New York Friends: I'll be on a panel at the first annual Symposium on AI & Sports hosted by the Columbia-Dream Sports AI Innovation Center on Thursday, September 12.

Registration is free! (and there is a remote option)

lnkd.in/g2deuMMS

From novice to expert—STaRT@Rice has the workshops to match your research journey. Grow your skills and network with us! I’m excited to lead a session on ridge regression and the lasso in R. Register here: start.rice.edu #STaRTatRice2024 #Shapingthefuture

Let's compare batter results on their "fast" swings (above that batter's average bat speed) vs. their "slow" swings.

MLB avg (FB only, min. 50 comp. swings)

"fast" swings
84% contact
44% hit into play
31% squared up

"slow" swings
85% contact
34% hit into play
23% squared up

What is going on with those long Cruz swings? Are they misses because they are long? Or are they long because they are misses? Here's the distribution of contact/miss swing length by pitch type. Miss-swings are longer for breaking balls and off-speed but shorter for fastballs!

Here is a comparison of Cruz's and Soto's swings. At the risk of making the figure too complicated, I also annotated contact/miss. The upshot is that Soto never takes swings > 8.5 feet like Cruz does, and these swings are almost always misses.

By contrast, here is a visualization of within-player variance, specifically for Oneil Cruz. We see there is not much relationship between bat speed and swing length across Cruz's swings. Cruz doesn't seem to swing harder when swinging longer.

I'm particularly interested in comparing *between-player* variance in bat path and *within-player* variance in bat path. For example, here is between-player variance in bath path. We see that there is a correlation: Batters who swing hard tend to swing long.

First-years Brady Detwiler, Devin Abraham, Tyler Emanuel and Wyatt Bellinger developed Points Saved by Tackle (PST). For made tackles, they trained a neural network to predict the counterfactual EPA of the play if the tackle had been missed.
www.kaggle.com/code/bradyd/...

First-years Lou Zhou and Rahul Herrero developed Play Value Without Penalty (PVWP). They used random forests to gauge yardage penalties based on the expected yardage had the defender needed to make a tackle instead of committing the penalty.
www.kaggle.com/code/louzhou...

Juniors Jonah Lubin and Charlie Wells developed TacklrTrackr and Adjustable Tackle Metric Hub, two apps that allow users to visualize eight attributes they developed to measure tackling; find similar players; and create custom leaderboards.
www.kaggle.com/code/jonahdl...