The Overrated Value of Catcher’s Throwing Arms
If you are familiar with my previous studies on the battery, I have often struggled with preconceived notions regarding the relationship between the pitcher, the catcher and the running game. I have previously concluded that it is the pitcher who has more influence on the caught stealing percentage of the battery than the catcher. In addition, I’ve concluded that it is the pitcher who has more of an impact on the passed ball. Meanwhile, in box scores and in broadcast booths all around the country we continue to reward caught stealing and responsibility for the passed ball to the catcher. Fact is, there are many variables at play and as a result, there is a battery effect that must be considered.
In my continuing study on the relationship between the pitcher and the catcher, this article addresses one specific area of the battery effect and will question the conventional wisdom that the catcher’s arm is the determining factor in the outcome of a would be base stealer.
While there are many variables in play, for today we will solely look at timing of the battery and the past success of the battery, the pitcher, and the catcher in controlling the running game.
There is no question that having a catcher with a great arm has great value in the battery dynamic, but is that enough to constitute a battery combination that will prevent the running game? It comes down to timing on both sides of the rubber, but whose speed, or lack thereof has greater influence on the outcome? Because timing data for each respective attempted steal is not information that is readily available, the following is the methodology that I painstakingly undertook in order to give credence to, or refute “conventional wisdom” that a caught stealing rests with the catcher’s arm.
Methodology
I will quote what I have previously written in my prelude to this study — which you can find here: http://www.fangraphs.com/blogs/a-prelude-to-a-study-caught-stealing-variables-and-assigning-responsibility/
Between the 2011 and 2012 seasons there has been 7757 stolen base attempts — specifically from first to second base. 5641 of these have resulted in a stolen base — 2116 of these have been caught stealing. Our population will act as all the stolen base attempts from first to second base during this time period — primarily because that is how far MLB.TV goes back. I cannot go through all 7757 stolen base attempts, so for the sake of my sanity, instead we will create a representative sample of the population. First, I have pulled all the stolen base attempts via retrosheet, and created two different groups — stolen bases and caught stealing. From those stratified samples, I randomly selected 50 stolen bases and 50 caught stealing and combined them into one large “representative sample”. Now, the number of stolen bases to caught stealing is not proportional to the population, but we want to get a good feel for the distinction between both — so it will serve us well enough.
Using the above methodology, I timed all the individual attempts on a frame-by-frame basis (40 fps), and split each attempt from a pitcher’s first movement to the time where the catcher caught the ball. Then, I timed how long it took between the catch and transfer to the point where the ball was caught at second base. For statistical purposes, the sample was taken from the population –all 7757 stolen base attempts from 2011-2012 – at random. Therefore, the sample is more or less a representative sampling of stolen base attempts from first to second base. I’d hate to bore you with any more technicalities, so let’s move on to the fun stuff.
Timing the Battery
The goal of this study is to answer whose timing is more indicative of the outcome – the pitcher or the catcher?
From our sample, below is a table and a chart of the catcher’s “Pop Time” — the time from catching the ball to the ball reaching the infielder’s glove at second base:
Disclaimer: I have chosen not to include outliers in the below correlation calculations, but they will remain in the charts and tables below. Reason being, given the small sample size we are working with, small outliers have a disproportionate influence. In doing so, we are keeping this analysis statistically sound.
Pop Time (s) | CS% | SBA |
---|---|---|
1.6/1.7 | 43.56% | 25 |
1.8/1.9 | 51.00% | 58 |
2/2.1/2.2 | 53.00% | 17 |
What is important to note from this chart is that a catcher’s pop time has a correlation of 0.01 — which is strange considering that we would expect to see is a negative correlation which would suggest that a catcher with a smaller pop time would be more likely to throw a runner out. Here when binning similar times in the chart we see that generally a larger time equaled a larger CS%, very surprising. In this representative sample as a whole, a catcher’s time did not have a strong relationship with the CS%, going against conventional wisdom and suggesting that there are other variables influencing the outcome of a stolen base attempt. Perhaps it comes from the other side of the rubber?
To explore further, let’s take a gander at pitcher release times and their relationship with the CS%, with the same sample.
Release Time (s) | CS% | SBA |
---|---|---|
0.9/1.2/1.3 | 67.07% | 15 |
1.4/1.5 | 59.00% | 37 |
1.6/1.7 | 41.00% | 39 |
1.8/1.9/2 | 22.00% | 9 |
In contrast to the catcher’s chart, you can see this faux probability distribution is skewed left towards the smaller times. What does that mean? Well, that a smaller time for a pitcher’s release equals a greater probability that a stolen base attempt ends up in failure. A pitcher that releases the ball in a shorter amount of time — with a slide-step perhaps — he is giving his catcher a better shot at throwing out the runner. The correlation backs up our point as it sits at – 0.88 — a strong negative correlation. So, in short, if a pitcher like Wide Miley is on the mound with a low release time (a 1.2 s release time) there is a far greater chance that a stolen base attempt will result in a caught stealing. Conversely, if you have Tim Lincecum (a 2.0 s release time) on the mound, you better have a catcher with an exceptional “pop-time” to have any chance.
