The Weekend Report comes back as the return of a classic. There are a lot of new people on the site since last year, so I'll give a brief introduction of what I try to do with the Weekend Report. I try to focus on the big leagues because most of the other things that go on at Brewerfan focus on the minor leagues. I award a player of the week, who receives nothing other than my congratulations. Despite my dislike of the signing, last year's most frequent award winner was none other than Eric Young. I do a fair amount of analysis from a statistical (dare I say Moneyball?) stand point. What I really try to focus on is bringing all of these new complicated metrics to those individuals that didn't take three semesters of calculus in college (or those that got through because of the miracle of alcohol). That statement is worth repeating: I try to make a wide variety of new statistical concepts understandable to average baseball fans. This is not because I think they are all great, but to truly evaluate how useful a particular metric is you need to understand its strengths and weaknesses. I covered a number of statistics last year and I invite you to take a look at the archives for an increased understanding.
This week I'm going to take a look at what the folks at Baseball Prospectus
call third order winning percentage. A slightly different take on the same idea
can be found at the Diamond Mind (http://www.diamond-mind.com/articles/tmeff02.htm)
website. I encourage you to read it. I referenced that article because it tries
to get at the same issues as BP does with its third order winning percentage.
Third order winning percentage's name, while a bad name for aesthetic reasons,
is insightful into how this statistic comes into being.
It starts with a simple older idea: the best predictor of a team's record next year is not their won-loss record this year, but a winning percentage generated based on how many runs they scored and how many runs they allowed. This little formula is known as the Pythagorean winning percentage for its similarity to the famous geometry formula. When it comes to predicting future success a lot of other factors can be important, but that's not relevant just yet. Generally, most teams will finish the season with a Pythagorean winning percentage that is within 3 games of their actual record. Every year though, teams do much better or much worse than the underlying data suggests. Statistical minded analysts generally attribute these large swings to luck - often times it shows up as a really good or really bad record in one-run games. What is frequently seen with these teams is a pronounced tendency to get much better or worse the following year and move towards their Pythagorean record. Some quibble with this interpretation of so much luck, but most of it is luck though intelligent managing, a good bench, and bullpen can all have impacts here that are real and not based on luck.
The second idea is that of EQuivalent Average or EQA, and its sister statistic
EQuivalent Runs or EQR. I wrote a number of articles (http://www.brewerfan.net/fullArticle.jsp?articleId=83,
etc) on the subject last year. Briefly, EQA is meant to incorporate all offensive
production (hits, HRs, BB, SB, CS