Why I don't fully trust Catch Probability

[Frobby]

Watch these plays and tell me what you think.

5%: Colton Cowser 4/26/24 4.5 seconds, 73 feet

20%: Colton Cowser 4/2/24 5.1 seconds, 84 feet

45%: Colton Cowser 4/24/24 4.6 seconds, 74 feet

50%: Anthony Santander 4/6/24 3.5 second, 49 feet

50%: Cedric Mullins 4/6/24 3.6 seconds, 51 feet

55%: Anthony Santander 4/20/24 4.4 seconds, 73 feet

65%: Cedric Mullins 4/15/24 4.2 seconds, 67 feet

70%: Anthony Santander 4/1/24 3.5 seconds, 45 feet

70%: Cedric Mullins 4/14/24 4.5 seconds, 74 feet

75%: Cedric Mullins 4/22/24 4.8 seconds, 80 feet

80%: Colton Cowser 4/2/24 5.2 seconds, 91 feet

85%: Cedric Mullins 4/15/24 3.9 seconds, 53 feet

85%: Anthony Santander 3/28/24 5.0 seconds, 83 feet

85%: McKenna's missed catch 4/26/24 (time and distance not available)

[Fiver6565]

I only watched the first three, and then the 85% Mullins catch.  What is the highest % these can go - is it 100?  I'm not sure how that Mullins catch wasn't 100 if so.  It was a lazy pop fly.

The Cowser catches at the top of the list probably required a good jump, but those low % numbers are a bit much.  Especially the 5% one.  How is that a 5% catch while the Mckenna catch is 85?  They aren't that dissimilar other than one being in RF and one in LF. 

Maybe the deep wall somehow plays into the numbers on LF balls?

[Fiver6565]

*the McKenna NON-catch, I obviously meant above

 

[Warehouse]

23 minutes ago, Frobby said:

Watch these plays and tell me what you think.

5%: Colton Cowser 4/26/24 4.5 seconds, 73 feet

20%: Colton Cowser 4/2/24 5.1 seconds, 84 feet

45%: Colton Cowser 4/24/24 4.6 seconds, 74 feet

50%: Anthony Santander 4/6/24 3.5 second, 49 feet

50%: Cedric Mullins 4/6/24 3.6 seconds, 51 feet

55%: Anthony Santander 4/20/24 4.4 seconds, 73 feet

65%: Cedric Mullins 4/15/24 4.2 seconds, 67 feet

70%: Anthony Santander 4/1/24 3.5 seconds, 45 feet

70%: Cedric Mullins 4/14/24 4.5 seconds, 74 feet

75%: Cedric Mullins 4/22/24 4.8 seconds, 80 feet

80%: Colton Cowser 4/2/24 5.2 seconds, 91 feet

85%: Cedric Mullins 4/15/24 3.9 seconds, 53 feet

85%: Anthony Santander 3/28/24 5.0 seconds, 83 feet

85%: McKenna's missed catch 4/26/24 (time and distance not available)

McKenna’s missed catch was 3.9 seconds, 53 feet (need to look on Toro’s statcast page in batter’s perspective fielding to see it).  It was also too much movement toward first base to get extra credit for “going back”.

[dystopia]

You shouldn’t blindly trust any advanced stat, especially ones that manipulate other more basic stats, which is the vast majority of them. 

[sportsfan8703]

It was hit right at him, he got a terrible read and jump. 

[Frobby]

6 minutes ago, sportsfan8703 said:

It was hit right at him, he got a terrible read and jump. 

Interesting that Statcast (per @Warehouse) has wildly different numbers (3.9 seconds, 53 feet) than this bot does (3.4 seconds, 31 feet).  

[sportsfan8703]

5 minutes ago, Frobby said:

Interesting that Statcast (per @Warehouse) has wildly different numbers (3.9 seconds, 53 feet) than this bot does (3.4 seconds, 31 feet).  

It would be nice to know what defensive tool/site the analytics department uses. We’ll see with McKenna. Options are out there. 

[AlbNYfan]

There was a diving, foul ball catch in front of the 3rd base dugout by AR a few weeks ago...statcast gave it a high probability of catch number 80% ????

McKenna?? not an easy catch, but if your forte is defense, it should have been made. 

[Moose Milligan]

1 hour ago, sportsfan8703 said:

It was hit right at him, he got a terrible read and jump. 

I can't remember who was doing the color commentary that game, ( I think it was McDonald) who said that the hardest ball an outfielder will have to catch is the one hit directly at them on a line.

Anyone who grew up playing baseball and played the outfield knows this is true.  I'm not sure how catch probability works this into the equation because I don't think it can.  

I'm not making excuses for McKenna, he should have caught it...especially if he's considered a defensive replacement.  But I also don't believe that catch probability was 85%.

[ChosenOne21]

All I have to say after watching those is man, Cowser's routes are something else. I don't think that guy can take a straight line to a ball.

[Tony-OH]

I do think that statcast needs some tweaking. While it's promising and I love the concept, there is no doubt it's kicking out some wonky catch percentages. If you check out the graphs of the catches, they graph correctly to 1-5 star catches, but the catch percentages can be all over the place on some balls. I also have some concerns over the infield OAA.

I don't know if MLB is working to fix these or not since they longer give you an opportunity to reach out to them. I wonder if the issues can be traced to the Hawkeye system?

[Tony-OH]

24 minutes ago, ChosenOne21 said:

All I have to say after watching those is man, Cowser's routes are something else. I don't think that guy can take a straight line to a ball.

His -0.8 ft/sec for routes confirms this and backs up what I saw in the minors. Cowser catches everything he gets to and his acceleration and top running speed on the go allows him to catch a good amount of harder balls, but he rarely looks smooth doing so.

[baltfan]

Please tell me how you find these.  I can't seem to find individual plays with catch probabilities.  

[Warehouse]

1 hour ago, baltfan said:

Please tell me how you find these.  I can't seem to find individual plays with catch probabilities.  

Actually, the catch probability seems to have been updated to 44% based on wall proximity.

Go to Ryan McKenna’s statcast page.  You’ll see four boxes that say Statcast, Standard, Splits, and Game Logs.   Below that is a dropdown box for Batting, Fielding, Running, etc. Select Fielding.  Scroll down to Statcast Fielding Breakdown.  Make sure Fielder’s Perspective is selected. Click on the one dot.  The catch probability, hang time, and distance needed will pop and you will be redirected to a clip of the catch.  This is from the mobile web interface so it might be different from a desktop.