First Take: "Giannis is the better player, but Harden is the MVP"

I think the problem with Giannis is that because he can defend 1 to 5, people claim he can play 1 to 5. If you categorise him as a PF or C like Shaq or even Hakeem then it's quite clear how dominant he is. He can't handle the ball like a PG or shoot like an SG. And he only plays 32mpg.

But he doesn't need to because Bledsoe and the rest of his team shoot lights out... But mostly because of his gravity

Budenholzer should get plenty of credit as well.. But just like Shaq, everybody knows he's going to drive down the lane and dunk but they can't stop him

 

Amen. Finally, someone stating the obvious.

 

15 wins Missed Playoffs
41 wins L First Round
33 wins Missed Playoffs
42 wins L First Round
44 wins L First Round

And if not for Coach Bud, they'd be winning another 40 games with another first round exit.

"Better player"

 

But, but, but .... Poopoo is a better passer!?!

.......
.......
.......

 

First Take Twitter Poll: More important to team
62% Giannis
38% Harden

 

fwiw: When aggregation formulas include defensive metrics, keep in mind that the Sports leading authority on metrics -- Daryl Morey -- consistently says there are no defensive metrics that are useful, right now.

My understanding is some of the authors of those metrics are hobbyists in college. They certainly have a youthful exuberance with applying their impressive knowledge of classic statistics and math to their passion -- basketball. It is impressive. But I have to strongly consider Daryl's criticism of defensive metrics first and foremost, along with my own personal opinion of how horrible those defensive lists are, especially for guards and wings. Morey's take puts me at the "that's cool -- but dangerous if misused" level of interest on Off + Def aggregate stats, and certainly not the level of actually talking about the stats or sharing them, much less voting with them, or betting with them.

I like that they are now trying to implement some tracking data, beyond the boxscore (like touches), and get to do stuff with the PbP now, like differentiate between good and bad turnovers. Unfortunately, they can't run 15-20yr tests on any hypothesis on how to apply things like Touches, Contesting Shots, Bad Turnovers, etc, because that data doesn't exist in the past.

I have a question: some of their explanation (in papers I've seen) is over my head wrt classic Statistics Methodologies (my work uses other math), but I recall them talking about subjectively tweaking some coefficient (I might be using wrong term here) to get the results to align with historical expectations of who's best. On some of the formulas, don't they run models of 15-20yr data to see resutls, then have the ability for fine-tune the formulas to better align with expectations?​

 

I think I'm not quite as down on defensive metrics as what he has said. I would put it this way. I would put more trust in an expert or avid fan who really pays attention defense over them. But I think defense is just really hard to evaluate on a subjective basis for most of us (I'd at least include myself in this group). Look at the All-Defense teams and tell me how accurate you find them to be, historically? So, I think they are helpful as guides for more casual basketball observers. I don't think they should be seen an definitive, however. More like, if someone you expect to be high ranks kind of low, maybe that points to something he's not doing or an inefficiency in how the team is using him; and conversely if a player you don't expect to be high is ranked high, maybe there is something interesting there that he's doing that is worth paying more attention to.

For the purposes of making "best player" lists, we can first recognize that such lists are of little practical value to teams. It's just a recreational thing that some of us like to debate. And I think it is reasonable to consider defensive metrics that look at +/- data and other things that show up in a stat sheet into our assessment for such discussions, with the healthy recognition that these numbers don't tell the full story.

What that list from the site I linked to did is combine a bunch of different defensive rankings together to try to reduce the noisiness of the defensive value estimate. If I look specifically at defensvie rankings, by position, here is the result for SG/SF:

Top SG/SF

Code:

    Name                Team    CraftedDPM  DBPM    DRPM    MYDRPM    C19DPM    DPIPM   MYDPIPM
1   Robert Covington    MIN     4.05        2       3.48    4.13      3         3.1     3.5
2   Kyle Anderson       MEM     2.81        3.4     2.54    3.45      3         1.7     2.1
3   Paul George         OKC     2.66        0.8     2.92    1.72      0.6       2       1.2
4   Thabo Sefolosha     UTA     2.53        2.6     2.02    1.92      1.8       2       1.7
5   Andre Iguodala      GSW     2.33        1.8     2.42    1.64      1.1       1.6     1.4
6   Danny Green         TOR     2.1         1.1     2.04    1.8       1.4       1.5     1.4
7   Joe Ingles          UTA     1.98        0.9     1.91    0.43      0.9       1.6     1.3
8   Maurice Harkless    POR     1.89        2.3     1.74    0.05      0.7       1.5     1
9   Justise Winslow     MIA     1.72        0.8     1.4     2.48      1.1       1.3     1.4
10  Stanley Johnson     NOP     1.69        0.9     1.75    0.75      0.7       1.2     1.3
11  Victor Oladipo      IND     1.69        1       1.19    3.2       1.5       1.2     1.8
12  Josh Hart           LAL     1.57        0.9     1.67    -0.58     -0.7      1.4     0.8
13  Royce O'Neale       UTA     1.49        2.1     1.5     1.66      1         0.8     1.1
14  Jayson Tatum        BOS     1.42        0.7     0.72    2.5       1.7       1.2     1.5
15  Jrue Holiday        NOP     1.41        -0.1    1.13    2.19      0.6       1.1     1.6
16  Marcus Smart        BOS     1.32        1.3     1.74    1.61      0.8       0.5     0.7
17  Bruce Brown         DET     1.22        1.5     0.73              -0.6      1.2     0.9
18  Garrett Temple      LAC     1.15        0.6     0.93    -0.54     -0.3      1.2     0.7
19  Josh Richardson     MIA     1.15        -0.5    0.73    0.5       0.6       1.2     1.2
20  David Nwaba         CLE     1.14        -0.4    1.2     -0.96     -0.6      1.2     0.5

What sticks out, to me, is that this appears to be very team-context dependent. Players on very good defensive teams, even those who don't have a great defensive rep individually, are present. Maybe that's appropriate, and coaching and team setup determines how well a player performed on defense relative to other players on other teams. It also means that you can't assume these rankings would be stable if players shuffled to different teams.

