question:Before the season, Andy MacPhail said that the team - and more specifically, Dave Trembley - would be judged more on wins and losses this year. Well the W's were few and far between, the L's piled up, and Dave was shown the door. Now a lot of fans are calling for Andy's exit, citing the team's record, which is worst in the majors. How strong of an argument is that? MGL from The Book Blog asked the
"Let’s say that you knew the exact skill level of every manager and GM, with respect to their contribution to their team’s winning... And let’s say that you were able to rank each manager and GM or assign some other quantitative value to that skill level. Now, imagine that we had lots of data along these lines and we were able to run a regression of skill on team WP for one year. What do you think the correlation coefficient would be for one year of a team’s w/l record for the managers and also for the GM’s? How about for 5 years and for 10 years?"That is, how much can you tell about the quality of a manager or general manager from a team's record? Phil Birnbaum, from Sabermetric Research, runs a few numbers:
"Suppose the best GMs can add 3 extra wins, and the worst cost 3 extra wins. That means the SD of GM talent is roughly 1.5 wins. The SD of actual wins in the majors is a bit less than 12. So the one-season correlation would be 1.5/12, or .125. So if a team goes 2 SD above average, going 105-57, the GM was likely responsible for .125 of 2 SDs, which is 1/4 SD, which is about 2/5 of a win."Tom Tango speaks to MGL's point:
"If you have the perfect lab experiment, that if you rely solely on team wins, you will still have a low correlation. Therefore, if you use real-world situations, the team wins will be such a laughably poor variable to use."Which hits the nail right on the head:
"Exactly. I have an extensive background in blackjack/card counting. It would never even occur to a serious blackjack player to evaluate someone’s blackjack “skills” by their w/l record in ANY amount of time. I have always found it laughable when people evaluate managers and GM’s by their team’s w/l record, or even their team’s w/l record as compared to the mainstream expectations (which is even more laughable as the mainstream expectations are often completely wrong). If you want to begin a discussion about the “skill” level of a manager or GM, and you start by compiling a list of pieces of evidence supporting that skill, please have team w/l records at the bottom of a very long list…"So if you want Andy MacPhail gone, it would be good to hear a reason (well, actually several) that don't stop at how many wins and losses the team had. A case can still be made for it - and I'm getting closer to being willing to do so - using the moves he's made (or not made).