mikey

Baseball analytics discourse died 2 years ago when people stopped feeling like they could contribute to the frontier. When the plus metrics (e.g. stuff+) came out machine learning became the bare minimum and it was basically over. Honestly analytics became too glorified even though it's just learning about the game that everyone watches. Like it's not cool anymore to post linear models or something. This wasn't so 5 years ago when I learned to code hoping one day I could be like one of those people on fangraphs putting players into buckets and seeing their differences. Maybe people forgot that the fun part is the analysing the game and not the statistics. And for some reason it became okay(?) to gatekeep the full process. Next month we'll see another person post the nth stuff model and not the source code. Baseball analytics was always about 'baseball' not 'analytics'. And baseball is still so, so far from being solved. Just look at the massive RMSEs on any stuff model. But at some level, people need something to build on to stretch the frontier, and as long as people are required to build their stuff models from scratch to talk about stuff, the discourse won't happen. That's why open sourced my command project, anyways. I meant it when I said hopefully more people can pursue and share cutting edge baseball research. Honestly I'm just looking for more friends to talk baseball. Hopefully I can release a frontier stuff model soon, too. Command tracking has massive potential. It comes with so many interesting questions: - Do (starting) pitchers stop throwing pitches they can't command that day? - Do some pitchers command better under stress (e.g. playoffs)? - Are miss patterns dependent on arm slot? - Does command have aging curves like velo? - Do long at bats hurt command? - Do mound visits help command? - Are 1st inning struggles command issues? - Are consistent pitchers just ones with good command? - Does slowing down velo help command? - Do pitchers have command issues early in season? I came up with these last week. There's gotta be so many more. If you're considering trying baseball analytics, this is your chance to work on the frontier. These questions are up for grabs. You have like 5 people competing with you right now. No ML required. All you have to do is go to @open_command repo/releases and download csv to get started. If you're struggling to even download full season savant data DM me and I'll help you. And if you want to see more analytics discourse (or the answers to these questions) repost this so it can reach as many potential analysts as possible!

Some news: I decided to open source my command project. It's called OpenCommand. This means anyone can see the code (and even the data)! I hope this gives more people a chance to pursue and share cutting edge baseball research. Anyways, here's how to measure command: 🧵 1/N