lineups with for example SGA's lineups in offensive and defensive side. Good luck with that. Like I explained earlier, when I compared Mavs and Warriors team D, Luka needs just good team D and Mavs results and +/- will skyrocket with him playing exactly the same.
When you're looking at MVP candidates, one of more important criteria is team results, and in that criteria is basically +/- already included. Just look at Boston, OKC, Clippers...they all have multiple players with around 10. Do you think it's coincidence that 3 Clippers players have almost identical +/-? Are all those players that more impactful than Luka? Of course not, their lineups are just so much superior than Mavs, who played with 30 different lineups this year. So Luka is first penalised, because his team has not good enough results, fair enough, but then is again penalised, because his +/- is not good enough, but that's basically the same thing. If you're losing your +/- is by default bad. Just look how Luka's +/- dramatically changed in last 6 games, when they started to play better D and Kyrie had few great games.
Scientists can't do miracles, they need robust data to make any kind of plausible conclusion. And +/- is not only noisy, it's not measuring impact of a single player.
3. If I have to explain why Luka had average numbers in his whole career, why you don't explain, why SGA's 50 games data is more representative than his first 5 years? Nobody makes that incredible jump over night. And you as analyst should be first to question that question. What happened?
1.
I don't recall seeing the post but I'll speak to Curry's +/- here since you ask:Fundamentally, there's a reason why I haven't been bringing Curry up in the '23-34 MVP: I don't see indicators of him having the impact necessary to be a serious candidate.
To speak to what's actually going on behind the Curry have negative on/off stats after being so dramatically positive as a matter of course for so long, well the biggest thing here is that the lineups with Curry in them have been really not so effective on defense compared to past years.
And that's noteworthy I'd say primarily because it's always been the concern about Curry. He has a reputation as an exploitable defender that underrates how good he typically is able to hold up, but it's not a surprise that if the bottom drops out of a guy's weakness as he ages out, it could fall hard.
Now, as we say, these are stats of correlation not causation. There's the further question of what specifically is going on in Golden State that is causing this. One natural thing to evaluate is just how much of a role Draymond Green's absence might be playing a part here.
Last year Curry played more minutes with Green than any other teams.
This year, Curry's played more with 4 guys more than Green (Klay, Wiggins, Looney & Kuminga). Yup, that's going to hurt.
As is the fact that Curry's not the only one aging.
There's also the Chris Paul factor of course, which is something worthy of bringing up as something of an argument against Curry as Paul - as much of a +/- master in his own right - is being significantly staggered with Curry, but we're not talking about a situation here where the team is being taken to a new level by Paul so much as the level of the team around Curry is falling.
Now, that all might seem confused to you given the things we've been debating on, but please try to focus on something other than a gotcha on me as potentially contradicting myself. There are other sub-questions you might want to ask though. To just take one chain of thought:
Am I saying that the drop in raw on/off from last year to this one represents a direct drop in Steph Curry's ability to play basketball? No.
Am I saying that its clear that Curry has aged out and will never put up his old style +/- numbers again? No.
Do I think Curry is most likely showing us the downside of the rainbow laid out before us? Yes, I think age is catching up to him to some degree.
Do I think that this is an indicator that Green really helps support Curry defensively? Yes.
2. So we have a clear, objective disagreement about the fact that +/- is first and foremost about lineups makes it something that says nothing about the value a player adds.
Fundamentally: Correlation is not causation, but consistent causation with sufficient sample yields correlation, hence correlation with sufficient sample actually does imply the existence of cause.
3. I actually did already respond to this so clearly we're both missing each other's posts. Understandable. Challenge to keep up when a thread is rolling.
Here's my 2 days old post,
Thanks for explanation. Now I will explain to you why I'm talking about anomalies, why they're so important to understand what is going on.
It's the fact that players of the best teams have in vast majority of cases the best +/-. We can see that normally multiple players of the best teams have great +/-, some are role players with great +/- too. Because of the nature of +/- it's impossible to understand what's is going on with looking at those teams/players.
So the only possible solution, if you're not total believer in +/-, you need to find anomalies and try to find out, why +/- is behaving like this.
