American Soccer Insights Summit. Save the date: January 30 - 31, 2026. Registration Open.
We are thrilled to announce that the American Soccer Insights Summit will return in 2026. The summit will be held at Rice University on January 30-31. Register today at:
www.ticketleap.events/tickets/amer...
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
I'll be at Saberseminar this weekend! If you'd like to meet, book some time with me using my Calendly.
calendly.com/saberpowers/saberseminar
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
The American Soccer Insights Summit is seeking sponsors to keep registration costs accessible and to fund travel for deserving students. If you are interested in being a sponsor, please reach out: americansoccersummit@gmail.com.
Re-posts are appreciated!
americansoccerinsights.com/sponsors
That did not occur to me. I will check that out. Thanks!
I just checked. There is a slight exacerbation with two strikes, but only slight. It's not strong evidence in favor of my hypothesis.
So what's going on here? I *think* I did the calculations correctly. Further investigation is required.
Perhaps batters are sometimes adjusting their swings just to make contact and foul the ball off. If so, we would expect to see this phenomenon exacerbated with two strikes ...
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
Adi Wyner asked a good question on the Wharton Moneyball show last week: We observe that bat speed correlates negatively with squared up rate ACROSS batters. But what about WITHIN batter? For a given batter, when they swing harder, are they less likely to square the ball up?
I'll be there!
Conclusion: Let's exercise caution with our interpretations of swing length. It's probably true that longer swings lead to more swing-and-miss. But I need more convincing that the new data provide strong evidence in favor of this hypothesis.
When hitters are fooled by off-speed pitches, they start swinging early. The "point of contact" ends up out in front of the plate, increasing the measured swing length. Here, being fooled is correlated with both miss-swings and swing length. We see the opposite for fastballs.
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.
I wondered whether we could use this to measure swing adaptability: Which hitters are appropriately adapting their swings based on the pitch?
Standard Deviation in Swing Length (min. 150 swings)
#1 out of 186: Oneil Cruz (0.9 feet)
...
#186 out of 186: Juan Soto (0.5 feet)
This isn't unique to Cruz. It seems generally true across players. Players who swing hard tend to swing long. BUT for an individual player, swinging harder doesn't correlate strongly with swinging longer. I am curious what causes differences between swings for individual players.
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.
HUGE thanks to all the folks at MLB who put in the work necessary to bring us pitch-by-pitch bat-tracking metrics publicly available on Baseball Savant. This is an awesome dataset, and I appreciate the opportunity to play around with it.
Better late than never, mRchmadness (w/ Eli Shayer) is up for 2024!
We're NOT in the business of predicting outcomes. But if you have those predictions from another source, our tool maximizes your chances of winning your pool (using pick data from ESPN).
saberpowers.shinyapps.io/mRchmadness/
Just got my ticket!
⚾ Save the date - April 13!
Join me at the UConn Sports Analytics Symposium for a historical journey through Baseball Analytics. We will discuss its present state and open problems, how we got here, and envision the future. Don't miss out!
statds.org/events/ucsas...
#UCSAS2024
Luke Stancil, Naomi Consiglio and I estimated a touch-by-touch point win probability model for women's college volleyball. We measured core skills (serve, receive, set, attack, block, dig) all on the scale of Points Gained.
The preprint is now on arXiv:
arxiv.org/abs/2402.01083
Another one of my favorites:
pebblehunting.substack.com/p/baseball-a...
Having worked in MLB front offices as a data scientist for several years, sometimes it *is* hard to be romantic about baseball.
Thank you @sammillerbb.bsky.social for rehabilitating that feeling within me over the past 11 months.
I loved this piece.
pebblehunting.substack.com/p/mike-trout...
💯
This is my biggest lament with the task of creating player projections. You're always wrong, and all you can do is try to be (a little) less wrong.