The big day has finally arrived: introducing StuffPro and PitchPro!
Getting these pitching metrics ready for prime time has taken a ton of effort from the team, and we're excited to share them with you!
@eveewing
My wife and I will call out what we’re doing like we’re getting ready for launch.
Me: Seatbelts..
Her: On.
Me: iPhone cable..
Her: Engaged.
Me: All systems are a go, preparing for ignition in five, four, three....
it wasn't just good luck or facing the Marlins; Fried looked truly phenomenal last night. idk what a hitter is expected to do when someone is locating six plus pitches perfectly
Some exciting news from R&D at
@baseballpro
. We've joined forces with
@srbrown70
. Steve has created the best in class 'pitch stuff' metric and we'll be sharing that (and more) over at this season. Follow
@baseballpro
and
@pitchinfo
to be ready for launch.
It may be petty but I’m still not over Freddie dragging Acuña in public last October while playing like shit because he’d been covering up an injury for months.
@EliBenPorat
part of it is aesthetics- guys with an 80 hit tool are fun to watch- but also 33rd all time in fWAR from age 27 on plus 10 gold gloves. also think cultural impact should play a part in voting, and there's no denying that for him.
I think public stuff models are missing something about these "bridge" cutters (movement between a pitcher's FA and SL) that some teams have figured out. StuffPro, Stuff+, and PitchingBot all see his FC as merely average, but the results are solid and the Yankees believe in it.
Here's a story about Carlos Rodon's cutter... sort of.
Really, it's a look at how technology is used to develop new pitches so rapidly, and whether that's a good thing.
Insight here from Rodón, Matt Blake, Gerrit Cole & more.
🆓
@NYDNSports
|
#Yankees
reminder that the winner of a kaggle competition for predicting likelihood of a swing found that swing decisions on non-fastballs are based on if the pitch at release looks like it would or would not be a strike if it were a fastball with average movement
full grid of xwOBA against a pitcher's primary fastball by it's velo relative to their average FB velo that day (columns) and by that pitcher's average FB velo that day (rows)
on average hitters perform like Matt Olson on 95 mph fastballs from pitchers who typically throw 96
Everyone give a warm welcome to Georgia Rose and August Rae Sutton-Brown!!
Mom and babies are healthy and doing great! I couldn't be happier or more excited about finally getting to meet them. They are perfect in every way
Folks, you don’t expect us to blindly acknowledge a bogus record, right?
Ronald Acuna Jr. is a transcendent talent.
But he’s not the first 40-70 guy.
He’s the LATEST 40-40 guy.
Stolen bases are ⬆️ 40% this season.
70 SB’s - 40%=42 swipes
Love Ronald, but wake up!
#ForTheA
The share of houses being built with 10 or more bathrooms has gone up nearly seven-fold since the 1970s. "If he were so inclined, Alessandro Giacometti could stay two weeks at his Southampton, N.Y., home and use a different bathroom every day."
STUFFPRO LEADERBOARDS ARE LIVE!
Welcome to your new one-stop-shop for pitch evaluations. We've provided a ton of exciting features and information, so let's dive straight in to see what you can find 🧵
Pitch targets should change based on the level of command that a pitcher has.
This gif shows how the expected run value of targeting a particular location varies as the pitcher is more likely to miss their spot.
Notice how the backup/high slider is more reliant on command.
something to keep an eye on: Luis Guillorme's fastball has been the worst pitch in baseball this season according to both StuffPro and PitchPro (min 20 pitches thrown)
interesting tidbit here. I remember reading that the Blue Jays do the same thing. anyone who has a stuff model has the framework to make a swing decision model too, so not surprising teams have this, just love the idea of an app to review after games.
I wrote a blog post digging into the Stuff+ team switching debate that I hope you all will read! I tried to make it accessible to the lay person while still providing adequate depth to the discussion. Please let me know your thoughts!
taking a break from the Strider show to toot my own horn: no matter how you prefer to run your correlations, my version of Stuff+ predicts next-season ERA better than any available public metric
"i don't know who needs to hear this but" here are correlations (r²) with next-year ERA (60+ IP in current and next year, n=305)
0.216 Stuff+
0.197 pitchingbot stuff
0.185 Pitching+
0.144 SIERA
0.131 bot overall
0.101 pFIP
0.099 bot ERA
0.091 xFIP
0.090 pFIP_EV
0.074 FIP
seems more likely to me that folks consistently overperforming their xwOBA is due to things xwOBA doesn't consider- like batted ball spin or spray direction- rather than confidence. interesting piece nonetheless
first was on a much lower and steeper pitch than the others, which in my head makes me think less backspin? wonder if including pitch height and/or VAA on a p(homer) model with EV and LA would capture this
one thing that makes it difficult to assess the value of swinging hard is that hitters will slow their swing down if they realize they got fooled. so it's not simply that fast swings create more value but also that swings that were on track to create more value stay fast.
