Perhaps misguided, but I wonder if guys like Bradish and Sonny Gray whose arsenals lean so heavily on glove side movement see benefit from the relatively unexpected hard armside ambush.
Could be! I think it’s also just natural that if you have heavy glove side, you probably cut your fastball. And cut fastballs are usually better versus lefties, therefore you need something for righties - sinker.
Can someone get our guy Lance a grant and data access to research this and present at Sloan? I love this kind of breakdown, I hope you make more like it.
are there any "weird" pitches other than sinkers? like other outlier pitches that dont look impressive on a public model but are able to still be effective, maybe due to late seam shifted effects in ball flight?
Vertical approach angle is big. If a guy is throwing an “average” four seam FB shape with 14IVB & 7 vertical, but it comes from a lower armslot with good extension it is a good pitch even though the movement and velo could say no
Splitters are still pretty weird to me. They’re almost universally effective, spin far lower than anything else and hitters seem to always chase them at above average rates. In terms of things that buck the model, it’s likely outlier release traits where models don’t really have comps and therefore just don’t know how to grade it (mostly weird shapes from low releases, but I wonder whether we’ll see some high release popularity soon).
@@Mason18718this is totally true, but I would argue that most (all) models are capturing this effect. Release and movement are pretty integral components. So it’s already kind of baked in? I think he means weird as in the models are indifferent and it universally carves.
Despite some improvement, models still have some trouble with traditional changeups as well, don’t they? I wish I could look under the hood at something like Eno’s Stuff+ so I could see for myself what’s going on. As someone with a marketing analytics background, I can’t help but wonder if there might be some methodological issues that could be addressed via alternative statistical techniques, or maybe even alternative model-building approaches/philosophies. Sometimes, we stick to doing things a particular way because they’ve generally worked well enough, we’re comfortable with it, etc., and handwave away issues even though they might pop up over and over again, but we might be better served to reexamine our approach and maybe try something different. Maybe I’m just wishcasting (or wishsomething-ing), and there really isn’t a better approach, particularly on the public side given given the available data, but I also don’t know to what extent people have tried. (I believe Cameron Grove took a different approach with PitchingBot, which probably explains why it sometimes differs wildly from Stuff+, but the existence of those differences seems instructive in and of itself.
The cause of late movement is probably caused by the change of direction when the ball is travelling. So for example as the ball drops, the air rushes more in the bottom of the ball which might mean a seam might now be exposed to the rushing air thus causing additional seam-shifted wake movement
Sure, but as I mention in this video, we don’t really know when that effect is happening. And how it changes in magnitude and “unexpectedness” based on orientation.
We’ve seen a few 3700+ but no 4000 I would imagine the ceiling is sub 4K, but that’s more a guess Velo and spin are correlated. So perhaps we see a huge spin hard thrower soon that can get to 3800+
Perhaps misguided, but I wonder if guys like Bradish and Sonny Gray whose arsenals lean so heavily on glove side movement see benefit from the relatively unexpected hard armside ambush.
Could be! I think it’s also just natural that if you have heavy glove side, you probably cut your fastball. And cut fastballs are usually better versus lefties, therefore you need something for righties - sinker.
Can someone get our guy Lance a grant and data access to research this and present at Sloan?
I love this kind of breakdown, I hope you make more like it.
Hahaha likely something a lot of MLB teams have (kinda) figured out. I’m just curious what their findings are !
are there any "weird" pitches other than sinkers? like other outlier pitches that dont look impressive on a public model but are able to still be effective, maybe due to late seam shifted effects in ball flight?
Vertical approach angle is big. If a guy is throwing an “average” four seam FB shape with 14IVB & 7 vertical, but it comes from a lower armslot with good extension it is a good pitch even though the movement and velo could say no
Splitters are still pretty weird to me. They’re almost universally effective, spin far lower than anything else and hitters seem to always chase them at above average rates.
In terms of things that buck the model, it’s likely outlier release traits where models don’t really have comps and therefore just don’t know how to grade it (mostly weird shapes from low releases, but I wonder whether we’ll see some high release popularity soon).
@@Mason18718this is totally true, but I would argue that most (all) models are capturing this effect. Release and movement are pretty integral components. So it’s already kind of baked in?
I think he means weird as in the models are indifferent and it universally carves.
Despite some improvement, models still have some trouble with traditional changeups as well, don’t they? I wish I could look under the hood at something like Eno’s Stuff+ so I could see for myself what’s going on. As someone with a marketing analytics background, I can’t help but wonder if there might be some methodological issues that could be addressed via alternative statistical techniques, or maybe even alternative model-building approaches/philosophies. Sometimes, we stick to doing things a particular way because they’ve generally worked well enough, we’re comfortable with it, etc., and handwave away issues even though they might pop up over and over again, but we might be better served to reexamine our approach and maybe try something different. Maybe I’m just wishcasting (or wishsomething-ing), and there really isn’t a better approach, particularly on the public side given given the available data, but I also don’t know to what extent people have tried. (I believe Cameron Grove took a different approach with PitchingBot, which probably explains why it sometimes differs wildly from Stuff+, but the existence of those differences seems instructive in and of itself.
First
The cause of late movement is probably caused by the change of direction when the ball is travelling. So for example as the ball drops, the air rushes more in the bottom of the ball which might mean a seam might now be exposed to the rushing air thus causing additional seam-shifted wake movement
Sure, but as I mention in this video, we don’t really know when that effect is happening.
And how it changes in magnitude and “unexpectedness” based on orientation.
Two things have we ever seen a 4000 rpm pitch. And what do you think is the highest rpm possible for a pitcher.
We’ve seen a few 3700+ but no 4000
I would imagine the ceiling is sub 4K, but that’s more a guess
Velo and spin are correlated. So perhaps we see a huge spin hard thrower soon that can get to 3800+