New year, new resolution. I hand craft a lot of data insights and datasets. Hoping to use agents more to help. Thinking of writing a skill.md for each database schema (dataset) so an agent knows how to join views and tables, the main filters to use, and the key columns. #databs

I released an update to mplsoccer that should make it easier to add other sports.

I can confirm this works and looks ace. I may be able to finally release a Statsbomb football data parser with no dependencies apart from duckdb.

We've open sourced QuackStore - a block-based caching extension for DuckDB! 🦆 QuackStore dramatically speeds up queries on remote data by intelligently caching only the blocks you need. Now available as a DuckDB community extension: github.com/coginiti-dev...

Has anyone got any good examples where they've used AI in data work? #databs

I herby restart my campaign for my modest proposal to reduce the size of the penalty area to 30m wide. It might be more palatable to fans than Ted's arc from this pod.

Dev and deploy require of diff skills. @getdbt.com bridges the gap for #data pipelines, but what about MLOps? I explored how @posit.co {orbital} + dbt delivers "good enough" zero-infra MLOps for batch model scoring in database from python or #rstats models www.emilyriederer.com/post/orbital... 1/

I have finally published my post about Moneyball, as promised (a long time ago). If you're interested in baseball, numbers, or movies, please take a look. I'd love to know what you think. cc @matsonj.com @alexnoonan.bsky.social djpardis.medium.com/revisiting-m...

Wild. Need to try this!

🚀 mplsoccer 1.5.1 released @dmitry.mclachbot.com adds the capability to use custom slice labels in the pizza charts

🌀 𝐄𝐅𝐏𝐈: 𝐄𝐥𝐚𝐬𝐭𝐢𝐜 𝐅𝐨𝐫𝐦𝐚𝐭𝐢𝐨𝐧 𝐚𝐧𝐝 𝐏𝐨𝐬𝐢𝐭𝐢𝐨𝐧 𝐈𝐝𝐞𝐧𝐭𝐢𝐟𝐢𝐜𝐚𝐭𝐢𝐨𝐧 ⚽ The newest release of the 𝒖𝒏𝒓𝒂𝒗𝒆𝒍𝒔𝒑𝒐𝒓𝒕𝒔 Python package (v1.1.0) now includes the functionality of my latest research paper. 🔗https://tinyurl.com/unravel-efpi

Tried this out on mplsoccer. Definitely impressive deepwiki.com/andrewRowlin... But , I think the generated docs are more interesting for someone looking to contribute to the codebase, rather than users.

🚀 @unravelsports.com added some new formations to mplsoccer for his formation detection work. Hopefully will see some results some day.

@unravelsports.com just seen your PR today. Of course adding extra formations is helpful and happy to add it in. I'll try and look at it soon 🙏

Yamal is so good he pretty much broke my axes when I looked at young attacking outliers. Streets ahead.

It's early days, but seeing if I can adapt mplsoccer to other sports

I have successfully switched mplsoccer docs to install via uv rather than conda. It took a while to fix some install problems with lxml. I think Conda still wins is installing the complex C dependencies in some data science projects (e.g. PyMc), but worked around it here with apt install.

The margins between winning the Premier League and being runner up are incredibly tight, just seven points separates the teams on average.

If learning from survey data, how do you feed in the non-response weights? #databs

Musing about the dark patterns you could build into a stadium design to cement home advantage 1/ screens showing the opposition making mistakes or missing penalties around their dressing room

🔥🎉New library: boosting for survival analysis, including multiclass (competing risks) Survival = missing outcomes because limited observation window (common in medicine, marketting...) soda-inria.github.io/hazardous Gives very fast boosted-trees for survival

uv init

🚀 Great to see this improved data structure! Glad we could be of some help with this. Thanks @pff-fc.bsky.social!

👏streamlit-soccer will solve some things mplsoccer isn't great at like animations and interactivity.

New blog on booting it long.

Set pieces seem to be hot at the moment 🔥 so I decided to investigate whether lumping it long into touch is a good ⚽️ strategy andrewrowlinson.github.io/blog/footbal...

This holidays, I'm going to attempt to learn enough javascript/typescript to build an API 🤞.

Spotify label fail 🤘 #divorceddadrock

Anyone have a good resource for customers who bought this product also bought these products. How is it usually implemented in practice? #databs

I just played floorball/ salibandy and most of the goals came from wide areas. Does anyone know is there any similar expected goals metrics for floorball?

Often details make a great #dataviz. A few things I noticed here: a) the pentagons are facing the flow of the lines b) the group A..H are left/right aligned c) the bold font for the knockout and group so it's easy to see 👀 d) choice or circle for representing the group stage

I'm the king of unfinished side project.. mplgrid is 80% there just need to figure out a better API 1/3 github.com/andrewRowlin...

Could be useful for the football crowd. The code is available too. Who is going to be the first to try it with @carloncarpenter.bsky.social wide angle footage?

Trying something new: A 🧵 on a topic I find many students struggle with: "why do their 📊 look more professional than my 📊?" It's *lots* of tiny decisions that aren't the defaults in many libraries, so let's break down 1 simple graph by @jburnmurdoch.bsky.social 🔗 www.ft.com/content/73a1...

Where did our lost rhinos roam? Distribution charts based on bone finds (individual dots) and habitat suitability for the four fluffy rhinos of late Pleistocene. These are of course approximate and might not include all obscure finds. #Sciart

I still think outlier detection is under utilised 🤪

👋 new followers. I am going to write a few #dataviz, python and data cleaning tips that I use regularly. Anything you want to hear about?

I have been thinking of making a duckdb backend for requests-cache for a personal project. To query APIs and cache the raw data. Does this make sense? #databs

Newsletter friendly website good open.substack.com/pub/absolute...

Copying over a July blog from the 🐦 site on scouting football/soccer players using outlier detection. andrewrowlinson.github.io/blog/footbal... #databs

Thinking of LLM-as-a-judge and wondered if anyone had tried building a supreme court of 9 LLMs with different instructions.

Hi football peeps. 👋 👋 How are you finding me?

I often name files wtf.csv if I want to quickly eyeball data 👀. Anyone do anything similar? #databs

Super hard in aviation... Trips have one or more bounds, bounds have one or more segments, segments have one or more legs. Then there are passengers (fly one segment), customers (fly 1+ trips), and tail numbers (aircraft). I always start problems with which level should I model at #databs

Loads of great #rstats and #dataBS starter packs going around, here's an attempt at a #SportsAnalytics starter pack - please DM if you think I've missed you/anyone else!

@markposts.bsky.social @henshawanalysis.bsky.social @timkeech.bsky.social anyone made a starter pack for the football analytics crowd?

Is anyone part of a data science community at work? What do you do to make it successful? #databs

Get started with football analytics today with mplsoccer