It is possible, but I haven't read any of the recent papers that you may be alluding to. I'd question trying to fine-turn a formula just to make it align to historical, subjective evaluations. What would it use, All-Defense selections? I don't think those are very reliable.

 

Giannis hasn't even had their most points in their last 2 games. I'm sure these people voting for Giannis do not watch the Rockets.

 

A new game plan for the only team that has beaten Golden State 6 times out of the last 11 times they've played?? Riiiiight.

 

One thing you can count on.....if there is an *internet poll* Harden will lose it regardless of what it is about.

 

Maybe I'm misreading how they tune the formulas after doing test runs on 15+yrs data. Let's consider offensive stats only: My understanding is the purpose of regression analysis or solving linear equations (for instance) is to weight the various component stats in importance. Example: Boxscore stats need to solve -- Rebounds are weighted like this; Assists are weighted like that, etc. Advanced +/- stats need to solve the value of teammates, or mythical "replacement players".

Maybe they are just tuning the weighting in a purely mathematically manner, in effort to find good correlation (for instance) -- R-squared stuff that is over my head.

Question: So, from your experience, you haven't read them doing any tweaking against subjective expectations?​

I would think at the very least, if the results of running tests on 15yr data provides ugly lists, they would not publish that formula.

Rhetorical questions: What is totally over my head is when is an R value of 1 a perfect correlation?

• Is it a correlation to one's hypothesis that a meaningful linear relationship could exist?
• Is it a correlation to an observable fact?
• Is it a correlation to one's expected top player?

​

Seriously, thanks, any education you can provide is much appreciated. I'd very much like a better high-level understanding of how they test their formulas, so I can better understand their explanations of how the formulas came about.

I'll hang up and listen ......

 

I know a lot of people say "OMG, why are you watching this stuff?? Just ignore it and have fun watching our team play. Who cares what the say!?" Look, I get that argument, and I certainly wouldn't say they are wrong to have that opinion.

That being said.....just to play devil's advocate. Other comparisons though, would be "Well, they are saying things that are just clearly wrong about one of our favorite players. These comments are being spat out over and over to possibly new fans, and they believe it's just common knowledge." You see it obviously by comments on videos with 90 percent being "Harden flops, Harden travels, Harden gets every call." It wouldn't be that big of an issue if someone in the media had the guts to just flat out call these so called journalists wrong. Not by just saying they're wrong, but actually using stats, actually using video clips proving it isn't travel."

The worst I've seen was an episode of the jump, where they actually showed the video of someone showing how Harden wasn't actually traveling, and at the end, all 3 hosts on the jump said "Yeah..but it's still traveling." Why is Harden just so freaking hated? He plays the game the right way despite what Scottie wants to say. He doesn't run his mouth, he's not a problem in the locker room....he's great for the NBA. He's showing players that they don't have to be the most athletic to be the best in the league. All these idiots in the media, ESPECIALLY former players, are just jealous at the fact that it comes so easy to Harden. The fact that Harden is truly smarter than these former players. It's horrible that he's become the enemy. He should be looked at as someone other players should try to be like.

 

My “experience” is just reading online articles on this stuff over the last 15 years or so, and perusing the public domain stats websites that you are probably familiar with. I used to participate in some forums that focused on these advanced metrics years ago, but not so much recently. So, there may be stuff out there that’s doing what you are describing, but I’m not aware of it.

The closest thing I can think of is boxscore +/- (BPM), which regresses boxscore stats onto an adjusted +/- rating. A linear model is often used, but I’ve seen people try other more fancy stuff. The idea is to come up with an estimate of the player’s true impact by combining numbers from the boxscore. I find that the results often align well with my subjective impressions, even when compared with adjusted +/-, but subjectivity is not used for the regression itself.

You’re probably right. If the numbers are really off, it could point to some methodological problem or a bad model used for the formula. So subjective impressions can operate as a final filter for whatever gets published publicly.

​

So, I’ve sometimes seen people try to assess the value of a boxscore stat like PER or BPM or WS/48 by finding how well it correlates to adjusted +/-. You wouldn’t get r=1, or even close to that, because adjusted +/- is so noisy. But the higher the correlation, the idea is the better that stat is at capturing true impact of a player.

 

Last year when MVP was undeniably James Harden, the media started talking about who the best player in the game was and largely awarded that title to LeBron James. Now, the best player on the healthiest high end team is better.

James Harden is clearly the MVP this year again but people try to argue why he isn’t because he doesn’t have a chip. Once he does, they’ll talk about ‘giving it to someone else’. This is the nature of Houston sports. The nation despises our greatness starting with Moses Malone and ending with the next great after the next great after the next great etc etc etc Rocket.

Being a Houston fan is all about digging into your own personal fandom. Hopefully your fandom respects how good Calvin, Moses, Elvin, Rudy, Hakeem, Clyde, Yao, Scola, CP, EGo, Clint, PJ, and Harden really are and were. This IS our year again.