In the suggest-a-topic posts after Week 6 of the Baseball Prospectus Idol competition, “bravejason” asked one of the Idol contestants to explore the idea of whether swinging to protect the runner was a valid tactic. Before we can answer that question, we need to ask ourselves does swinging help the base stealer at all. This starts to open up an even more interesting discussion of what effect does the batter (if any) have on stolen base success rates.
Does Swinging Help the Base Runner?
I examined each stolen base attempt (removing those when a pitch wasn’t thrown, i.e., pick-off attempts) using Retrosheet play-by-play data from 2004 to 2008. I also filtered out all instances when the batter had two strikes as they are likely not to swing to protect the runner, and when they do swing they are simply protecting themselves.
Over those five years, these are the stolen base success rates when the batter does not swing versus when the batter swings.
Initially, it seems we have a counter-intuitive result: the batter swinging reduces the success rate. Obviously, the more likely scenario is that when the batter swings and misses, it is the result of a botched hit and run and not simple protection. In the hit and run situation, one either has a poor base stealer and/or a worse lead and slightly delayed start by the base runner.
To eliminate the first possible issue (a poor base stealer), I only considered situations where there was a prodigious (75 or more attempts over the five-year span) base stealer with a better than average success rate (77.0%). The difference reduces significantly, but still the counter-intuitive result holds:
It seems that there will be problems trying to tease out whether a batter swinging is actually helpful to the base stealer using this approach.
Does Swinging Affect Certain Catchers?
Another way to consider this is to find if certain catchers seem more affected by the batter swinging than not. For this, I looked at the 48 catchers who have attempted to throw out at least 100 base runners over the five year span from 2004 to 2008. The table below shows the top catchers who seem to be the most bothered by a batter swinging.
Batter
Catcher Doesn’t Batter
Swing Swings Difference
Chad Moeller 77.6% 93.3% +15.8%
Mike Napoli 73.5% 82.4% + 8.8%
Javier Lopez 85.7% 88.2% + 2.5%
Toby Hall 74.7% 76.5% + 1.8%
Chris Snyder 72.9% 72.2% - 0.7%
Joe Mauer* 61.8% 59.0% - 2.8%
* On a side note, one way to examine a catcher’s reputation is to examine the ratio of the number of stolen base attempts where the batter doesn’t swing to when the batter swings. Overall this ratio is 5.6 with a usual range for an individual catcher of 4.2 to 7.5. This suggests a level of how many times a team tries to advance a batter by a straight steal versus the hit and run. Joe Mauer is in a league by himself with a ratio of 2.8, i.e., teams realize that stealing on Mauer is so difficult that they rely much more on the hit and run for runner advancement.
The effect of the batter swinging was only statistically significant for Moeller, Napoli and Lopez. But here’s the important point. Given that the confidence level was 95%, we would expect from pure randomness 2.4 catchers out of 48 would be found to be statistically significant even if no effect existed. Therefore, only finding three catchers thought to be statistically significant is likely due to noise, and not due to any real signal.
So bravejason, it seems that swinging to protect the batter would be a bad tactic, as the evidence suggests that there is no consistent measurable benefit to the runner, and therefore the swing would simply be a wasted strike.
Handedness of the Batter
As a follow on, I was also interested in investigating the adages that having a left-handed batter benefits the base stealer when trying to swipe second, or having a right-handed batter improves the success rate of steal attempts at third.
The logic is that on an attempted steal of second base having a left-handed batter at the plate helps the base runner since most catchers are right-handed and the potential to bother the catcher’s throwing side will help. Unfortunately, there isn’t any evidence for this, as the success rate with a lefty batter is a mere 0.5% better. Harking back to presidential election polls and the classic “margin of error”, the margin of error on 11,462 stolen base attempts (with a success rate of ~77%) is 1.1%. This suggests there is no real difference between the handedness of the batter on attempted steals of second base.
Similarly, the notion of having a right-handed batter helping in the attempted steal of third base because he blocks the vision and throwing path of the catcher is unfounded as well as the success rate is essentially the same.
This was cool enough to be an article in its own right, because the concluding arguments are significant enough to merit shout-outs in a more main-frame venue. Really very outstanding stuff, Tim.
