The Bucketless
2023 BDB Finalist
Github: the-bucketless
Email: thebucketless @ protonmail . com
He/him
- Looking through my backlog and came across this: huggingface.co/SimulaMet-HOST Models and data for computer vision for hockey. Haven't really looked through it, but public data is cool to see.
- Reposted by The Bucketless✨new scraper✨ I will write some docs, improve the scraper, etc. eventually ✨it notably includes an xG model i built (model gives lower xG values than most other models [can mean I am dumb lol], so expect lots of goals scored above expected) github.com/maxtixador/s...
- Very minor update to the rink package for an error caused by a matplotlib update to the contour plot github.com/the-bucketle...
- There was a WHL game where Codey Burki scored in the first shift of the 2nd period that got taken back when the refs remembered they'd given him a penalty at the end of the 1st
- Reposted by The BucketlessNHL Draft model v2. Wrote a rough draft about what I think was going wrong the first go. Not there yet. drydan.github.io/posts-hockey...
- Saw a goal with a great water bottle pop tonight. I miss those.
- We're talking football! Well kinda. @centralwinger.com asked if anyone was interested in explaining The Zoo model, so I figured I'd give it a go. thebucketless.wordpress.com/2025/04/01/e...
- Reposted by The BucketlessIt’s hard to put into words what building and scaling the DynastyProcess shiny app has meant to me, but I took a crack at it in this talk and shared some of the lessons it taught me along the way! #rstats
- "Ottawa leads the PWHL in time spent on the penalty kill" Well they did name their team after a penalty
- me, writing code: I am so smart me, trying to run the code I just wrote: I am such an idiot
- Reposted by The BucketlessHey y'all its a shot-plotter.netlify.app update! This one features the addition of: - football/soccer net (IFAB/NCAA) - indoor lacrosse (NLL, customizable) and: - ice hockey (IIHF) is now customizable All additions were suggestions from people, so let me know if you have thoughts of things to add!
- most places in the us have a bad/boring joke food that they like to fool everyone else into thinking they like. for example louisville has the hot brown, an extremely off-putting meal no one there ever eats but tricks visitors into trying. if your community doesn't have a shitty joke food, start one
- Just hit Ctrl+C on one computer and Ctrl+V on another. It, uh, didn't work.
- New version of hockey rink with this, a new wavy_arrow plotting function, and some example notebooks github.com/the-bucketle...
- The wavy arrow can also be used for normal matplotlib plots
- Tried out the NHL's ice image with the hockey-rink package. Might have to include that as an option.
- I uploaded all the goal tracking data for 2023/24 to GitHub (across multiple files). I also included a csv of issues I've found that I may update as I play around some more. github.com/the-bucketle...
- Had to look up at least one
- The 8581 goals last year amounts to a little over 13M rows. For reference, BDB this year is almost 60M. Fewest frames: 89 Most frames: 123 3 goals show the puck recorded as having moved less than 10 ft total. The one in the QT is at 2:31 of the 2nd from 2023020040.
- While giving half a foot leeway around the edges of the net, there are 38 plays that didn't have the puck in a net at any point
- Seems fitting that I'm doing this within a day of BDB data being released
- Thanks to @nguyenank.bsky.social for the SVG size (2400x1020)
- Good morning, it looks like the NHL has posted player and puck tracking data and it’s publicly accessible. Going to write up some notes on how to use it today! Here’s Caufield’s goal as a sample, there’s some transformation we can do to more easily work with it. wsr.nhle.com/sprites/2024...
- Here's the code I used. A couple notes: 1) Coordinates are image coords, not plotting (need to flip the y-axis). 2) The id column is, presumably, used to identify separate shifts for players (don't need to track them on the bench). Not as useful when all we have are goals.
- Threw together a quick Python Shiny app to do head-to-head comparisons for last season using this thebucketless.shinyapps.io/nhl-bayesian...
- For a bit of fun, I tried out some Bayesian RAPM
- New version of the rink package is up. Improved the binning code and fixed an issue with hexbin. github.com/the-bucketle...
- Based on Sportlogiq data, middle entries in the PWHPA were more likely to lead to shots, but not necessarily goals thebucketless.wordpress.com/2024/07/19/p...
- Smoothed out some RAPM numbers while focusing on Mo Seider thebucketless.wordpress.com/2024/06/26/r...
- Turnovers at the top of the ozone are around twice as likely to result in shots against as pucks that make it to the neutral zone (using Sportlogiq PWHPA data) thebucketless.wordpress.com/2024/06/06/t...
- I used the Sportlogiq PWHPA data to see how valuable pre-shot movement from passes was to an xG model thebucketless.wordpress.com/2024/05/16/e...
- More PWHPA powerplay work using Sportlogiq data. This time looking at ozone play. thebucketless.wordpress.com/2024/05/05/p...
- I used Sportlogiq data to cluster some powerplay breakout patterns thebucketless.wordpress.com/2024/04/24/p...
- Quick post comparing Platt scaling, isotonic regression, and Venn-ABERS for calibrating an xG model thebucketless.wordpress.com/2024/04/12/c...
- I compared shot data from @gdignard.bsky.social to the PWHL's to see whose numbers might be getting exaggerated thebucketless.wordpress.com/2024/03/27/p...
- Once again trying to find where the camera's pointed thebucketless.wordpress.com/2024/03/19/t...
- Very much a fan of the PWHL adopting every fun rule change the NHL is too scared to try
- Feel like I should put some PWHL fun here as well, so here's goal locations by team
- Reposted by The BucketlessTake a look at PREDATOR our Big Data Bowl project.
- Trick to speeding up neural network training: have the model output all NaNs
- Why do models keep doing what I train them to do instead of what I want them to do
- Not what I was looking for, but ghosts on the field are fun too (the sky really needs some videos/gifs)
- Time for my new annual hobby: finding plays in the BDB data that track the wrong ball
- Playing around a bit with the data from the new NHL API. I built a model for last season's shooting/goalie talent and the list of the worst shooters includes a couple 30 goal scorers, a former 40 goal scorer, and a couple other fun names.
- The pbp in the new NHL API has timestamped shots-on-goal going back into the 90s and goals/penalties going back to 1917/18
- When using the rink package with larger figures, you can set linewidth to 0 to get a little finer detail on the rink
- Reposted by The BucketlessTADA I made a whockey BlueSky feed: https://bsky.app/profile/did:plc:nwceptoqh7h2foyap7trf5l6/feed/aaajzzgt6lj6u Right now it has these case sensitive keywords so if you're using anything different, lmk and I can add it: women's hockey, woho , whockey, pwhpa, PHF, nwhl, cwhl, NWHL, CWHL, PWHPA