Two days before the playoffs start in Yokohama and Fukuoka, I’ve come to a fork in the road. I was about to buckle down with my season-ending valuations, when I decided to test out a new model for park adjustments and realized I know precious little about how to make really good ones.
I suspect that thousands of people understand the following problem better than I do, so if you are one of them, please help, because I am stuck.
All along I’ve basically followed the logic that the degree to which a park increases or decreases run scoring is the relationship between runs scored and allowed per inning at home, and runs scored and allowed per inning on the road. It’s vastly more complicated than that, but that’s the basic idea.
To make an absurdly long and painful story short, I observed that if you create a model with a group of identical teams of known quality playing in parks whose affect on offense is also known, the standard park factor formulas give you fairly inaccurate estimates of how much those parks actually affect scoring. And if one’s park factors and park adjustments are off by more than a little, one’s estimates of player value can be off by quite a bit. Essentially, the more extreme a park is, the more inaccurate its park adjustments will be, and this is more so for hitters’ parks than those that favor pitchers.
In my overly simple knucklehead model, I created a six team league in which each team would score and allow exactly 3.5 runs per nine innings in a neutral park. The teams are now a known quantity. Their offenses and defenses are known. The only difference between them is their parks.
The six parks influence offense as follows: Two are neutral, one will see five runs scored per 18 offensive innings, another six, another eight and another nine. Let’s say each team plays 10 games, five at home, five on the road. The games played by the team in the most extreme pitchers’ park would see 62 runs, the next more pitcher friendly park’s club would see 66 runs scored, the two neutral park clubs’ games would have 70 each, while the two teams in hitters’ parks would see 74 and 78 runs scored.
Not surprisingly, park factors calculate the neutral parks as perfectly neutral. The others work out as follows:
|Affect on runs scored in park||Runs at home||Runs on road||PF|