Friday, May 25, 2018
Descriptive v Predictive
?Based on these polls that I ran:
Where in every stat except one, the Predictive was the choice in 65-95% of the time, and FIP is close to 50/50, I can only conclude thusly: because FIP was (a) made to look like ERA and (b) does a better job than ERA itself, for a single season anyway, it's being treated as a Predictive metric.
Even though I have FutureFIP and there is a Predictive FIP metric out there that does BETTER THAN FIP in predicting next year's ERA.
The reality is that the weights of FIP have been set to match the DESCRIPTIVE. It is a descriptive metric. The metric exists to capture a COMPONENT of pitching. Just like OBP captures a component of hitting.
I did my best to explain that on Twitter, and I thought I gave enough illustration and prompts to kind of level the playing field. But when it came time to vote, the Twittersphere remained divided. Maybe I tipped the scales from a 30-70 to a 50/50. Nonetheless, there's more work to be done.
The main point is thusly: if I WANT to create a descriptive metric, and I want to use SO, BB, HB, HR, IP: am I allowed to ONLY if I do NOT make it look like ERA? Or is 50% of the readers out there simply telling me that if I try to combine those in ANY WAY that looks like ERA it can never be descriptive. Even if I show it game by game? Here's Scherzer: