Should we revise the error stat?

[grimshaw]

https://www.fangraphs.com/blogs/jackie-bradley-jr-and-the-error-rule/

The above article brings up an IMO fascinating point about the fairness of assessing errors on balls with low fielding probability,

Our own David Laurila brought up the play on Friday that clanked off JBJ' glove, which according to statcast, the typical fielder only had a 6% chance of making.

If the ball is a hit 94% of the time anyhow, should a fielder really be to blame when he is in a position to make the difficult play? What about a 25 or 50% probability? I should add that a much greater frequency of errors per play is a lot more for infielders than fly balls for outfielders.

Unlike the win stat discussion, reassessing error frequency would have a real impact when evaluating a pitchers worth (which I'm sure organizations already consider).

Also - holy shit Bradley made that look like a fairly routine play.

 

[AB in DC]

I think the flaws with the "error" stat are fairly well-known. Like a lot of stats, it doesn't tell you nearly as much as people think it does. And as with a lot of stats, "this stat is flawed" does not logically imply "we should change it". It just means that we eventually need to find (or create) a better one.

So if Statcast data (for example) lets us more accurately identify fielders who don't make plays that are made with 99% probability (or 90% or whatever), let's use it come up with a new metric and get it into popular use.

It's no different than with the RBI. It's a stat that existed for a while, but wasn't terribly useful. Eventually, once better stats were in common use, people stopped using it. But that didn't require changing the stat itself.

 

[Cesar Crespo]

Another problem with the error stat is that if a fielder doesn't touch the ball, it's not an error.

 

[Boggs26]

I really like the idea of a "simple" (ie not as complex as ZR or UZR) Fielding start based on statcast probabilities.

The simplest would probably be assigning an Error (or whatever it would be called) to any fielder failing to field a ball that is fielded 95% of the time or more.

A more interesting stat might be something where points are subtracted based on catch probabilities above some cutoff and added when below that cutoff. I'm completely spitballing here, but, for example perhaps the cutoff is 50% so failing to catch a ball that has an 80% catch percentage causes a lot of 30 (or maybe 3 or .3) points. On the other side of the slate making a catch that has a 10% catch percentage would add 40 (or 4 or .4). I haven't looked at fielding data enough to know where this would put most fielders or if the cutoff should be changed. In my mind the cutoffs could be picked in order to make 0 an average rating.

 

[SirPsychoSquints]

I really like the idea of a "simple" (ie not as complex as ZR or UZR) Fielding start based on statcast probabilities.

The simplest would probably be assigning an Error (or whatever it would be called) to any fielder failing to field a ball that is fielded 95% of the time or more.

A more interesting stat might be something where points are subtracted based on catch probabilities above some cutoff and added when below that cutoff. I'm completely spitballing here, but, for example perhaps the cutoff is 50% so failing to catch a ball that has an 80% catch percentage causes a lot of 30 (or maybe 3 or .3) points. On the other side of the slate making a catch that has a 10% catch percentage would add 40 (or 4 or .4). I haven't looked at fielding data enough to know where this would put most fielders or if the cutoff should be changed. In my mind the cutoffs could be picked in order to make 0 an average rating.

Isn't that latter part pretty much what UZR does?

 

[AB in DC]

Remember a good rule of them is that the vast, vast majority of plays have either a >99.9% or <0.1% out probability.

 

[SirPsychoSquints]

I really like the idea of a "simple" (ie not as complex as ZR or UZR) Fielding start based on statcast probabilities.

The simplest would probably be assigning an Error (or whatever it would be called) to any fielder failing to field a ball that is fielded 95% of the time or more.

A more interesting stat might be something where points are subtracted based on catch probabilities above some cutoff and added when below that cutoff. I'm completely spitballing here, but, for example perhaps the cutoff is 50% so failing to catch a ball that has an 80% catch percentage causes a lot of 30 (or maybe 3 or .3) points. On the other side of the slate making a catch that has a 10% catch percentage would add 40 (or 4 or .4). I haven't looked at fielding data enough to know where this would put most fielders or if the cutoff should be changed. In my mind the cutoffs could be picked in order to make 0 an average rating.

I also think statcast does do outs above average, which is what you're describing (I think).

https://baseballsavant.mlb.com/outs_above_average

 

[charlieoscar]

Isn't there a problem inherent with "fielding a ball X% of the time"? Give a fielder an error if he fails in fielding a ball that is fielded 95% or more of the time? What if there is a gust of wind in his face at the last second that means the ball goes off his glove tip instead of settling in the pocket? What if they just resumed the game after a heavy rain and the outfield is soggy? What if a player in the previous half-inning fouled a ball off his foot and doesn't get quite the jump he normally does?

Another thing with errors is whether the right person is charged? For example, in tonight's game, Devers got a throwing error but the throw was there in time. It hopped so it was considered a bad throw but I've certainly seen balls like that scooped by the first baseman. Why shouldn't there be cases where the first baseman gets charged when he fails to scoop and on-line throw?

 

[Lose Remerswaal]

Devers had time to set and make a good throw. Heck, he even looked at the ball in his hand before he threw it, and the throw hit the lip of the grass and took a very low bounce at Moreland. Quite a difficult scoop, especially compared to one that hits the dirt (like most do) that would take a more predictable bounce.

Moreland actually did a great job of getting behind that ball (it hit his leg) and not watching it go careening into right field or foul territory for extra bases.

 

[Reverend]

I kinda only care if an out is made or not--it's not terribly important to me whether or not there was an error involved in the outs not made.

I feel like they measured errors because that's all they had.

 

[BigJimEd]

Another problem with the error stat is that if a fielder doesn't touch the ball, it's not an error.

This is false. A fielder does not need to touch the ball for an error to be charged.

 

[charlieoscar]

Devers had time to set and make a good throw. Heck, he even looked at the ball in his hand before he threw it, and the throw hit the lip of the grass and took a very low bounce at Moreland. Quite a difficult scoop, especially compared to one that hits the dirt (like most do) that would take a more predictable bounce.

Moreland actually did a great job of getting behind that ball (it hit his leg) and not watching it go careening into right field or foul territory for extra bases.

Sorry. Just take it generically. You've certainly seen times when the first baseman failed to scoop a ball he should have fielded.

 

[wade boggs chicken dinner]

Devers had time to set and make a good throw. Heck, he even looked at the ball in his hand before he threw it, and the throw hit the lip of the grass and took a very low bounce at Moreland. Quite a difficult scoop, especially compared to one that hits the dirt (like most do) that would take a more predictable bounce.

Moreland actually did a great job of getting behind that ball (it hit his leg) and not watching it go careening into right field or foul territory for extra bases.

Yes Devers made a good throw but the ESPN announcing crew mentioned that Moreland makes that catch 9 times out of 10, for whatever that's worth.

The error stat doesn't need to be changed; people just need to be informed about its limitations.

As far as Statcast goes, there's an assumption that the out probability it calculates is generally correct. Is that assumption accurate? I know we have talked some about it terms of other plays of JBJ.

 

[Cesar Crespo]

This is false. A fielder does not need to touch the ball for an error to be charged.

They are never charged with one when 2 or 3 fielders decide to let a pop fly drop in for a "hit."

 

[Lose Remerswaal]

True, but a ball thru the wickets, untouched, gets charged