Bob8 wrote:Doctor MJ wrote:Bob8 wrote:What about SGA in previous year's? How you can totally dismiss his first 5 years, because he had great first half of this year? How is possible that your analytic mind doesn't at least have some doubt, how could he turned +/- overnight dramatically? Can someone have average +/- for 5 years and then suddenly becomes a monster? Why don't you look what has changed with Thunder's roster this season? Some new important players? Having the same starting lineup for 47 games? I'm asking you for the 3rd time, why SGA made this jump in year 6? Please for the answer.
I'm sure that you're not Luka hater, I just believe that your tool is wrong to evaluate single player impact. Your tool is useful for finding best lineups and evalute lineups in one roster. To have proper, realistic comparison between Luka and SGA, they should have played with the same roster. Or they at least have to have similar team construction, similar starting lineups, similar bench and similar problems with injuries. Mavs are total outliner in all all those areas in comparison to others playoffs teams. They have for years far the worst starting lineup and pretty good bench, Luka has for years great replacement, Brunson/Kyrie, on his position and they were hit the most with injuries this year. Please explain to me how +/- is adjusted to all those differences?
But what I accused you of, that you're totally dismissing all other data, eye test including, and blindly believe in just one tool. Yes blindly. Luka can average 70/30/30 with fantastic efficiency from now on and your numbers and you would still have SGA better.
Luka and SGA has played 2 games against each other this year. And it was total domination by Luka, not only by numbers, but with the way he led his team. Anyone could see the difference. Unfortunately for Mavs, roster construction and problems with injuries count in basketball too. Not even MJ could do much without proper help. Your answer would be small sample size, my would be, why would have meaningless blowout games bigger impact thsn head2head? This is basketball, the best players play, when game is on the line, when every possession matters, not when their team leads by 20+. You can have negative +/- and be far the most impactful player on the court. MJ was not special because he dominated scoring from his first year, but because he made the basket every time, when it was the most important and D knew that he's taking last shot, but couldn't do anything about that.
You're implying how in playoffs everything might change. But it shouldn't, if your tool is working right. The most impactful player of the RS, should be in the top in playoffs too. But I have a feeling that especially SGA's impact will fall off dramatically. No more blowups, more minutes, more double teams for him, less FTs...and very likely first round exit.
Re: what about Shai's previous years? I mean, a young star is supposed to get better.
Re: dismiss first 5 years. Not when I talk about Shai's career. But when talking about the MVP this season, I tend to focus on this season. (May seem different because of the Luka conversation, but we get into earlier seasons there largely because people try to use small sample size as a reason to dismiss this season's +/- data on its own.)
Re: how could he turned +/- overnight dramatically? Well, that's the most valuable question, and also a very complicated answer that I wouldn't claim to have any monopoly on. Thread-worthy, in other words.
Wherever you see a guy with high primacy & big minutes spike like this, there's a meaningful set of causes to look into and understand.
I can speak to it more later, but try to avoid looking to discredit the data by discrediting me. You should try to explain the data yourself without looking to dismiss it as being an improvement of everything except Shai and see what you come up with.
Re: 70/30/30 and you'd still ignore. I don't and wouldn't ignore any of this data. I just recognize that it doesn't capture the entirety of the competition, and that if that data doesn't match other data, there's other stuff going on that is significant.
Re: Luka beat Shai badly! Cool, and the data for those games exist, just as the data for all the other games do. The head-to-head may end up deciding a playoff series, but what happens against the other 28 opponents matters too.
Re: can be most impactful with negative +/-. In a game this is not common but possible. In a season, in the NBA, this is very, very unlikely.
Re: Jordan. Jordan has outstanding +/- by what we see, just like we'd expect.
If you're making a case about your model for 1 player, saying that 6 years data confirms your thesis, you can't dismiss first 5 years of other player and be impressed by only first 3 months of current year. At least you should give me a viable explanation for that big change. And btw, SGA is born in 98, so he's not that young either.
SGA's sudden rising is basically death sentence for your model, that's why you don't want to talk about it. Yes players get better, but you're not average, by your model, and then become monster overnight. Like Thunder is not become top Nba team, just because SGA's became better, basketball is a team sport, single player, no matter how good he's, alone can't do much. So I would look for explanation in the changes of Thunder's roster, development of other players and not put everything on SGA's overnight rise. But there's the catch, isn't it? If we say that roster is partially to blame for sudden rise of SGA's +/- than we admit that +/- is measuring impact of lineup not a single player.
If someone is averaging 70/30/30 and you don't see him as impactful player, I believe you should watch more basketball and less numbers.

Data for yesterday's game in which Luka had 32/8/9 with 81% TS, + 18, in only 3qs is saying that he had negative on/off, because he didn't played in q4, when lead ballooned from 21 to 35. Whoever has watched yesterday's game and saw what was Luka doing to Thunder as playmaker and as scorer can just laugh at that data. And no, you can't say it's only 1 game, problem is much bigger, if you can't trust single game data, you can't trust seasonal data either. +/- data is over the place, because it measures impact of 5 men lineup not impact of the single player. And yes, players of good teams have normally good +/-, because their lineups are the best.
You have by quality very different starting lineups, by quality very different replacement players, very different benches, different team success and data, which is over the place in single games and we're only at the players team lineup, what about opponents? They play with different lineups too, have good or bad night... And after all those variables you want to represent that data as all in one measurement for impact of a single player? Really?
I would say that knowledgeable observer can see pretty good, what impact the best players have on their teams. Seeing and measuring impact of role players is much more difficult.
