Were park factors considered?
No, I didn’t account for the different scoring environments of the various ballparks, at least not yet. I plan to give this more thought later.
What about the concept of pitchers “pitching to the score” with regard to its impact on this algorithm?
Do pitchers pitch to the score consistently enough to impact statistics in a measurable manner? I’m not sure this has ever been proven or can be proven. One could argue that practically all pitching statistics are vulnerable to this concept. That being said, yes, a pitcher could be slighted with this algorithm if he’s ahead by 12 runs and allows 5 ERs over 7 innings because of the slack his offense has given him. If he’s a Roy Halladay, that would hurt his seasonal WE-based grade. Arguably, Roy would have grinded more in a tie ballgame.
On the other hand, including run support defeats the stated purpose of the entire Ace Factor methodology, which is to focus only on the pitcher’s work and not his team. At least for this algorithm, the benefit of this isolation is that teams that are offensive juggernauts won’t make #3 starters look like aces, and anemic offenses won’t make aces look like #3 starters. From the sWEZ historical ranking results, I think the algorithm sans run support is on the right track as a true measure of ace-ness.
This algorithm doesn’t use Win Expectancy in a way that it was intended.
I agree. The Ace Factor algorithm doesn’t promote the use of Win Probability as a means of quantifying the impact of play-by-play events on the outcomes of games. There are other tools that do that. Nonetheless, I think I’ve shown here that tagging a pitcher’s outing with the notion of “win expectancy” relative to a proposed outcome based only on his performance is a valid app for grading starting pitchers.
It’s arguably easier to achieve a higher sWEZ score in the modern era because starters aren’t taxed as much as they used to be, so it’s easier for outliers to stand out (e.g. Pedro Martinez in 2000–sWEZ: 1.003, Sandy Koufax in 1966–sWEZ: 0.805).
We’ll get to this and other ideas of outlier relevance when I re-run the data for hWEZ and compare the results with the sWEZ results. But keep in mind that there are multiple forces at play not explicitly factored in that will impact the WEZ scores. A starter going every fifth day might seem to have an advantage over earlier generations because of the additional day of rest. But the advent of the relief specialist—long and short—has also kept pitchers from going deep into games like they used to—and the late innings is where the higher WE numbers are. It’s hard to say which force had a greater impact. So as with most statistics you take the numbers as indicators, not gospel. Pedro’s 2000 sWEZ score doesn’t prove that he was more dominant in his era than Koufax was in his—it merely suggests that he was based on the measurement process.
