Posted by jtrinsey on 3/21/2013 9:51:00 AM (view original):
Yeah I used the league average CS rates.
Basically I did: [ .4*(CS - E) - .2*(SBA - CS + PB + E) ] - [ .4*(.25 * SBA - 5) - .2*( (.75 * SBA ) + 8 ) ]
The first bracket is the player and the second bracket is the "league-average. So the first term .4*(CS - E) is the value of the outs he created and the second term .2*(SBA-CS+PB) is the value (negative for a fielder) of the bases he allowed. I guess 0.436 would be more accurate. The first bracket is the value of the outs a league-average player would have created, assuming 25% CS rate, which seems about average for the No Quitters world. And likewise for the second term, using 5 as a league-average number of errors and 8 as a league average number of passed balls. I guess I improve that by doing errors and PB for catchers as a rate function, but honestly I was just kind of lazy by the time I got around to catcher defense so I took a shortcut there. I then took all that and then divided it by Innings/1200 to normalize it somewhat (ie, a guy that allows 4 PB in 600 innings would be average rather than above average).
This assumes that all errors create an out, which might not be true. Some of them might be overthrows that just add another base, but I think it's close enough. Thoughts?
Let me play around with some of this tonight.
I like the idea of passed balls being a rate function in the sense they are rare, and I don't think have any real correlation to any of the defensive ratings, but I haven't done any test to prove this theory. So I like the idea of saying, Player X played in 1100 innings and had 7 PB, while league average catchers had 6 PB in the same time frame.
First part: Would we want to subtract errors from CS? We're already giving catchers who commit errors a .2 run deduction for each, which assumes the error would've been an out, or allowed the base-runner an extra base. So I think we it would be best if we said [CSvalue * CS - .2(SBA - CS + PB + E)] for the first part.
Second part: No real qualms about this. Just change the 5 for errors and 8 for passed balls into an equation that would give us league average numbers over a given player's playing time. The 25% CS rate and 75% SB Allowed rate could be modified to incorporate the league's average that season, so this statistic could be used to compare players across multiple years.
So I think this is a good start.
Now, do we even care about PO and A? I say no simply because most assists are from throwing base-runners out during SB attempts, and we obviously don't care if a catcher caught a lot of pitchers who have high strikeout rates.