Ah, I thought you might be confused on this point. You think I'm changing the sample I care about to suit my bias, but I'm not.
- I'm literally saying I think each guys' this-year MVP candidacy is pretty well-represented by what this-year's +/- data tells us about them right now.
- The only reason I'm bringing up years before this-year in a thread about a this-year award, is because you and others are trying to use arguments to dismiss the stats' validity when talking about Luka Doncic, and when you do that, you talk about what's going on this year as if these things are flukes. But if your thesis - to use your parlance - is that these things are flukes, then looking to other years of Luka's data as experiments is the natural thing to do, and those other years tell a similar story.
- I would acknowledge that there's more uncertainty about Shai's data because the sample of him doing this is still short relative to someone like Luka, and if someone wanted to say that they had no issue with my assessment of Luka's data, but that I was being too hasty with my assessment of Shai, they would be logically consistent, but to be clear, this is not what you're doing.
- "At least I should give you". If to you, this is about you and me, then you're approaching the conversation wrong. That might seem I'm taking words out of context, but honestly, it really seems like you feel like if you can only discredit me then you don't have to deal with the data. And I'm saying: Deal with the data without trying to excuse Luka. Ask yourself honestly: If, hypothetically, there were things about Luka's play that were holding back how much impact he has on his team in a typical regular season game, what might they be?
- But to try to speak some to what all is causing Shai's On/Off to look so much better than in previous years, I think the thing to focus on might be his improvement in ability to handle the ball a great deal, push the pace, and massively diminish turnovers for his team, particularly by avoiding steals against.
Now, like anything else in this team sport, team context has effects here, but so does a player just getting that much more comfortable and as they say "seeing time slow down".
- "SGA's not that young". He's young enough to be of an age where we'd still expect him to be pre-peak traditionally, particularly if he's a guy able to demonstrate improvement toward mastery at things rather than being a fast-twitch superstar.
- "Not average and then a monster overnight". In terms of impact, I'd say that's sometimes basically how it works sometimes. A player can't literally become 200% as good across all skills overnight, but sometimes you figure a particular thing out, and sometimes your team figures a particular thing out making use of what you can do. I do appreciate your skepticism here though. When I see data along these lines, I try to keep my mind open as to what all is causing it.
- If teammates matter, then with +/- then... Value is about teammates too. This is not the Best-Player-in-a-Vacuum Award.
- Re: "If 70/30/30 not impactful...". Okay so, as I add up your numbers, I should walk back what I literally said because I was speaking in principle there that probably wouldn't hold up to such unrealistic numbers. Why?
70 points scoring plus 60 points assisting would give us the greatest offense in history, so while it's in theory possible the defense would be so bad to overwhelm this, when you're talking these types of numbers, the offense just has to be great.
But if I chop those numbers in half, to 35/15/15, I'd get numbers that I think you'd assume must be creating a huge positive impact, while I would say it's not plausible that it isn't. Why?
35 points scoring plus 30 points assisting is a mere 65 points. If I run my offense through a player like this for 36 MPG, and that's literally the entirety of the scoring my team gets, then I have a terrible offense.
This in a nutshell is why it's so problematic to look at individual box score as proxy for impact: "Huge" player numbers are still only a fraction of the team's total, and so the question becomes:
Since most of a team's scoring has to come in possessions where the star is neither scoring nor assisting if we want our offense to be successful, what is our starring doing to help for success in those possessions?
- "+/- all over the place because measuring lineup". So to be clear, I do understand this. I've been explaining to people about the difference between validity and reliability in statistics to RealGMers since well before you were on this site. The nature of +/- is that it requires greater sample before we can use it as confidently.
It feels to me like you're under the impression though that this means we should look to utterly dismiss it until we reach supreme confidence in our interpretation of it without any other supplementary knowledge, and I would object to that as both impractical and counterproductive.
The fact, for example, that Shai's +/- data largely aligns with the box score and occurs as the team is doing great, means what we're actually choosing between is:
a) Great guy looks great across multiple first-pass perspectives, yup, 'bout what we'd expect.
b) Yes, the guy who looks great in every other metric also looks great by impact metrics, but it's merely a coincidence that these things agree because of some mysterious balance of forces.
I'd say (a) is the more likely truth. You are of course free to disagree with my assessment.
Re: one measurement for everything? That's not what I'm doing. What I'm saying is that this is data that needs to be explained in order to understand what's going on with Luka's impact, and that arguments based on the specifics of this year's Mavs are simply possible explanations which cannot explain why the data looked similar in earlier years.
Actually, let me flip something there: You should understand that I'm coming at all of this time-forward. I've been paying attention to all of this data since long before Luka's career, and that means that when I'm talking about what I see this year, it's something that I saw after having seen the previous experimental data - meaning, watching previous NBA seasons.
So I'm someone coming into each season asking: "Is this the year Luka's +/- data looks like I'd expect for a superstar-level player?". We're now most of the way in the '23-24 season and the answer happens to still be "No", and so you and I have this conversation in a way that has moved time-backward. I talk about the past because you bring up arguments about the present that seem simpler to you but they complicate things for me:
If the injury-etc arguments about this year are knocking Luka's +/- in weird ways, then we're getting two things happening at the same time: a) Luka finally figured out how to be much more impactful than ever before, b) but he's getting absurdly unlucky and so all the +/- data still looks about the same.
Is this possible? Yes. But it's complicated and unlikely in my estimation. And further, I don't think most Luka supporters would agree with the assessment that Luka had an impact problem before this year. In my experience, people who believe in Luka's MVP-ness have been trying to dismiss +/- data as a matter of course for years and years.