[IBL] IBL Range Ratings

[Noel Steere]

Thanks for the mini-primer at the end, Sean!

I mean that sincerely; I noticed the change in data source when you announced it two years ago, but didn’t have a good idea of how IE was utilized, leaving me to guess based on comparing players (how does Segura stack up to Lindor in these probability buckets?).

Any thought of moving outfield range towards Statcast?  That would seem to be the ultimate measure of defense.

Sent from my iPhone

> On Mar 27, 2018, at 9:59 AM, Sean Sweda <sweda at ibl.org> wrote:
> 
> The primary data source for range ratings was changed to the Inside Edge data published on Fangraphs, as announced here:
> 
> https://lists.ibl.org/pipermail/members/2016-February/001551.html
> 
> Fangraphs UZR range runs (RngR) is also used as a secondary data source, primarily for validation but also because using it allows some in-season visibility given that the IE data is not synthesized by Fangraphs into an aggregated metric.
> 
> When the data source was switched the mapping methodology was also changed.  Plays made above/below average are converted into a runs saved metric and then mapped into a range rating based on the number of expected runs saved based on projected distribution of IFR/OFR plays.  For example, given the average expected distribution of IFR/OFR plays the difference between A (100%) to K (0%) range at 3B would be approximately 44 runs.  That means in order to gain an A range a 3B would need to grade out over +20 runs per season.  For positions that have more plays (e.g. SS, CF) this number would need to be even higher.
> 
> Jace Peterson got a good range rating because he made 1/1 of plays rated "unlikely" by IE and 2/2 rated "even" by IE.  An "average" result would probably be about 1/3 of those plays.  When you pro-rate the runs saved from his 66 innings at 3B to a full season you get an very large value.  This is pulled back down by combination with the secondary data source (UZR) and the capping mechanisms that are applied to low playing time players, resulting in "D".
> 
> My intention has been to create a detailed write-up on the blog, but I've only got so much time...
> 
> TLDR:
> 1) range is based on IE (and UZR), UZR is a decent approximation in-season
> 2) A/B ratings are very rare now
> 3) small sample size "flukes" are mitigated but still exist
> 
> 
> Sean
>