Let's start with SGA,
He's total anomaly in my eyes. Firstly, he doesn't give me MVP vibes at all, when I watch him. Great midrange shooter and great driver and that's it. He doesn't do much for his teammates, is not good passer and is not very good defender. And then you look at his +/- and it's great. Something doesn't add up for me. So I looked at his past +/- and there's even bigger shock, he had awful +/- in first 5 years, he was well in -. His career +/- is still 1.7 worse than Luka’s.
When you see such a discrepancy between 5 years data, more representative, because it's much bigger, and 50 games data, you should ask yourself, what is going on here? How is possible that SGA came
from - 6.8 to + 11.1 in less than 2 years? There are only 2 possible explanations,
1. He became that much better in his year 6. You defended that thesis with some minor improvements in his game. But those improvements can hardly explain 18 points swing. For example, Jokic has only 7 points better +/- from his rookie year. And he’s totally different player now. 18 points swing is a difference between G-league player and GOAT. SGA has jumped from the worst leader of the Nba team, with disastrous impact of -6.8 to the best with 11+ in less than 2 years, by minor improvements in his game? Really?
2. He has much better players around him now. That thesis looks much more plausible to me, especially when we looked at what happened in last 2 years in OKC.
- Second best player Chet, who's averaging 17/7 and 2.6 blocks and has + 9.2 on court, is playing first year in Thunder.
- 3rd best player G-Will is averaging 19/4/4 and is playing second year in Thunder. He is having + 8.7 on court.
So in those 2 years Sga has got legit second and third opinion, that’s huge. And interesting enough they started with fantastic +/- from the start? Why didn’t they need those 5 years like SGA? Because they have great team from the beginning.
Thunder had 17th team D in 2022 and has 5th team D in 2024. And look at this, they had
30th offensive rtg in 2022 and in this year they had 5th!In those 2 years SGA went from - 6.8 to + 11.1 in +/-.
Now tell me, can those a little better ball handling and less TO's bring OKC's offensive rtg
from 30th place to 5th? and defensive rtg to 5th place? Or maybe getting great second and 3rd option is a game changer for OKC. I have no doubt, and I believe anyone honest doesn't have no doubt, what is the reason for OKC's incredible rise.
That should explain SGA's atmospheric rise in +/- and maybe show that good teammates are essential for having good +/-.
But we need some other conformation of my thesis, how essential are good teammates for having good +/-. If I'm right, we should see the same effect, if some starter from bad team, having bad +/- there, comes to a good team.
So let's look at KP, his career +/- is +1.5 and he has fantastic
+ 10.1 in Celtics. Do you believe that he had in his 8th year breakthrough season, like SGA had in OKC? Of course you don't. His great +/- is derivative of playing in the best starting lineup in Nba.
I believe I made pretty good case how essential are teammates for having great +/-, but I still didn't addressed enough Luka's particular situation.
I would use another anomaly for trying to address that. And I would use one of the best guards ever playing the game to help me. Because he's in one way very similar to Luka, generational offensive talent and not good defensively.
What is going on this year with Curry's is basically the other side of the same SGA's coin, sudden fall in +/- of the player, who had always fantastic impactful stats, but this year - 0.2; -4.9?
Like I don't believe that a player can rise like SGA overnight, I can’t believe that all time great player can fall like that overnight, on/off numbers would even suggest that he hurts Warriors, while he’s still producing very impressive numbers. So what's going on here?
Curry's offensive rating is even better than last year, when he had + 5.8; + 8 on/off. So he's still doing what he's paid for, but how is he negative player suddenly? Because his defensive rating fell for more than 5 points. Is this primarily his fault?
Do you agree that Curry was never good defender? And somehow Warriors had always one of the top D. in the league. And this year Curry's defensive rating plunged, taking his +/- together in the hole. Do you think that's because Curry overnight became a lot worse defender? Or maybe his defensive rating has something to do with bad D of Warriors' starting lineup? Warriors team D is only on 20th place this year, so we can safely say, knowing that he was never important player for Warriors D, that's very unlikely that he's the reason for that.
What can we find out from this case? How important is good team D for having good +/-. Which club is notorious for bad team D? Maybe Mavs?
So let's look at Steph and Luka's ofensive rating,
Seth Luka
2019 119.5 107.6
2020 108.5. 116.7
2021 114.2. 116.6
2022 114.1. 113.8
2023 118.1. 118.1
2024 118.3 119.7
Except Luka's rookie year and the year Curry was injured soon, they were more or less on par, Curry having better teammates. Interesting enough, the difference in +/- was always huge. How is that possible? Both great offensive players and both similarly bad in D, although I would still have Luka better there, he can at least rebound.