However, do pitchers that throw harder have a better chance of throwing out the runner? Or is the advantage simply in the pitcher’s quickness to the plate?
Well, because we have the overall time of the pitcher from release to the plate all we have to do is subtract the time the ball spent in the air after release. In doing so, we find the time it took each to release the ball — we will call this “move time”.
When we calculate the “move time”, its relationship to velocity of the pitch is very weak at 12%. However, a relationship of the “move time” and the CS% sits at -90%, while velocity and CS% correlates at a -4% clip. This means you can have a hard thrower like Chapman, or a soft tosser like Mark Buerhle on the mound, and despite one throwing harder than the other, the success will be dependent on who releases the ball first, not who gets it quicker to the plate once the ball is released.
So what about the overall time of the battery? How does overall time affect the probability of caught stealing in this setting? The following is the table and chart of the overall battery time and the corresponding CS%.
Battery Time (s) | CS% | SBA |
---|---|---|
2.8/3/3.1 | 85.71% | 7 |
3.2/3.3 | 52.78% | 36 |
3.4/3.5 | 48.72% | 39 |
3.6/3.7/3.8 | 33.33% | 18 |
So here we have another distribution that is more or less skewed left towards a smaller time where a lower time equals generally a better CS%. Up until 3.5, we see a gradual descent; all totaled, we see a negative correlation of -0.81. From this chart we know that from 2.8 to 3.2 seconds you will generally see a caught stealing, and between 3.3 to 3.6 seconds you will see variations.
As expected, timing of the overall battery naturally has a lot to do with the probability of a caught stealing. However, our study suggests that one of the two battery mates has more of an effect on the overall outcome than the other. Given that we already know that a pitcher’s release time has a strong negative correlation with the CS%, we would expect to see that a pitcher’s time would have a strong positive correlation with the overall time of the battery — which would mean smaller release time, smaller time of the battery.
Running the numbers substantiates this conclusion; a correlation of 80% between release time and overall time of the battery. Conversely, a catcher’s pop time has little impact on the overall time of the battery, with a 38% correlation.
Another variable to consider is reputation; if a catcher’s arm is inconsequential in the terms of the overall time of the battery, will reputation keep base-runners on their toes? While this is not a timing variable, it is important to consider for our study since we are questioning the value of a catcher’s arm within the battery dynamic.
Past Success/Reputation
To quantify “reputation” of the battery and its components I will implement my metric (bBRS) that estimates the number of runs saved in limiting the running game.
As expected, when a catcher with a high bBRS is behind the plate it is more likely his reputation will limit the number of stolen base attempts — a – 0.21 correlation compared to a 0.30 positive correlation for pitchers. While the relationships are not as strong here, the catcher is the one with the slight edge. This makes sense to me. When you have Yadier Molina behind the plate, a runner is less likely to steal in fear of being caught, so there will be less stolen base attempts against someone of his caliber.
But what happens when we compare past successes of the catcher and pitcher respectively versus the past success of the battery as a pair? Do we expect to see that the past success of the battery will be highly influenced by the pitcher, the catcher, or both?
In order to answer this question, I binned the sample into CS and SB, then compared pitcher bBRS to battery bBRS. For the CS bin the correlation was 88%, and for SB it was 80%. Meanwhile when comparing the catcher bBRS to the battery on both CS and SB the relationship was much different. For CS there was a 57% correlation and for SB there was a measly 13%.
In other words, a pitcher who had past success maintaining the running game had more of an impact on the battery’s ability to throw out runners. This leads us to surmise that the pitcher’s success of holding runners on has more influence on CS% than a catcher’s past success in throwing runners out. In general, these findings back up what I previously found in larger samples where the pitcher had more to do with most battery outcomes than the catcher. For instance, in 2013 a pitcher’s bBRS correlated to the battery’s bBRS at 70%, while a catcher’s bBRS correlated at a 30% clip.
In short, when it comes to the timing variables within the running game and the reputation of the battery mates, our study refutes the conventional wisdom that the catcher’s arm is primarily responsible for caught stealing. While there are other lurking variables at play — like pitch location and handedness of the batter — surface value says that a pitcher’s quickness to the plate is a whole lot more influential than a catcher’s arm in the battery dynamic. Said lurking variables will be topics for future installments and will help us dive deeper into assigning credit to one of the two battery mates. When it comes down to the timing variable, the need for speed is on the pitcher’s side of the rubber.
Max Weinstein is a baseball analyst. He has written for Fangraphs, The Hardball Times, and Beyond the Box Score. Connect with him on Twitter @MaxWeinstein21 or email him here.
Max, really like this. Two ideas to think about. Couldn’t you run a model with SB/CS as the dependent variable and pitcher and catcher pop times as separate independent variables? Should give you a better idea about how to assign “credit.” You could also model this as a two stage process, with reputation determining the probability that a runner attempts a steal and the pop time predicting whether or not that attempt is successful.