This article is part of our The Z Files series.
One of my in-season duties for the site is generating weekly pitching rankings. The process is transparent, having been detailed in the space over the years. In brief, the starting point is the season-long expectation for the pitcher. This is then adjusted for park and opponent and ranked with all the other expectations.
Each offseason, the process is reviewed. Is the projection properly adjusted by quality of opponent? Do the rankings accurately reflect how the weekly expectation will affect end-of-season numbers? Are changes in the pitcher's skill set adequately captured by the rest-of-season algorithm?
The third of the above will be the focus of today's discussion. As outlined in this space, I have developed a formulaic means of altering my initial projection, concentrating on skills rather than outcomes often out of the pitcher's control. A pitcher owns his in-season skills at different rates. Each of these are regressed to initial expectations based on how rapidly they become real. It usually takes five or six starts to really move the needle.
I trust the system; it's served me well over the years. That said, in today's offensive environment, streaming pitching is more integral than ever. In addition, there's more information available, with smart individuals helping to decipher the data. As such, I decided to be more liberal this season when it comes to overriding or perhaps accelerating changes to the rest-of-season expectation. However, I made myself promise to wait three starts, and manually adjust solely in the event of something tangible