The answer is really easy, Warriors was always great defensive team and Mavs not.
Let's look at their defensive ratings,
Seth Luka
2019 105.8 110.6
2020 123.7 111.4
2021 109.6. 112.7
2022 103.4 110.3
2023 112.4 116.0
2024 117.6 116.9
We can see that all the difference between Curry and Luka in + /- is generated in defensive side. Watching Curry his entire career, because he was really entertaining to watch, I never thought that he’s a generational defender. But his career def ratings are great. How is this possible? My guess is that Warriors had exceptional team D, and indeed, they had top 5 D numerous times.
Is fair to say that one player is extremely impactful player and other has 0 impact, both being fantastic in offensive side, if 1 plays in team with great team D and other not? I believe we got answer on that question this season. No good team D and no +/- impact for Curry. He's imho still incredibly impactful, but +/- just doesn't recognise that, because it's flawed methodology for measuring impact of a single player.
To some it up, Luka’s average +/- is derivative of his bad defensive ratings. Would it helped, if he was great defender? Absolutely, but Mavs still wouldn’t be great defensive team. We have seen in Curry's case, that you can have great defensive ratings, even if you’re not good defender, if your team is build properly. Having Powell as starting C, no defensive bigs and being to small on wing position, make things incredibly difficult in D. It seem that Mavs are going in the right direction finally, so my prediction is, if Mavs build proper defensive team, Luka’s +/- will skyrocket.
Few things to add to my old post,
Team playing well in the season causes good +/-. Team playing bad causes bad +/-. So there's clear correlation between +/- and how team is playing. But at the moment we go to single player performance, everything falls apart. Players A playing good does not necessarily causes his team playing good and having good +/-. Players B playing bad doesn't necessarily causes his team playing bad and having bad +/-. There is no clear correlation between how player is playing and team results/ +/-, if you look at random team. But when you're searching MVP you're not looking at random teams, you look only at best teams, and there's your crucial mistake, you think that you have found correlation between +/- and how good single player is playing. But you basically only found out that star players, who're playing the most minutes in those winning teams, are having the best +/-. You found correlation between minutes played in winning team and +/- and you find correlation between domination of that team and +/-. Better team results equals better +/-. And with wrong understanding of what causes something, you try to correlate Luka's +/- on a bad team, especially with incredible problems with injuries, and +/- of the players playing in good teams. You're comparing incomparable things. Like saying someone is much better driver than other. One driving formula 1 and other normal street car.
To explain on example. We can look at what is happening in last month with Luka and Giannis. Luka + 103 in last 6 consecutive wins, he has got 80% of his +/- in his last 6 games, when his team, not necessarily him, played much better, especially in D. In some games Luka was great, in some Kyrie, especially the last, and in some games team D was great, pretty big influence in D had Gafford and PJ and in last, return of Lively. Basically the whole team played exceptionally well.
Now look at Giannis. Bucks struggling a lot, but Giannis is still mostly playing well and is having good +/-, when Bucks are winning. But when they're losing, a lot recently, Dame and others struggling, his +/- is very bad, in last 6 losses is - 59. And you can see how over the place is +/- in last Bucks game. Giannis with 35/4/12 and fantastic 15/17 shooting, - 14, far the worse among the starters, and Dame with
7/21 shooting and playing 0 D, +10. 
Sorry but you just can't build any kind of scientific argument on that bad data. When you see something like that, and those anomalies are basically in every game played, you should just conclude that model is not robust enough and dismiss results you're getting.
Did Luka's and Giannis's impact on the team suddenly changed dramatically? No, their team's success just went in opposite direction and because basketball is team sport, their +/- went in opposite direction too.
Or look at Dame. Does anyone believe that he's having great season? No one, except his +/-.
He has far the best +/- in last 5 years although all his other numbers and stats, raw, advanced, efficiency...has fallen off the cliff. What could cause that incredible discrepancy between +/- and his real performance? Only one answer, his teammates, which are light years better than in Portland. So what can you do with a model that brings you so wrong and inconsistent predictions? Nothing, you can just throw it in a bin.
