Some changes to defensive metrics

E

Enscheff

Well-known member
FG just made a post about how they have improved their defensive metric, UZR. Long story short, they now base hit designations on hang time rather than someone deciding to call it a LD, FB, or GB.

https://www.fangraphs.com/blogs/instagraphs/2018-uzr-and-uzr-update/

They then retroactively applied the results all the way back to 2012. This obviously changed some players' defensive ratings...sometimes by a lot. How much? Here's a sample of some of the largest changes from recent seasons:

2016 Cespedes from -12.1 to -4.0, a 8.2 change for the better.
2016 Pillar from 21.5 to 12.3, a 9.2 change for the worse.

2015 Uribe from -1.3 to 6.7, a 8.0 change for the better.
2015 Kiermaier from 23.2 to 11.8, an 11.4 change for the worse.

We see a lot of the extreme defensive values brought closer to the middle. Inciarte from 11.8 in 2014 to 4.1. Simmons from 22.8 in 2013 to 11.7. Search for Kemp..many of those terrible seasons weren't so terrible after all.

My biggest issue with UZR was always the crazy outlier seasons it spit out for excellent and terrible players. It looks like now they have finally figured out how to reign in those seasons in question.
 
Enscheff said:
FG just made a post about how they have improved their defensive metric, UZR. Long story short, they now base hit designations on hang time rather than someone deciding to call it a LD, FB, or GB.

https://www.fangraphs.com/blogs/instagraphs/2018-uzr-and-uzr-update/

They then retroactively applied the results all the way back to 2012. This obviously changed some players' defensive ratings...sometimes by a lot. How much? Here's a sample of some of the largest changes from recent seasons:

2016 Cespedes from -12.1 to -4.0, a 8.2 change for the better.
2016 Pillar from 21.5 to 12.3, a 9.2 change for the worse.

2015 Uribe from -1.3 to 6.7, a 8.0 change for the better.
2015 Kiermaier from 23.2 to 11.8, an 11.4 change for the worse.

We see a lot of the extreme defensive values brought closer to the middle. Inciarte from 11.8 in 2014 to 4.1. Simmons from 22.8 in 2013 to 11.7. Search for Kemp..many of those terrible seasons weren't so terrible after all.

My biggest issue with UZR was always the crazy outlier seasons it spit out for excellent and terrible players. It looks like now they have finally figured out how to reign in those seasons in question.
Click to expand...

Overall, my eyeball complaint about WAR was that defense was being over-emphasized.

My second complaint was how to really credit the wild shifts in defensive values from year to year of players where there was not necessarily any real thesis for decline setting in.

I guess if this is flattening out defensive outliers on both ends of the spectrum, it might be smoothening out both of my major complaints in effect.



Would this adjustment to the defensive ratings tend to decrease the defensive contributions to WAR overall? IT seems like if you are pushing the outliers back to the pack that's probably going to be the overall effect.
 
Southcack77 said:
Overall, my eyeball complaint about WAR was that defense was being over-emphasized.

My second complaint was how to really credit the wild shifts in defensive values from year to year of players where there was not necessarily any real thesis for decline setting in.

I guess if this is flattening out defensive outliers on both ends of the spectrum, it might be smoothening out both of my major complaints in effect.



Would this adjustment to the defensive ratings tend to decrease the defensive contributions to WAR overall? IT seems like if you are pushing the outliers back to the pack that's probably going to be the overall effect.
Click to expand...

It seems to be pushing outliers from both ends to the middle. I imagine it results in a net change of nearly zero, but that’s pure conjecture in my part.

I still don’t like defensive metrics. The sooner metrics based on Statcast data are publicly available, the better. We need to move past UZR/DRS already. The data is there to do it.
 
Defensive metrics are in their infancy. To treat them as the gospel now is a fools errand.
 
You must log in or register to reply here.
Back
Top