A couple of weeks ago we estimated that catchers may generate up to a full extra win by their ability at framing pitches. While the final number was intended to be conservative, Peter Jensen pointed in the comments to the fact that the value had been estimated neglecting at-bats by left-handed hitters.
Thus another three borderline pitches per game should be added to the back-of-the-envelope calculations. That would make for a contribution of nearly 18 runs for Russell Martin, who was shown to be a top-tier catcher at framing.
However, while the number of borderline pitches per game for which a catcher can make the difference has always been estimated conservatively, the assumption that the framing skill is equal on either side of the plate and regardless of the batter’s handedness has not been tested, and that could be a cause for inflated values.
This issue will be explored in this follow-up article.
Inside, outside, lefty, righty
Using the same methodology outlined in the previous article, catchers have been ranked according to their framing ability on inside pitches to right-handed batters, and on both inside and outside pitches to left-handed batters. The following scatter plot matrix should help identify the how the four ratings are related.
Ratings on pitches coming to the same side of the plate (outside to lefties and inside to righties, or outside to righties and inside to lefties) show positive correlations, while rating on pitches on opposite sides show mild negative correlations.
Also, the charts below show there’s a higher variation in ratings on outside pitches (both against right-handed and left-handed batters), while the spread on the inside part is negligible. Given the umpires’ reluctance to concede the inner border to the pitchers, this was probably to be expected.
All the above said, it’s not advisable to extrapolate a global framing rating from the values obtained on the outside part of the plate against right-handed batters, as was done in the previous article.
Recalculating the run contribution due to framing skills considering the separate ratings (by side of the plate and batter handedness), we get that the top catchers (Jose Molina, David Ross and, again, Russell Martin) are worth eight/nine runs per 130 games (data from 2008 to May 2011).
That’s a lower value than the one estimated in the previous article, but it’s still close to one win per season.
Best framing catchers player R/130 Jose Molina 9.14 David Ross 8.39 Russell Martin 8.20 Paul Bako 6.75 Brayan Pena 5.16 Michael Barrett 5.16
Worst framing catchers player R/130 Mike Rabelo -7.73 Jason Kendall -7.65 Eliezer Alfonzo -7.18 Brandon Inge -6.55 J.R. Towles -6.19
Framing on high and low pitches has not been considered, and catchers have been evaluated on an equal number of “frameable” pitches per game, while it’s likely that good framing catchers will receive more borderline pitches.
Since on an outside borderline pitch to a right-handed batter, a catcher can increase the probability of a called strike by as much as 20 percent, one can wonder whether hitters adjust to this and expand their strike zone accordingly, especially on two-strike counts.
Using the same technique to predict a probability of a swing, it turns out that only the identity of the pitcher and the hitter alter the outcome. On the contrary, neither a pitcher-friendly umpire nor a good framing catcher induces the hitter to have a more aggressive approach, even on two-strike counts.
There is still a lot of work to be done on the subject.
First, the sample size can be expanded. In fact, for ease of interpretation of the results, the analysis has been performed only on borderline pitches (which, however, are the ones that make the difference). Though the variability in the likelihood of a called strike diminishes both toward the fat part of the plate and way out of the strike zone, those pitches can add some information to the model.
Then, season to season rating variations can be analyzed. In particular it will be useful to see what happens to catchers who move to a new team in comparison to the their colleagues who stay with their employer, as suggested on The Book’s Blog.
Also, until now the run contribution has been estimated by arbitrarily assigning an average number of borderline pitches to each catcher. Once all the pitches get into the model, the difference between observed and expected called strikes can be calculated for each catcher.
Finally, the issue has been explored only on the horizontal plane. A separate analysis needs to be performed to understand the catcher effect both on high and low pitches. The ultimate analysis would simultaneously account for both the horizontal and the vertical planes; unfortunately that would require a to add a spatial component to the multilevel logistic regression and, to my knowledge, that has not been implemented yet to the statistical package I’m using.
References & Resources
I’d like to thank all those who have commented either here at The Hardball Times or at The Book’s Blog for their input. I know I have not addressed many of the questions yet, but hope I’ll be able to do that before I move to another subject.