OOTP Developments Forums - View Single Post - An example of why I don't think DIPS will reflect pre-1960 pitchers accurately

gmo · 06-15-2004, 01:35 AM

Supposedly the DIPS basis research covered the entire 20th century, and it is not simply a case of taking today's data and imposing it on the past. Rommel - see page here at b-f.com - is an interesting case indeed with so few K's. For the most part in his career his yearly BABIP was lower than DIPS would predict, and in several cases much lower. The metric I used was (Hits-HR)/(Inn*3+Hits-HR-SO), and gives values a little to low, but I figure him as 25-30 points below league average BABIP for his career. The year you show was his year most better than league-average BABIP, 50+ points.

That said, his other numbers do not look that far off from what I would probably expect - generally at least a few more Hits than IP, good ERA because few HR and good control (generally <3BB/9Inn). You would not want him to have a very good Avoids Hits rating since he had more Hits than IP would you? Another thing to consider is what kind of park/defense he pitched in/with. The park was apparently more of a hitter's park, which is interesting. I have no idea about his defense - maybe one of the history expert knows the 1920's A's. Along with defense you have to consider whether he was more flyball or groundball and if the defense fit his pitching manner well.

All pitchers' results cannot be explained within the error bounds of DIPS, and Rommel definitely appears to be an outlier. But I think the system overall represents most pitchers very well with a minimum of variables, and overall better than the old system. Don't think I'm picking on you - just taking the flip side because I think the discussion can be interesting. A lot of the DIPS talk here a while back when it was the hot topic was how much effort, if any, should go into replicating the extreme outliers. The extreme cases may grab attention, but I think they are few enough that the current system is fine considering how good it appears to be for the majority of cases.

But I'm far from the final authority on this, especially since I'm a fictional league guy.

06-15-2004, 01:35 AM	#2 (permalink)
gmo Hall Of Famer Join Date: May 2002 Location: Grand Forks, ND Posts: 2,788 Thanked 78x in 46 posts	Supposedly the DIPS basis research covered the entire 20th century, and it is not simply a case of taking today's data and imposing it on the past. Rommel - see page here at b-f.com - is an interesting case indeed with so few K's. For the most part in his career his yearly BABIP was lower than DIPS would predict, and in several cases much lower. The metric I used was (Hits-HR)/(Inn*3+Hits-HR-SO), and gives values a little to low, but I figure him as 25-30 points below league average BABIP for his career. The year you show was his year most better than league-average BABIP, 50+ points. That said, his other numbers do not look that far off from what I would probably expect - generally at least a few more Hits than IP, good ERA because few HR and good control (generally <3BB/9Inn). You would not want him to have a very good Avoids Hits rating since he had more Hits than IP would you? Another thing to consider is what kind of park/defense he pitched in/with. The park was apparently more of a hitter's park, which is interesting. I have no idea about his defense - maybe one of the history expert knows the 1920's A's. Along with defense you have to consider whether he was more flyball or groundball and if the defense fit his pitching manner well. All pitchers' results cannot be explained within the error bounds of DIPS, and Rommel definitely appears to be an outlier. But I think the system overall represents most pitchers very well with a minimum of variables, and overall better than the old system. Don't think I'm picking on you - just taking the flip side because I think the discussion can be interesting. A lot of the DIPS talk here a while back when it was the hot topic was how much effort, if any, should go into replicating the extreme outliers. The extreme cases may grab attention, but I think they are few enough that the current system is fine considering how good it appears to be for the majority of cases. But I'm far from the final authority on this, especially since I'm a fictional league guy.