Bryan Shalloway

{funspotr} -- #rstats function spotter -- is now compatible with quarto documents and is available on CRAN. Rstudio conf 2022 pres: youtube.com/watch?v=c9oU7A… package on github: github.com/brshallo/funsp…

I'm referenced here:

@cruickshank @exante_me @thebethocracy may also find notes of interest.

@cruickshank (I don't have strong feelings on the debate but had been planning on pulling together data and doing some analysis for my own interest but hadn't gotten around to it. I've been following you and @exante_me on this topic and thought you might find these couple notes of interest.)

The @SeaPubSchools board has posted a resolution that will be voted on today, with less than 12 hours notice, authorizing the superintendent to propose closing 5 schools and setting up a process to close more in coming years: drive.google.com/file/d/1jgOaje…

@exante_me Thought you may also be interested in some of my notes here: x.com/brshallo/statu…

Must read op-ed from new board member Sarah Clark. I hate those little gifs but “This. All of this.” seattletimes.com/opinion/sps-bo…

Reading "Everything is Predictable," is interesting how some of the origins of probability came about by identifying "fair" ways to split a gambling pot part way through a game. Could be interesting to calculate in the context of a game of pick-up... applying pascals' triangle

@NoahKuntner @predict_addict @YannickKae @ProfASoares @kaggle It's back up.

@predict_addict @YannickKae @ProfASoares @kaggle sad, but understandable. your repo used to be my go-to place to check for updates in the space. all the best!

Awesome Conformal Prediction has a small but growing section referencing @Kaggle materials. And yes conformal prediction often improves even ROC AUC and other non probabilistic metrics. Submit a quality notebook and get it featured on the repo! #kaggle #conformalprediction github.com/valeman/awesom…

Gave a lightning talk at @cascadiarconf on how prediction intervals are helpful both in informing intermediary agents and in setting the bounds within which you want those agents to operate : youtube.com/watch?v=jkxp98… #rstats

When playing make-it-take-it pickup basketball the expected value of shots can be approximated as a geometric series: bryanshalloway.com/2024/08/27/how…

#rstats #sportsanalytics post on expected values under different possession and scoring systems in pickup basketball. One takeaway is that, once you leave "balanced" FG % conditions, shooting incentives in make-it-take-it can get out-of-whack pretty quickly.

1s and 2s full court is a scourge that must be defeated but if you think you "get it" yet are playing make-it-take-it by 2s and 3s, you are missing something. bryanshalloway.com/2024/08/27/how…

If you're curious why and what you should do with this information, I wrote a short blog post on the topic: bryanshalloway.com/2024/08/27/how…

When you work it out under these conditions, it turns out 1s and 2s likely offers more balanced incentives than 2s and 3s.

However 2s and 3s isn't always better. If you're playing "make-it-take-it"--where scoring comes with retaining possession--expected points are not just {expected points of shot} but also {expected points of future shots you may get if you score and therefore retain possession}.

When I play pickup basketball I like to go by 2s and 3s because it gives more similar expected values to a wider range of shots compared to the more commonly played 1s and 2s (where taking long range shots often becomes the dominant strategy).

Inconsistent rules are like plot holes in a movie, you get pulled out of the experience... and then, what's the point? Why play? (... yeah, yeah, time with friends / family, blah blah blah...). If I have a say in the rules, I like to choose rules that make for a balanced game.

I don't like the phrase "it's just a game." Yes, the stakes may be low but the zest of life comes from *getting into it.* I tend to be a stickler on rules because rules facilitate balanced games. I'm fine with house rules but be consistent.

A lot of armchair AI psychologists out there...

Pascal's triangle is awesome: youtube.com/watch?v=XMriWT…

Very good visual for understanding epistemic and aleatoric uncertainty (youtube.com/watch?v=Cla7b6…).