APBRmetrics

Just a quick note, I'm not sure if folks saw, but I took over as editor of Nylon Calculus today. Ian is now running all of Hardwood Paroxysm so needed someone to handle the day to day. I just wanted to say that the virtual door is always open for questions, comments, suggestions and even the occasional submission if you have something interesting to share. Thanks.

Looking forward to more great work....congrats!!

I checked Nylon Calculus and Hardwood Paroxysm for comments. In each case I found 150 articles in a row with zero comments and stopped looking further. I raise this not to embarrass but to ask what is going on and what if anything to try different. My hunch is that most have migrated to chat on twitter. In the last year I have tried that but the debates are so scattered, chopped up and limited by the antiquated 160 character cap and die out easily. Would NC and HP considered a central hub forum at one of their sites or here instead of hundreds of empty pockets? Are you happy enough with the feedback at your twitter sites? Is discussion dead because there are so many good articles that no one has time left to discuss?

I like the Nylon Calculus articles, but I'm not a huge fan of Disqus.

Yeah, me either. I have thought of posting occasionally and then decide not to wait for it to open. My login had issues too. Could re-register but if nobody else is commenting, pass for the moment.

A quick check of twitter action @nyloncalculus: In last 10 days about 80 tweets, almost all from NC, NC writers and a few from friendly bloggers. 5-10% drawing tweet response; none more than a few of course brief tweets. Maybe that was enough, but it is not a whole lot. What is, what was, what could be.

NateTG wrote:I like the Nylon Calculus articles, but I'm not a huge fan of Disqus.

A known issue, I hate it too.

sethypooh21 wrote:Just a quick note, I'm not sure if folks saw, but I took over as editor of Nylon Calculus today. Ian is now running all of Hardwood Paroxysm so needed someone to handle the day to day. I just wanted to say that the virtual door is always open for questions, comments, suggestions and even the occasional submission if you have something interesting to share. Thanks.

Hey congrats. Good luck, have fun, here's hope you get tons more people over to the dark side - basketball analytics.

http://nyloncalculus.com/2015/10/08/str ... -creators/

Couldn't get disqus to load on my phone, so commenting here.

Creation % might be a good new tool. Would be interesting to see correlation with team efg% RAPM factor. And compare with correlation / possible impact of usage, total usage, % "Moreyball" shots, etc.
Perhaps also look at each of these by a split of the shot clock. The impact of creation shots could well vary by when they are taken.

Working the interface between box score / optical tracking data and RAPM could give greater meaning to both. They have what each other needs: significance measures and tangible action analysis.

i have a question on the 7/6/2015 nylon calculus article on robin lopez and rebound value...

i'm guessing the gest of the article is yes C robin lopez is a poor defensive rebounder, but because the team rebounded better defensively with him than without him, he must be doing something to improve that defensive rebounding (boxing out opposing rebounders)...

C joel anthony (mia) from 09-10 to 12-13 was an even worse defensive rebounder per minute than was robin lopez, but each year the heat grabbed more defreb/48min with him on the floor versus off. also those 4 years lebron james had 4 of his best 5 years for defreb/min. is the logic also that anthony was doing something to improve the team's defensive rebounding (other than grabbing them) such as boxing out opposing rebounders?...

I'd look at the adjusted rebounding factors. It by itself may not be enough to answer the questions but I'd value a judgment that included consideration of both raw and adjusted data more than one that didn't. Raw on/off data alone for lineups can easily create misimpressions here and there. If the two data sets agree, you probably can go with it. If they disagree, more study is required. I'd generally go with the adjusted but not always.

– The (Updated) Math Of Transition Defense (from Joseph Nation, Nylon Calculus:
Read it here: https://fansided.com/2018/07/31/nylon-c ... n-defense/
– Game Theory And The Deep Three (from Senthil Natarajan, Nylon Calculus):
Read it here: https://fansided.com/2018/07/31/nylon-c ... ry-deep-3/
– Real Plus-Minus Is Less Real When Players Change Teams (from Andrew Johnson, Nylon Calculus):
Read it here: https://fansided.com/2018/07/31/nylon-c ... nge-teams/

Thanks for the heads up.

The third link got my first attention.

The third link for me as well, as it relates to a follow up comment owed here (on what is an important topic beyond the specific focus): http://apbr.org/metrics/viewtopic.php?f=2&t=9454

First and foremost, I am grateful for the interesting and informative contribution.

A few comments on the data and approach taken (https://fansided.com/2018/07/31/nylon-c ... nge-teams/) and hopefully Andrew Johnson can be here to answer.

First, as an aside, the overall data presented graphically were a bit surprising to me. If one draws the line of unitary slope (representing no change in RPM year to year) seemingly a very large majority of points lie below. Perhaps my intuition is incorrect, but there is the adding-up constraint based on the fact that the sum of minutes times RPM in any given season (approximately) equals zero. And I don't think that the fact that sub-800 minute seasons are out of the sample, given where they would lie, would be an explanatory factor.

Regarding these omitted player seasons, a question is why? Were they initially included and the results seen as too noisy?

Were I to exclude one subset of the data it would instead be healthy, high-minute players where the role in changing teams is assumed and revealed not to have changed. This would not just include (most of) the anomalous RPM elite, noted in the article, but players in the next tier down, such as Kyrie Irving, where anticipated RPM stability was clearly realized.

Next, a preferred univariate regression would add the age-adjustment to the base year RPM, as that is a part of the statistic. I suppose the reason that this wasn't done is that this information is proprietary? Perhaps it could be made not so or the previously published xRAPM aging curve formula serve as a supposed close substitute?

Another data question: My particular interest relates to explaining the well below-average seasons of players lost by the Celtics in the Kyrie trade, however I cannot locate the data for Jae Crowder (3.89, -2.13) or Isaiah Thomas (1.83, -4.23) on the graph. What am I missing?

Finally, could the standard errors of the slope of the regression be shared? One can eyeball the scatter plot and get a pretty good idea of what that is. But this statistic is important in terms of how people's thoughts should be guided in terms the expected consequence of player trades/movement, so perhaps it could be shared?

Thanks.