Quality of Opponent Splits, For Hitters
Jul 10, 2019 11:00:50 GMT -5
philsbosoxfan, incandenza, and 1 more like this
Post by ericmvan on Jul 10, 2019 11:00:50 GMT -5
How do hitters perform against different qualities of pitchers? Are there guys who hit good pitchers particularly well, versus guys that fatten their stats against bad pitchers?
I've been into that question since I worked for the Sox. For instance, I found that J.D. Drew (before we signed him) was perfectly ordinary against better than average pitchers and earned all of his numbers against below average guys. (Theo's response was, well, he'll be facing them half the time, right?)
One problem back then was the incredible noisiness of the data. xwOBA fixes that.
The other problem is ... how do you define your buckets for pitcher quality? Aren't they going to be arbitrary, and doesn't that introduce noise?
I've finally come up with a kick-ass methodology to solve the bucket problem. You measure how many standard deviations a guy is better or worse than average. (Of course you do this separately for hitters of each handedness.) The trick is to then take a guy who is, e.g., 1.6 SD better than average and put the results against him 60% into the +2 bucket and 40% into the +1 bucket. Weighted buckets! One of the best sabermetric ideas I've ever had.
You end up with 6 buckets. There aren't enough guys +3 SD or better to warrant a bucket. But you end up with buckets for Great (+2 or better), Good (+1), Average (0), Subpar (-1), Bad (-2), and Awful (-3 or worse).
The data runs from after the ASB in 2015, when they changed the ball, to today. So 4 seasons exactly.
Here's what MLB RHP have done against RHB:
.249 Great (5.7% of pitchers)
.285 Good (25.3%)
.313 Average (29.8%)
.343 Subpar (23.2%)
.376 Bad (4.5%)
.439 Awful (1.4%
The numbers for a given hitter will be xwOBA+, where 100 is league average.
The first guy I ran was Xander Bogaerts vs. RHP. I got:
92 Great
108 Good
100 Average
99 Subpar
100 Bad
101 Awful
I started to worry that maybe everyone was like this!
Then I ran Mookie vs. RHP:
127 Great
122 Good
118 Average
114 Subpar
97 Bad
92 Awful
The smoothness of the data here is impressive. It's really hard evidence that the better a pitcher is, the better Mookie hits him relative to everyone else. And in fact, he's been a below-average hitter against the 6% of pitchers who are bad and awful! And that data represents 142 PA.
I haven't even built my league table for R pitchers vs L hitters, or either table for L pitchers. And right now it takes me a few minutes to run one guy's splits vs. one handedness.
But I certainly intend to construct a spreadsheet where I can calculate a guy's splits with a simple pair of copy and pastes from Baseball Savant. I'll do all the Sox and then, before each series, run the opposition guys.
I've been into that question since I worked for the Sox. For instance, I found that J.D. Drew (before we signed him) was perfectly ordinary against better than average pitchers and earned all of his numbers against below average guys. (Theo's response was, well, he'll be facing them half the time, right?)
One problem back then was the incredible noisiness of the data. xwOBA fixes that.
The other problem is ... how do you define your buckets for pitcher quality? Aren't they going to be arbitrary, and doesn't that introduce noise?
I've finally come up with a kick-ass methodology to solve the bucket problem. You measure how many standard deviations a guy is better or worse than average. (Of course you do this separately for hitters of each handedness.) The trick is to then take a guy who is, e.g., 1.6 SD better than average and put the results against him 60% into the +2 bucket and 40% into the +1 bucket. Weighted buckets! One of the best sabermetric ideas I've ever had.
You end up with 6 buckets. There aren't enough guys +3 SD or better to warrant a bucket. But you end up with buckets for Great (+2 or better), Good (+1), Average (0), Subpar (-1), Bad (-2), and Awful (-3 or worse).
The data runs from after the ASB in 2015, when they changed the ball, to today. So 4 seasons exactly.
Here's what MLB RHP have done against RHB:
.249 Great (5.7% of pitchers)
.285 Good (25.3%)
.313 Average (29.8%)
.343 Subpar (23.2%)
.376 Bad (4.5%)
.439 Awful (1.4%
The numbers for a given hitter will be xwOBA+, where 100 is league average.
The first guy I ran was Xander Bogaerts vs. RHP. I got:
92 Great
108 Good
100 Average
99 Subpar
100 Bad
101 Awful
I started to worry that maybe everyone was like this!
Then I ran Mookie vs. RHP:
127 Great
122 Good
118 Average
114 Subpar
97 Bad
92 Awful
The smoothness of the data here is impressive. It's really hard evidence that the better a pitcher is, the better Mookie hits him relative to everyone else. And in fact, he's been a below-average hitter against the 6% of pitchers who are bad and awful! And that data represents 142 PA.
I haven't even built my league table for R pitchers vs L hitters, or either table for L pitchers. And right now it takes me a few minutes to run one guy's splits vs. one handedness.
But I certainly intend to construct a spreadsheet where I can calculate a guy's splits with a simple pair of copy and pastes from Baseball Savant. I'll do all the Sox and then, before each series, run the opposition guys.