538 Basketball posts

[Bobbofitos]

So far I'm very disappointed in 538's basketball content:

http://fivethirtyeight.com/datalab/advi ... l-jackson/

http://fivethirtyeight.com/datalab/lets ... th-the-3s/



2 fairly weak articles thus far. Very little data or original "digging". The 1st essentially says that Phil Jackson has won a lot of championships; the 2nd says we may be at the apex of taking a lot of 3s. Wondering what you all think.

(I should add I have a lot of faith in Nate Silver, so I believe it will get better...)

[mtamada]

Yeah, pretty weak so far based on those two articles. On the one hand, they're just getting started. OTOH they've had plenty of time to prepare some killer articles so they could've come out hitting the ground running. There's no shortage of opportunities to do some innovative research. Or at least frame the questions or results in interesting ways. He did make a point which is under-appreciated in the second article, namely the difference between the average probability of a shot going in and the marginal probability. However though under-emphasized, it's not an innovative point; it's something that basketball researchers have been struggling with for decades.

But as you say, Nate Silver has a very strong track record. Nobody hits a homerun every at bat (or a 3-pointer every FGA), so I think it's a wait-and-see situation for a few months.

[Mike G]

Neil Paine hasn't written anything there on basketball yet, has he? Several on baseball and football, one on hockey.
Depending on how they are perceiving their readership, once he starts it'll be clear he's their best on that subject.
With 7 contributors, does it seems they're unusually deep in basketball writers? -- Neil is 3rd string.

[schtevie]

Wow. Thanks, Bobbofitos (or should I say, no thanks?) for that reference to the article about three point shooting.

Let me begin by saying that I've always been a Nate Silver fan (dating back to his anonymous Poblano days, when he was kind enough, from his undisclosed location, to take the time to reply to an email asking for guidance in understanding some of the empirical issues related to the Obama/Clinton primary battle) though I guess I'm more of a Sam Wang guy now, but I digress.

Paul Krugman, in his blog, recently took issue with a 538 economics article for it being theory (hence context) free, exposing a particular weakness of taking a "data driven" approach. A related and equally serious problem is revealed in the three point article.

In addition to it being weak on what is rather simple theory, it is historically uninformed, and really weirdly so. And this is perhaps a more damning criticism, because we're talking empirical history here, and if you're data-driven, to not avail yourself of this publicly available-knowledge is a bit difficult to forgive.

Basically, the article is a retread of the cited Goldsberry Grantland piece (http://grantland.com/the-triangle/trio- ... nge-shots/) to which the same criticism applies.

This aside, the bottom line on the 538 piece is this: if you want to talk about the marginal vs. average value of three point shooting and then speculate about what this bodes for the future, how about structuring your argument around a simple graph showing 3PA and 3P% over time. To have done so would have both informed the audience as well as inevitably resulted in a completely different article.

[Kevin Pelton]

However though under-emphasized, it's not an innovative point; it's something that basketball researchers have been struggling with for decades.

I would argue that's a goal for a site like FiveThirtyEight -- bring issues that researchers are struggling with into the mainstream, so that a larger group of people understands them. In my experience, familiarity with this concept is far from widespread, even among researchers.

[AcrossTheCourt]

Those were only datalab posts. The datalab is like their blog, and it's for shorter-form articles. I would like to see a longer feature from them first.

Though if they ever need help, I'm obviously in favor of writing something for them. (Said slightly in jest.)

Also, I'm sure they're busy with March Madness.

[schtevie]

However though under-emphasized, it's not an innovative point; it's something that basketball researchers have been struggling with for decades.

I would argue that's a goal for a site like FiveThirtyEight -- bring issues that researchers are struggling with into the mainstream, so that a larger group of people understands them. In my experience, familiarity with this concept is far from widespread, even among researchers.

Hold on now. The difference between the marginal and average value of a shot is not something that researchers are struggling with, at least at the conceptual level. This is a distinction where a solid middle-school education offers sufficient expertise.

The 538 piece, but more so the Goldsberry piece it cribs, is, for reasons unclear, a bit of an intellectual disaster (the allure of slatepitchiness?) And bringing this into the mainstream, in such a state, is, quite simply, a disservice.

Those were only datalab posts. The datalab is like their blog, and it's for shorter-form articles. I would like to see a longer feature from them first.

