Wake Forest MS student. @CalPoly alum. Formerly @DrivelineBB, @SLOTribune, Sounders FC.
Seattle, WA
Joined September 2015
- Tweets4.4K
- Following370
- Followers601
- Likes17.5K
Scott retweeted
Think you have what it takes to solve important problems in professional baseball? Help us solve pitcher workload.
(Link in next tweet)
I was curious to see what a vision model could pick up.
Pose tracking says the tip was consistent all year. With a runner on 2B, SL averaged further forward at the elbow in 22 of 23 regular season games.
It’s small, but if you knew where to look could you see it?
The Brewers stopped swinging at Mason Miller's slider. San Diego swore he was not tipping.
So which is it? Is he tipping, or are you just seeing ghosts?
Being early and being wrong look identical until the numbers show up. I went and got the numbers.
ericniesen.substack.com/p/ch….
Scott retweeted
this is so sick. nice work @sdmiddlecamp !!
Uploaded the splat to Github. It has every angle but is rough at the edges where no camera could see.
Drag and zoom while it plays
sdm-commits.github.io/gaussi…
Uploaded the splat to Github. It has every angle but is rough at the edges where no camera could see.
Drag and zoom while it plays
sdm-commits.github.io/gaussi…
After nine games, the AFL rules are doing for offense exactly as the model predicted. Only the magnitude was wrong.
My original model under predicted XBH since it was trained on MLB data without marks.
A simple lookup table (green bar) is much closer to actual.
MLB's testing of new outfield positional rules in Arizona seems to be working exactly as intended.
Offense is WAY up.
Is that a good thing?
baseballamerica.com/stories/…
The week 2 changes moved outfielders ~5° towards CF, taking away more XBH in the gaps while adding some down the line.
Week 1: +1.42 runs per team-game on the old marks Week 2-present: +1.26 runs
1
672
Interesting use cases for player interviews, podcasts, etc.
Attribution was always a problem with media scrums.
Meet NVIDIA Nemotron 3.5 Transcribe 🎙️
An open-weight, promptable 30B MoE model with ~4B active parameters that brings transcription in 40 languages, speaker attribution, translation, contextual prompting and audio understanding into one checkpoint.
Watch it keep track of who said what when everyone’s talking at once.
1/ 🧵
4
537
Payroll tracks talent even more closely over longer stretches: .41 for one season, .51 averaged over 3 years, .56 over 5 (1997-2026).
190
The top is back to mid-2010s levels: 2.31x the league median in 2014, 2.06x now. The bottom five fell from 0.71x to 0.49x.
MLB’s $171M floor would lift them to 0.81x. 12 teams are under it, including every team still playing.
1
101
How playoff catchers stack up on ABS challenges vs. an average playoff catcher
Hicks (TB) +0.22 runs, 3-2
Contreras (MIL) +0.06, 7-7
Bailey (CLE) +0.01, 4-7
Hedges (CLE) −0.39, 2-0
Smith (LAD) −0.48, 2-2
Above the line = better than average
1
1
294
