Enter a scenario we have all seen before. Man on first via the leadoff walk to begin the inning. The pitcher, relegated to pitch from the stretch, begins to fret over the baserunner and the prospect of a stolen base. First pitch, strike one, runner stays put at first. Next pitch, ball up and away, a subliminal pitchout. Pitch three, the runner takes off, and the pitcher throws abnormally quickly to the plate with a ball up near the letters of the hitter. The catcher, with a perfect location for a seamless transaction, throws a bullet resulting in the would-be base-stealer being out by a mile. The threat of a stolen base is forgotten and the pitcher returns to the wind-up with the bases clean.
A sequence of intertwined and equally important events, such as these, occurs. Often, the credit due for this episode is not easily quantifiable, and much less clear.
In breaking down the phenomenon, the pitcher is the one who walked the batter and is responsible for the baserunner and the subsequent possibility of a stolen base attempt. Meanwhile, the catcher has little impact on the execution of the strike throwing, but is responsible for calling pitches. However, in sabermetrics, we delegate walks as one of the few controllable outcomes of the pitcher and therefore it is reasonable to say that the baserunner is the responsibility of the hurler.
Now, conventional baseball knowledge suggests that once the baserunner reaches base it becomes the task of the catcher to limit the stolen base and throw out any attempts of a would-be base-stealer. After all, our scenario ends up in the box score the next morning as a caught stealing by the catcher, suggesting that all the credit for the caught stealing is for the catcher and the throw.
Going beyond the box score, we know that it was the pitcher who threw the pitch in the location that allowed for a quick transfer -- whether intentional or not, remains another story. It was the pitcher's quick release to the plate that gave the catcher a shot to throw the runner out. The pitcher sets the stage for the catcher to execute the throw to catch the baserunner stealing. Much like the pitcher in our scenario, typically a catcher in his calling and framing pitches gives the hurler the opportunity to execute. Thus, going beyond conventional thinking, we know to question whether or not the catcher should gain all credit for the caught stealing as the pitcher's involvement in the play prior to pitch is possibly influential (positively or negatively) in the outcome.
Our scenario hardly scratches the surface of variables that can ultimately determine a few valuable steps and the difference between a caught stealing and a stolen base -- the most well-known of which is a pitcher who wields a formidable pick-off move.
Now that we have covered the hypotheticals, we cannot know realistically how much of an impact a pitcher and the catcher have on the stealing play until we look further into the study. So let's begin:
Skill of Throwing a Runner Out
Before we get into further business, it is important to know how CS% for a pitcher and a catcher correlates year to year. This will give us a good idea of how much of a skill it is for the pitcher and the catcher to throw out a base-runner. Using a 2002-2012 sample with at least 100 IP or IC, the results are as follows:
|Correlation year to year: CS%
We can interpret, given this data, that a catcher's CS%, year to year, is more sustainable than a pitcher's yet both are hardly significant enough to call throwing a runner out a significant skill. This does not tell us anything significant about who is responsible for catching a base-runner, but rather who is more likely to sustain their CS%, so let's move along.
For our sample we will look at every battery combination from 2002 to 2012 with at least 200 innings together. This means we are looking at 310 different battery combinations over the course of our sample. We will pull their CS, SB, and SBA from Retrosheet, and will use these numbers to assess who has the largest overall impact on the base running game in the battery, the pitcher or the catcher.
B_CS% compared to overall C_CS%
To begin let's look at a chart plotting the battery's CS% as B_CS% and the correlating catcher's CS% as C_CS%:
The red line, indicating the catcher's career CS% from 2002 and on, looks highly unrelated to the battery's performance as a whole. The subsequent correlation for B_CS% and C_CS% is a mere 0.39. Now, later we will isolate a catcher's performance outside the battery, and compare it to the battery's performance but for now, the numbers say that a catcher's overall performance compared to the sample of the battery do not correlate well.
