Twins Caught in the Middle of ABS Controversy After Razor Thin Overturn

[jorgenswest]

37 minutes ago, Squirrel said:

I get what you are saying but then why use it. Honestly, if we are going to get into the territory of ‘let the really close ones go’, it gets too subjective. What’s too close? Everyone has a different opinion. Then you start getting into measuring fractions and what fraction still counts as a strike and what doesn’t? I think the system is working. It evens the playing field for everyone. No more unintentional biased calls for or against a specific team or player. 

The system doesn’t know whether a close pitch is a strike or a ball. It isn’t consistent within that range. There will still be bias and possibly stronger motivation for bias when challenges are exhausted.

If they were to immediately change the call for every missed pitch outside 0.4 inches we would have fixed many more calls that are very likely incorrect. There would be no opportunity for bias. Umpires should be very motivated to not get overturned and every overturned call should be confidently incorrect.

There have been 506 overturned calls this year and the vast majority have been 0,4 inches or more away from the plate. There have also 1716 pitches that the system would have overturned had they been challenged. That isn’t a consistent strike zone.

[Richie the Rally Goat]

On 4/12/2026 at 10:01 AM, bunsen82 said:

Its also at point is the System calling the strike.  From the front of the batters box to the glove some of these balls are moving an an inch horizontally.   

The zone is two dimensional at the middle of the plate

[WYOTwins]

2 hours ago, Richie the Rally Goat said:

The zone is two dimensional at the middle of the plate

Not according to MLB website:

https://www.mlb.com/glossary/rules/strike-zone

[Richie the Rally Goat]

10 hours ago, WYOTwins said:

Not according to MLB website:

https://www.mlb.com/glossary/rules/strike-zone

https://www.mlb.com/interactive/mlb-abs-system-explainer
 

“Simplifying to two dimensions

Think it’s hard to call a ball or strike just based on the height and width of the zone? Now throw in the third dimension – of where over the plate the zone should be called.

While the rule book has long officially defined the zone as being a three-dimensional pentagon shape hovering over the plate, ABS testing in the Minor Leagues led to unsatisfying results when using the 3D zone. Therefore, the ABS zone you’ll see in the Majors is a line set at exactly the midpoint of the plate, 8.5 inches deep. This change eliminated some odd outcomes in testing and produced results more in line with the expectations of players and fans.”

[WYOTwins]

Interesting, theoretically that would suggest umpires and ABS are using two different definitions for the strike zone.

 

[laloesch]

On 4/11/2026 at 11:14 PM, jorgenswest said:

ABS had no idea if that was a strike. The technology can’t determine the exact location of the ball. The tolerance for the 99% confidence interval is 0.48 inches. At that close it was probably 50% confident it was correct.

MLB’s graphics would have us believe ABS has found the exact location of the ball. Those graphics should come with the confidence interval. If not 99% then 95% which is .39 inches. Anything less than that and they should stick with the umpire and state that the ABS system was inconclusive.

Isn't it measured with a laser?  It's gotta be more accurate than 0.48 inches.  0.48 inches off in that proximity would be HUGE.  

[jorgenswest]

5 hours ago, laloesch said:

Isn't it measured with a laser?  It's gotta be more accurate than 0.48 inches.  0.48 inches off in that proximity would be HUGE.  

The 99% confidence interval of 0.48 inches is later in the article. That data came from MLB.

[MN Brian]

7 hours ago, laloesch said:

Isn't it measured with a laser?  It's gotta be more accurate than 0.48 inches.  0.48 inches off in that proximity would be HUGE.  

It's all camera based during the games.  Though to determine the confidence interval they would have needed another "source of truth" during testing, I presume this was likely lidar (laser) of some type.

[dxpavelka]

To Matt Strahm's point, the system as ALREADY redefined what counts as a strike.  Look at the box (both the ABS box AND the box that's been on your television screen for years). Then pull out the rule book and look at what the rule is about the strike zone.  The top of the strike zone is supposed to be the mid point between the top of the batter's shoulders and his belt or waist line.  The top of that box is seldom above the belt.  I hate to come down on the side of the batters OR the pitchers but pitchers have been getting squeezed out of the top six inches of the strike zone for a long time.  

[laloesch]

7 hours ago, MN Brian said:

It's all camera based during the games.  Though to determine the confidence interval they would have needed another "source of truth" during testing, I presume this was likely lidar (laser) of some type.

Google AI states that the ABS system is highly accurate with a median margin of error of roughly 0.16 inches (4.1mm).  During 2026 spring training , the system showed 95% confidence in tracking to within 0.39 inches and 99% confidence within 0.48 inches. 

In other words the confidence tracking is not the same as actual measured accuracy which is to within 0.16 inches.  That's actually pretty darn good.   

[MN Brian]

1 hour ago, laloesch said:

Google AI states that the ABS system is highly accurate with a median margin of error of roughly 0.16 inches (4.1mm).  During 2026 spring training , the system showed 95% confidence in tracking to within 0.39 inches and 99% confidence within 0.48 inches. 

In other words the confidence tracking is not the same as actual measured accuracy which is to within 0.16 inches.  That's actually pretty darn good.   

I'm not a statistics major, but I have worked a bit with numbers.  The median margin of error is really not very informative here, the confidence intervals tell a much better story.

If you had 5 ABS measurements which, when compared to a perfect measurement system ("source of truth"), were off by 0.05", 0.08", 0.16", 0.5", & 0.75" your median (middle value) margin of error would be 0.16".  It would also be 0.16" if you had 0.14", 0.15", 0.16", 0.17", & 0.18".  It is challenging to make a prediction about the next ABS measurement with just this value.

The confidence values tell you statistically how likely the next ABS measurement will be within a given error.  So the system is 95% confident (19 times out of 20) that the next pitch will be measured within 0.39" of the "source of truth".  

[Esker Stannum]

I hope we understand that shrinking the zone doesn't increase precision.