It is ironic that the blog is titled "datalab" as the post had basically no data in it (certainly nothing original). Like I said, a plot of 3PA and 3P% (and throw in 2P% and mid-range 2P% when the data comes available) vs. time, and add a couple sentences of context and you have the makings of a great blog post - one in the spirit of the original 538. But this was not that, and the audience was fundamentally misled.

But let me extract myself out of the mire of negativity about this specific post, and return to the larger issue of what it augers for 538. For Nate's site to offer insightful sports journalism, it had better be based on a grounded knowledge of both relevant theory and history. If not, I fear, it will not be long of this internet, or at least a site worth checking in on.

[knarsu3]

Yeah, I was expecting more as well. The 3pt article was interesting for me because of my own research on the topic- contested 3s being better than open mid-range shots. But as Ben pointed out on twitter, the early shot clock contested 3s are going to be selected and therefore, better shooters would be taking those 3s. And as you get towards the end of the shot clock, where basically any one will be shooting contested 3s so they get a shot up, the %'s end up being about the same as shooting an open mid-range shot. Also, as I pointed out in the article, league average doesn't mean much since a team would only care how a specific player shoots.

I think the best way to look at this might be to look at specific players and how it varies over the shot clock. Or perhaps, the end of the shot clock should be used as a baseline since at that point, anyone will be shooting just to get the shot up. Even at that point though, it's still pretty even and rarely does anyone allow an open shot that late in the shot clock.

Anyways, it all depends on the players. Something like this but with larger sample sizes so it's actually meaningful (big problem would be it would take multiple seasons to accumulate that data at which point players' skill may change):

[Dr Positivity]

538's quality has been disappointing in general. The articles are as if they see the audience as n00bs who need short simplistic "fun" data pieces to understand anything, which happens to be tons of posts where small sample size makes it meaningless info

538 does not appear to have understood why Grantland is a raging success, not a surprise I guess because Silver is Simmons' Black Swan/Complete opposite

[Neil Paine]

Neil Paine hasn't written anything there on basketball yet, has he? Several on baseball and football, one on hockey.
Depending on how they are perceiving their readership, once he starts it'll be clear he's their best on that subject.
With 7 contributors, does it seems they're unusually deep in basketball writers? -- Neil is 3rd string.

I can't really comment a whole lot on the subject, but I am not currently able to publicly write about basketball. When that moratorium ends, I will be writing quite a bit about basketball.

[Mike G]

Haha, thanks, Neil. It's better just knowing there IS a reason, even if we don't know what it is.

For any who don't know, Neil was blogger and user advocate at b-r.com for a few years. One of his great strengths is his awareness of studies already done; so he won't tend to re-invent every inquiry from scratch.

[knarsu3]

Haha, thanks, Neil. It's better just knowing there IS a reason, even if we don't know what it is.

For any who don't know, Neil was blogger and user advocate at b-r.com for a few years. One of his great strengths is his awareness of studies already done; so he won't tend to re-invent every inquiry from scratch.

I think it's really obvious what the reason is. But it's interesting to hear that he'll be able to start writing about basketball soon- I guess consulting for a team and working for fivethirtyeight full-time is too much work. Certainly sounds like a ton. Maybe that's also the reason we haven't seen any longer pieces from him yet either.

[wilq]

So far I'm very disappointed in 538's basketball content:
2 fairly weak articles thus far. Very little data or original "digging".

Wait a second, aren't you omitting interactive tools and odds for NCAA bracketology?
Maybe it was supposed to be their huge basketball debut... and for mainstream point of view it probably was.

Overall, my guess is they have many basketball posts prepared or in the works, it's just not the time to publish them. During March Madness why would they focus on NBA playoffs or draft or some other random projects? They will have plenty of time and opportunities for that in the next few months...

[Crow]

knarsu3,did you assemble your Kobe chart from your own play by play database or from a public source? t is interesting to see an example chart but I'd also like to some day see a composite chart for the league or maybe split charts for PGs, wings and bigs.

[knarsu3]

knarsu3,did you assemble your Kobe chart from your own play by play database or from a public source? t is interesting to see an example chart but I'd also like to some day see a composite chart for the league or maybe split charts for PGs, wings and bigs.

It's from the Vantage data so not public (have the data for each shot). As for the league, it should be in here: http://blog.cacvantage.com/2014/02/the- ... fense.html

Towards the end. The graph is slightly different. Alternatively, here's another version:



I haven't done it split into positions yet but that'd be easy enough to do. Could even do defended by certain positions as well.