B_CS% compared to overall P_CS%
Now, using the same sample we look at the graph of a pitcher's overall CS% as P_CS% and compare it to the battery's performance. The results are as follows:
As you can see, from first glance, the pitcher's overall CS% seems much more correlated to the battery's CS%. This yields a correlation of 0.79, meaning that the pitcher's CS% is more than twice as significant to the battery than the catcher's. However, now we will look at how the pitcher and catcher performed outside the battery and compare those isolated totals to the battery totals.
B_CS% With C_CS% Isolated
To isolate the catcher's CS% outside the battery, we simply take out the battery totals from the catcher's sample totals which gives us the CS% for when the catcher was catching a different pitcher than the selected battery mate. This should give us a good look into how much a catcher is responsible for the pitcher's correlation to the battery's CS% and vice versa:
As you can see this chart looks remarkably similar to the other C_CS% chart. The correlation; however, decreased from 0.39 to 0.27, meaning that there is still little evidence to suggest that the catcher is strongly responsible for a battery's success in catching runners stealing.
B_CS% With P_CS% Isolated
We will do the exact same thing to P_CS% and isolate it for the pitcher's history outside the battery and then assess it next to the battery's performance:
Despite the P_CS% being isolated, the correlation still looks much stronger than that of the C_CS% -- yet the correlation dropped from 0.79 to 0.47. The difference in isolated P_CS% and before, is much more significant than the drop we observed in the same change in C_CS%. However, the correlation for the pitcher's CS% remains much more significant than that of the catcher's.
Conclusions and Context
Given our findings, it is reasonable to say, at the least, that the pitcher is more responsible than we conventionally think when it comes to catching base-runners stealing -- largely due to the fact that we know that a pitcher's CS%, isolated or overall, correlates highly with CS% for the battery as a whole and twice as much as a catcher's CS%.
Now for those who are skeptical and believe that a strong relationship between P_CS% and C_CS% is skewing the correlations to B_CS%, keep in mind that the correlation between P_CS% and C_CS% is a mere 0.19 for the sample, and -0.03 when it is isolated. This means there is hardly a relationship between the individual CS% numbers between the catcher and pitcher individually, suggesting that performance in the battery is highly unrelated to the individual performance of the two battery mates and is not capable of influencing each other's overall or isolated totals. Furthermore, a combination of the two variables when running a multivariable regression of P_CS% and C_CS% on B_CS% yielded a correlation was 0.26, further supporting our findings that the two variables do not have a significant enough relationship to skew the correlation to the B_CS%.
As it turns out, the pitcher, and not the catcher, is the player with the higher propensity to influence the running game, whether in a positive or negative way. For instance, Johnny Cueto and Ryan Hanigan together as battery mates have throw out 85% of would be base-stealers in upwards of 200 innings together. However on their careers Hanigan has thrown out nearly 40% of his base runners, while Cueto has thrown out roughly 66% of his base-stealers. Hanigan, is widely regarded as one of the better defensive catchers in the game, and being a common battery mate of Cueto does not hurt. Conversely, there are of course additional instances in which the catcher was the battery mate responsible for a tandem's success. For example, Yadier Molina and Jason Marquis together as battery tandem had a B_CS% of 64%. On his career, Molina has a C_CS% of 45% and Marquis has a 28% P_CS% and a 0.21% P_CS% isolated from Molina. Naturally, Molina is heavily responsible for the tandem's large success.
Like in any thing, there will always be extremes. Despite our findings, exceptional talents like Johnny Cueto or Yadier Molna will continue to do what they do best. Generally however, the pitcher's overall and isolated CS% will be highly correlated to the battery's CS%. This sheds light in the fact the science of pitching and catching is a two way street, as they are intertwined in very small subliminal ways. Disregarding one's influence on the other is negligible, much like we do under conventional baseball thinking that the catcher is responsible for CS%. We need to go beyond conventional wisdom and accept that CS% is not a skill directly related to the catcher, but largely a function of the pitcher.
. . .