Juany8: To add to what Milt already put extremely well (and the author of that particular study put even better) you’re not actually testing anything. You can’t just throw random numbers up against each other until you find an answer that suits your worldview, there are too many confounding factors to simply test numbers on a macro level. I don’t think the study I showed gives a conclusive answer, I just like the way he approached the topic and that he realized a simple graph of efficiency vs. possessions isn’t going to give you anything meaningful.

Since the only thing that really matters is how much a player improves the team’s offensive efficiency (not his own) looking at lineups instead of individual players gets rid of many of the confounding factors present in the individual stats. He also breaks it down even further, which won’t necessarily reveal anything but it does help me make my own decisions about the data I’m looking at. That’s pretty important, since for some reason you seem to ignore that the only things scientifically established are things that can be repeatedly proven by anyone who cares to try. You can’t just tell everyone else they are wrong and tell them to shut up, your model has to be able to withstand scrutiny and criticism. Considering your only response to the criticism is to make racist soccer jokes that don’t add to the discussion in any way, I’d say that’s literally all I need to know about WP’s validity

I’m not going to argue with someone who thinks that economics is a pseudoscience, but don’t accuse me of racist jokes because I picked “soccer” as my “sport that is not basketball” in the point I was trying to make. Your race (which I do not know — are you an Argentinian of Italian descent? German? Castillian? Oaxacan? Dominican? I have no idea) Would it have been racist had I said “baseball?” Latinos love…

Juany8: Nash obviously makes them better

A couple weeks ago, I posted that I thought that Nash’s injury was going to have a much bigger impact than originally reported (i.e., that he would be out 7-10 days). A few people disagreed because the bone that he fractured was “not a weight-bearing bone.” It’s been 4 weeks since he got hurt and, as of two days ago, he was still in too much pain to even begin jogging. I have a feeling that we are not going to see vintage Steve Nash this entire season.

http://www.basketballprospectus.com/article.php?articleid=1412

The Honorable Cock Jowles: Wait… what’s the difference, O Math Master, between the linear regression done in the study that supports YOUR hypothesis and the one that supports mine?

To add to what Milt already put extremely well (and the author of that particular study put even better) you’re not actually testing anything. You can’t just throw random numbers up against each other until you find an answer that suits your worldview, there are too many confounding factors to simply test numbers on a macro level. I don’t think the study I showed gives a conclusive answer, I just like the way he approached the topic and that he realized a simple graph of efficiency vs. possessions isn’t going to give you anything meaningful.

Since the only thing that really matters is how much a player improves the team’s offensive efficiency (not his own) looking at lineups instead of individual players gets rid of many of the confounding factors present in the individual stats. He also breaks it down even further, which won’t necessarily reveal anything but it does help me make my own decisions about the data I’m looking at. That’s pretty important, since for some reason you seem to ignore that the only things scientifically established are things that can be repeatedly proven by anyone who cares to try. You can’t just tell everyone else they are wrong and tell them to shut up, your model has to be able to withstand scrutiny and criticism. Considering your only response to the criticism is to make racist soccer jokes that don’t add to the discussion in any way, I’d say that’s literally all I need to know about WP’s validity

The Honorable Cock Jowles:
You’re saying “prove to me that something doesn’t exist” without having evidence that it does.

I think there’s another point to be made here. Someone much smarter than me (Kevin Pelton? Dan Rosenbaum?) already put it much better than I’m about to, but I couldn’t find their original quote.

The starting point for statistical analysis in sports has to be conventional wisdom. So: the hot hand exists, defence wins championships, scoring is the most valuable individual skill, there is such a thing as a clutch player/team, you need proven winners to succeed in this league, etc. These are opinions that have been around for ages, and are firmly entrenched in the minds of fans, players, coaches, and general managers.

I completely agree that many of these opinions are completely and utterly false, but the onus is firmly on the statistician to provide the evidence against them.

The Honorable Cock Jowles: Wait… what’s the difference, O Math Master, between the linear regression done in the study that supports YOUR hypothesis and the one that supports mine?

Short answer: it’s not the R^2.

Long answer:

Okay, let’s assume that the positive slope in your example is statistically significant. Even though the R^2 statistic is small, we would still have evidence of a relationship between two variables: (1) the year-to-year change in a player’s usage and (2) the year-on-year change in a player’s TS%. The problem is not the methodology (although not reporting the standard error for the estimate of the slope is pretty poor form); it’s that the analysis is not asking the right question.

A player’s usage/TS% can change for several different reasons. Maybe they moved to a much better/worse team. Maybe they’re being used in different lineups. Maybe the new coach has boosted/shattered their confidence. Maybe they’re gunning for a new contract, with amazing/awful results. There are so many confounding variables that it’s virtually impossible to draw a conclusion.

The clever part of Eli’s analysis is that it breaks down the data by lineup. This way we can eliminate many (though by no means all) of the other variables.

iserp: But they don’t “get a shot off” in actual games; they get lots of offensive rebounds and assisted buckets; but those aren’t going to magically make 100% of the shots if you combine these guys.

How do you quantify how well they are gonna fare when they have to “get a shot off” if you don’t have enough data from actual games?

So what you’re saying is that your assumption (that they can’t “create shots”) is more likely than my assumption (which is that a moderate increase in each player’s usage will not yield a significant drop-off in efficiency).

You’re saying “prove to me that something doesn’t exist” without having evidence that it does. The analyses I’ve cited may be crude, but they are better than a gut assumption that those players fundamentally lack the ability to take shots simply because other players on their teams take more shots than they do.

milt:

Seriously, this is why the online basketball analytics community is neatly divided into two camps:

1) Berri’s echo chamber.
2) Everyone else.

oh my god, I love you. if we had sig files here, this would be my new one.

Juany8: Thank you Milt, it’s really nice to talk to someone who seems to understand how science actually works. The WoW website doesn’t seem to comprehend how a hypothesis works, what constitutes valid data vs. invalid data, and that linear correlations are what you use when you literally don’t have something better to use.

The biggest fallacy present in society today is that attaching a number to your opinion somehow makes it more objective. Numbers are the easiest things in the world to manipulate, it is entirely possible, for example, to graph TS% against FGA, against TSA (True Shoot attempts), you can compare PPP vs. Possessions (including turnover ratio) It’s entirely possible to do all these correlations, pick the graph that fits your narrative the best, and just stick it on a website. Doesn’t make your work scientific, it makes it purposely misleading and very subjective

Wait… what’s the difference, O Math Master, between the linear regression done in the study that supports YOUR hypothesis and the one that supports mine?