Addendum to article MA-13 in the recent series, the article "True Levels".
The "True Levels" approach was fun because (a) the method worked twenty times better than I expected it to work, and (b) it yielded a list of pitcher rankings which obviously "works" to a certain extent, listing many of the greatest pitchers ever as the greatest pitchers ever. I can easily see how I could refine this approach to make it work better, but I did not choose to do so because I thought that there was a roadblock there that I could never get around. The roadblock is that the method depends on won-lost records, and won-lost records are "corrupted", by team support. Pedro Ramos may have been a #1 starter in some seasons, but he led the league in losses anyway.
But it has occurred to me that there actually may be a way around this roadblock. We know that when a #2 starter on a good team faces a #2 starter on a bad team, the #2 starter on the good team is more likely to win. This is the problem that we are trying to get around; the fact that the system is to some extent corrupted by the influence of the team.
But suppose that we set up the problem like this: suppose that a #2 starter on a team that finishes 96-66 faces a #2 starter on a team that finished 66-96. How often would each team win? We know that when a 96-66 team faces a 66-96 team, the good team will win 67.9% of the time. But suppose the starting pitchers are more or less even—that is, they are both #2 starters.
With work, you can get an answer to that question. And if you have an answer to that question, you can find a way to work around the roadblock. If you can say that this pitcher was 28-38 against #4 starters, but that is equivalent to 32-34 on a .500 team, then you can block out the team influence, and move forward.
It’s a big project, and I’m not going to do it, at least now, but there would be value in having that knowledge. There would be many things you could do with it, if you knew.
When you do a study you always find some ways that you could have done it better. If I was re-doing the study, I would use a different system to group the pitchers. One thing I would do different is not be quite as rigid about having the right number of #1 starters in each league. I would use a dual system that identified the pitchers by leagues, but also by quasi-fixed standards; in other words, a pitcher with this set of numbers is a #1, regardless of the league. Melding the two together would make a stronger system. You might have seven #1 starters in one league and nine in another, or even 6 and 10—because that’s real life, I think.
I also had a couple of additional thoughts about the Murry Dickson article.
· Regarding the Pittsburgh team for which Murry Dickson pitched for several seasons. I was just re-watching "On the Waterfront", a movie made in the heart of that period (1954). At the start of the movie Joey Doyle is murdered. Thirteen minutes into the movie, Doyle’s father gives a friend Joey’s windbreaker. "Thanks, Pop," says the friend. "Mine’s more full of holes than the Pittsburgh infield."
· Re-reading the article, I realize that I should have directed more attention to Dickson’s age. Despite outstanding performance in the high minor leagues, Dickson did not reach the majors for more than one game in a season until he was 25 years old. Then he was an extra man for two seasons (1942-1943), making only 7 starts each year, and then he was in the US Army for two seasons (1944-1945). In 1946 he started the season back in the same role he had had in 1943, an extra man. He did not make a start that season until June 9, having made 17 relief appearances with a 2.28 ERA. By the time he moved into the rotation, he was two months short of his 30th birthday.
So here’s a guy who didn’t become a major league starter until he was almost 30, pitched most of his career with terrible teams—and won 172 games. That’s a pretty impressive set of facts, isn’t it?