@tressiemcphd
we discovered our yorkie likes playing a game where we hide a piece of food in our hand and let her scratch our hand to get it. now at night she stands by her food bowl and yells at us until we come play the game
@janecoaston
apparently that weight was also the only way she could retain her title against her competitor, who was actively cheering for her as she pulled it for the win
Top 25 hitters with at least 150 PA in 2023 by Decision Quality and Batter Adjusted Decision Quality. Freeman and Seager stand out the most, per discussions from last night, in that they're too aggressive for an average hitter but perfectly aggressive given their skillset.
Just met my Navy Seal neighbor and talked to him about his dog Emperor who has a bronze star and found buried bombs in Iraq. Meanwhile I was constantly adjusting my glasses and holding our five pound dog Ellie who was in her pink harness.
ICYMI, bpro ran a piece yesterday highlighting a new decision quality metric, SEAGER. I wanted to see how well it compared to my Decision Quality metric, so I recreated the methodology and confirmed it does outperform my method pretty substantially.
ERA assumes the pitcher was sole contributor to all outs. FIP assumes fielders are the sole contributors to BIP outs. What if there was another metric that apportioned credit appropriately and methodically between pitcher, fielder, and park for settling Cy Young discussions??
another excellent piece here from
@Drew_Haugen
and
@bgrosnick
. pitching_bot found similar results in that most of what folks ascribe to tunneling can be explained by pitch location
this is incredibly exciting for me, i've been reading bpro for years and years and it's surreal seeing my name on an article there. please check it out! it's technical and in depth, but i worked hard to make it accessible to a lay audience as well
And (unsurprisingly given Pitch of the Day results) Aaron Bummer's sweeper is the highest rated pitch of the season with a minimum of 50 pitches thrown
@srbrown70
The logic is rooted in Mike Fast’s seminal article of losing a tick and gaining one - and the effect is amplified as you cross thresholds of velocity bands. Simplest is to just use a modifier rather than rebuild a model imo
@EricLevitz
SSC is like Lyman Stone in blog form: sometimes you learn something really interesting you wouldn’t have read elsewhere, sometimes you groan, but every single time you wish there were 4200 fewer words
Elder is my go-to example of why we should be careful crafting narratives from pitchers whose results don't match their peripherals. I'm the first to admit that every model has blind spots and weaknesses, but baseball is weird and sometimes guys stumble their way into 1.50 ERAs.
Ten simulated seasons for Bryce Elder using outcome probabilities for his pitches from my stuff model. Baseball is weird and random, even for "skill-based" metrics like FIP, so at 200 IP the same guy could have nearly a 1.5+ ERA/FIP range of outcomes by chance alone.
sorry one more nerdery for the rain delay: good catch from Ryan, as some of this could just be measuring the effect of the count: pitchers crank it up with 2 strikes.
I also made the same adjustment described in that tweet for times through the order & the effect was still there
@RyanCloude
great catch, check this out. instead of averaging across raw xwoba for each bucket like before, in this one I averaged across {raw xwoba - [average xwoba in that count] + [average xwoba for that season]}
looking at it this way the results are a lot more mixed
I've shared these before, but this is how well my model performs at predicting the outcome of a pitch, both unadjusted and adjusted for quality of the hitter. the model is really, really good, and includes a ton of info, so it's tempting to say what's left over is noise.
@jaseidler
tbf i had an opposite takeaway from this: adding spray to an xwOBA metric for a hitter without introducing noise is difficult because spray is already so highly correlated with ev and launch angle.
great piece here. the section below reminded me of this piece from Carleton last year, which said that pitchers throw harder when they know they're on a short leash, so it's hard to disentangle that from the decision of when to take them out
Lots of pretty plots, but if there's one I hope sticks with readers it's this one:
A single outing of StuffPro better predicts a pitch's value than a half a season worth of its actual results.
Ozzie's been outperforming his expected wOBA, but only really on ground balls. About half of that gap in expected vs observed performance on grounders is to be expected with his sprint speed.