Jack Lambert retweetledi
The Mets have been experimenting in spring training games with calling pitches from the dugout.
espn.com/mlb/story/_/id…
English
Jack Lambert
1.9K posts
Jack Lambert
@jacklambert__
Director of Baseball Operations @DrivelineBB
Cincinnati | Seattle Katılım Kasım 2014
726 Takip Edilen2.4K Takipçiler
Rowdy Tellez getting after it in the lab 🧪
Hard-hit ball launch angle averaged 20° with EV climbing all the way to 113.1 mph. Serious power on display for the 8-year major league vet 💪
English
Jack Lambert retweetledi
Automatic Ball-Strike (ABS) says Miller's pitch to Perdomo missed low by 2.64" inches.
Here's every comparable breaking ball against 6'2" hitters since 2020, with each pitch inching *closer* to the strike zone.
Jack Lambert@jacklambert__
How our models viewed this pitch: 60% chance of swing, 40% chance of take If no swing: 0.5% chance of CS If swing: 63% chance of whiff Model trained on MLB data so not accounting for WBC umpiring, but not surprised our swing decision model had this as an ELITE decision
English
@jacklambert__ @juanmarko @Tstokey So, the model would have output the same for that pitch metrics in the first inning?
English
How our models viewed this pitch:
60% chance of swing, 40% chance of take
If no swing: 0.5% chance of CS
If swing: 63% chance of whiff
Model trained on MLB data so not accounting for WBC umpiring, but not surprised our swing decision model had this as an ELITE decision
Just Baseball@JustBB_Media
MASON MILLER STRIKES OUT GERALDO PERDOMO! Team USA is heading to the championship! 🇺🇸
English
@Brayanmnz_ @juanmarko @Tstokey Only game state knowledge is balls and strikes. No knowledge of outs, runners on base, or the batter batting (other than it being a LHH)
English
@jacklambert__ @juanmarko @Tstokey But, does the model take into account the game state?
Or what does the model takes into account?
English
@AGPatriot1776 This model is on this specific slider from Miller!
English
Here are the outcome probabilities with resulting win probability. With these calcs:
A take has an expected WP of 19.5%
A swing has an expected WP of 6.5%
There would need to be a 67% chance of called strike for the EV of a swing to meet or exceed the EV of a take.
Jack Lambert@jacklambert__
How our models viewed this pitch: 60% chance of swing, 40% chance of take If no swing: 0.5% chance of CS If swing: 63% chance of whiff Model trained on MLB data so not accounting for WBC umpiring, but not surprised our swing decision model had this as an ELITE decision
English
@jacklambert__ You can let me know where in the rule book that describes this strikezone you speak of cause that is not what it says..
English
If you take the ball trajectory you can see when it is in Perdomo's rulebook strike zone (colored in red: roughly when it's 11 to 3 feet from it mattering).
Also, strike zone (by ABS standards) is measured as a 2D plane at the middle of the plate, as visualized here.
george@LoveTheseMets
@jacklambert__ @RobertStock6 Maybe I’m just not smart But looking at this - is the ball not at Knee level when it gets to the front of the plate? What am I not seeing
English
@juanmarko @Tstokey 98.5% chance of CS given take, 17% chance of whiff if swing; viewed as a very bad swing decision overall
English
@jacklambert__ @Tstokey Jack, what about the fastball that Geraldo let go by on the 1-1 count?
English
@jacklambert__ “The ABS system used in MLB tests and the minor leagues tracks the ball in 3-D with Hawk-Eye cameras and determines whether the pitch crosses the strike zone when it reaches the front plane of the plate.”
English
Middle of the plate actually:
"The ABS zone is set as follows: The width is 17 inches, identical to home plate. The top of the zone is set at 53.5% of a player's measured height without cleats. The bottom is set at 27% of the player's measured height. The strike zone is captured as the ball passes through the middle of the plate, not the front."
baseballsavant.mlb.com/abs
English
@jacklambert__ ABS is front of plate. Traditional strikes zone is a 3d cube.
English
@BlackCoffeeSobr That's the probability of any team winning with first and third, two outs in bot 9, down 1.
Definitely doesn't account for the fact that the DR lineup had Tatis coming up to the plate.
BUT also doesn't account for the fact that the US had Mason Miller on the mound
English
@jacklambert__ So you’re saying if the ump called that a ball the DR win probability is only 19.5% with the runners on the corners and two outs? That’s actually lower than I anticipated. My head had it more like 60% US / 40% DR win probability in that situation. But that was just my gut.
English
@BlackCoffeeSobr Yep - see here:
x.com/jacklambert__/…
Jack Lambert@jacklambert__
Here are the outcome probabilities with resulting win probability. With these calcs: A take has an expected WP of 19.5% A swing has an expected WP of 6.5% There would need to be a 67% chance of called strike for the EV of a swing to meet or exceed the EV of a take.
English
@jacklambert__ Do your models predict the probability of winning? Like if they had called that a ball what were the odds that the DR pulls out the win? I gotta believe the US is still a fairly solid favorite in that situation.
English
@BoltsandHalos85 Based on 3 years of data taking into account all pitch metrics (release point, velo, movement, location, etc) so we aren't *literally* passing in locations 3 feet from the plate, but very much get at similar trends
English
@jacklambert__ Thanks for the reply. Maybe I should have thought more critically. You’re saying that your model says this would be called a ball 99.5% of the time based on 3 years of MLB data when it was in the strike zone 3 feet from the plate? You’re going to blow my mind if you say yes lol
English
@clarkmunroe_ 100% agree on A and B - we have this as a 63% chance of whiff if Perdomo swung lol.
Agree this has to be an insanely difficult call to make as an ump, definitely something I couldn't do. Just contextualizing that almost no MLB umps call this a strike... 0.5% was our prediction
English
This. The reaction this pitch has been ridiculous.
A) that was a NASTY pitch
B) the late dip on it was NASTY
C) people are too entitled by digital strike zones and charts after the fact making them feel better than the live umpires in the moment with a split second decision.
Jack Lambert@jacklambert__
If you take the ball trajectory you can see when it is in Perdomo's rulebook strike zone (colored in red: roughly when it's 11 to 3 feet from it mattering). Also, strike zone (by ABS standards) is measured as a 2D plane at the middle of the plate, as visualized here.
English
Rulebook is changing to this definition:
"The ABS zone is set as follows: The width is 17 inches, identical to home plate. The top of the zone is set at 53.5% of a player's measured height without cleats. The bottom is set at 27% of the player's measured height. The strike zone is captured as the ball passes through the middle of the plate, not the front."
baseballsavant.mlb.com/abs
English
@CompositeTwins @jacklambert__ So the ABS zone is not a true representation of the strike zone as defined in the MLB rulebook?
Did they change the definition from it being 3 dimensional? I hate this lmao
Rulebook clearly states strike zone is a 3D space
English