Most Accurate MLB Experts Past 4 Years
Moderator: Baseball Moderators
RynMan wrote:Just curious, has there been any work done on determining what characteristics of teams correlate with their number of save opportunities? Surely you could just run a multiple regression to see if there is any relationships....?
mamorris wrote:RynMan wrote:Just curious, has there been any work done on determining what characteristics of teams correlate with their number of save opportunities? Surely you could just run a multiple regression to see if there is any relationships....?
At the start of the year I had a look at the previous 3 years and tried to link saves to wins, runs scored, runs allowed and run differential. I don't have the results any more, but from memory the results were pretty disappointing. There was a fairly weak link between wins and save opportunities, but that's about it.
RynMan wrote:mamorris wrote:RynMan wrote:Just curious, has there been any work done on determining what characteristics of teams correlate with their number of save opportunities? Surely you could just run a multiple regression to see if there is any relationships....?
At the start of the year I had a look at the previous 3 years and tried to link saves to wins, runs scored, runs allowed and run differential. I don't have the results any more, but from memory the results were pretty disappointing. There was a fairly weak link between wins and save opportunities, but that's about it.
Thanks for the replies guys and the links have been interesting.
This is more what I was asking, mamorris. I'm fairly sure it wouldn't take a very long time to run the figures thanks mainly to the Lahman Database, but a multiple regression that took into account all team based statistics may come up with something. Sounds like what you did MM was linear regressions to try and draw a relationship.
mamorris wrote:
*If you suffer from stat-ophobia, stop reading now*
Yeah, they were separate linear regressions. Turns out I just needed to go back further to find a relationship. I did the Lahman thingy like you suggested (but for saves, not save opps - Lahman doesn't have save opps to my knowledge). I looked at the years since 1969, when the save stat was officially recognised, and used all of their team-based stats. Eventually I reduced it to this:
Saves (predicted) = 0.485 * Wins - 0.37 * Complete Games + 6.31
The R^2 was about 67%, if you cared. I managed to get it slightly more accurate using runs scored and allowed, but it's probably not worth making it too complex for only a small increase in accuracy. Hope this is what you were looking for.
Users browsing this forum: No registered users and 11 guests