Wilson pitching change grade

[desbcoach]

Found this interesting considering the amount of discussion on this board and complaint both directions

[Captain 97]

How do you even begin to quantify that?

[Donn Beach]

TMBR basically represents the pitcher’s score against next three batters. So when Seattle makes a pitching change, the new pitcher is 14.4% more effective than the previous pitcher and 27.4% more effective than other options in the bullpen against the next three batters. They are also 35.1% more effective after the first three batters.

[Donn Beach]

How It Works
The framework evaluates each bullpen decision by comparing the expected outcome of the actual move to the outcomes of other realistic options - based on batter-pitcher matchups, game situation, and bullpen availability. These comparisons are powered by machine learning models trained on historical plate appearance level data and simulations that estimate the potential impact of different decisions. The result is a framework that helps teams (and fans) assess whether a decision aligned with the best available options, not just whether it led to a good outcome

https://www.uramanalytics.com/pitching- ... -framework

[Sibelius Hindemith]

That's a good attempt to quantify and evaluate a complex decision making process, but the main complaint with Wilson regarding pitching decisions has been that he generally leaves starters in too long after they show signs of losing it and that doesn't seem to be accounted for in this analysis. Also it says (in their site linked below) that they use simulations to generate expected runs but i couldn't find an explanation of what goes into their model. For instance, when making the comparison between how the first reliever pitched and how the starter would have pitched against the same batters, does their simulation attempt to take into account how the SP was (actually) pitching in his final inning and how much do pitch count and times through lineup factor into the calculations.

https://lookerstudio.google.com/reporti ... 33acEAIMMw

Another problem with evaluating the decision-making is that every bullpen is different and those differences affect how difficult the decision is when making a pitching change. If a team is carrying some really bad arms (like a Saucedo or Legumina) at the front end that are used less often than the other BP options (because of the high starter innings pitched), won't that skew the computation of "expected runs vs bullpen pitcher" towards being a more favorable decision when they are passed over in favor of a Bazardo or Speier? And if they are considered "available in the bullpen" late in close games where they are never used the impact on the decision evaluation could be huge given the disparity in their expected runs and what a Brash or Munoz actually allowed.

That isn't to say (in my opinion) that their analytical methods aren't advanced and useful at revealing something, but i don't think they take the full picture into consideration. Maybe Wilson has done a great job deciding which BP arm to bring in, but did those decisions come at the best time particularly when pulling the SP? And were the decisions as hard to make given the talent disparity between BP options like Saucedo/Legumina/Vargas/Thornton and Bazardo/Speier/Brash/Munoz? More balanced bullpens probably require more skill to manage because they're given less to a formulaic approach.