This is good work and well-written but easily could have been a book chapter rather than an Unfiltered post. It was a really, really long post, and at the end I still didn't feel like it was long enough to meet its stated objective.
Maybe he shoulda hit the 'send' button prior to getting into batter handedness. So think of it as one Unfiltered post, with a followup. It's still concise and demonstrates that swinging has no effect on protecting a runner trying a straight steal. Well done.
I agree with John Kearns. This is very good work and well-written but it's quite a bit too long for a typical Unfiltered post and a tad short to the point that it felt rushed.
I think it's long, but not overlong. I think Unfiltered posts should be something that comes up and has to be immediate or that don't fit into the normal work. So all in all, not out of place, which is what this was supposed to show.
Don't penalize the man because he used tables; tables are quick to read and don't have much prose. It is a long-ish BP:U post but don't confuse vertical length with prose length...
This was on the borderline of being my main article, especially if after reading my initial article the Indians decided they didn't want anything released. It felt "too light" for a main article but a little heavy (especially including the Handedness of Batter") for an Unfiltered post. The problem was addressing length versus a little more of "completeness" of the topic. Most likely I should have erred on the length side by leaving out the "Handedness of the Batter" section. Though I hope that if not on length and structure you all enjoy the content.
Tim,
Could it be that the batter handedness helps the base stealer by enabling them to make more attempts with the same success rate? Since all the percentages in the last table seem to be above break-even then more attempts would mean more runs produced. You could see what percentage of the time the runner attempted to steal 2nd/3rd with a left/right handed batter.
I don't care if an Unfiltered is long if it's this insightful and tackles interesting topics. Besides, it is effectively two small pieces, thus easy to digest.
Um. Did you also filter out times when the batter didn't swing because the pitch was in the dirt or out of reach? Those would tend to boost the success rate on non-swings, relative to swings, because they're also harder for the catcher to field and throw.
I don't have the data (or ability to get the data yet) to align this up with Pitch F/X location, but if I remove all instances of where the unswung pitch was a ball, so all we are comparing are batter doesn't swing (called strike) vs. batter does swing, the 2nd table (just focus on the good base stealers) above becomes:
So in 2004 - 2006 we have about a 2% benefit of swinging and 2007 a weird 12% detriment and in 2008 a 1% detriment which averages out to 1.5% detriment. Maybe we are seeing some slight benefit (i.e. with 2007 being really strange), but it seems likely that if there is it's at most 1% - 2%.
This begs going back in history further to see if 2007 to 2008 are the anomalies.
And a shoutout to my friend Brian K, who simply asked about the differences in success rate between Balls and Strikes when the batter doesn't swing. Here they are (for just the Pool 1 base stealers) -- Note I have also removed pitch-outs in this:
So throwing a strike seems to be a 2% to 3% improvement in catching the baserunner. Now the likelihood is that some of this is not necessarily pitch location but pitch speed as well, if I make the assumption that the situations where a strike was thrown a greater % of the pitches were fastballs versus when a ball was thrown.
I'm starting to see possibilities of lining this up with the Pitch F/X data.
On the side note, anyone who can help me to develop a Pitch F/X database would be great. I'm going to get through the Brooks presentation at the Pitch F/X summit to see if that helps.
Check out Eric's writeup of the summit. There's a link to the SportsVision site in which one of the presenters talked about the average Joe getting the data.
Problem is the presentation is in xml (and not ppt) so I think I need to load up Office 2007 on my computer.
I have no reason to complain about a topic being covered "too well" -- yes, this could have been a slightly short full article; isn't that a very good thing?
This is basically the "Mythbusters" of Baseball Prospectus -- evaluating a long-standing aphorism ("the batter should swing to protect the basestealer") using the scientific method. Doing it in a short, Unfiltered-style post is just the icing on the cake. I agree with the comments suggesting that this is good enough to have been Tim's main entry.
Interesting. I wonder, though, how many of these swinging situations are 3-2 situations, where perhaps you wouldn't normally have the runner going and where perhaps the batter is forced to swing due to a good pitch and the count, rather than swining to protect?
There is also the neglected aspect of hit-and-run versus straight steal, as the base runner tends to get a more conservative jump during a hit-and-run than during a straight steal attempt.
It wasn't neglected so much as that's going to be impossible to tease out (I mention that above). Unless we get to gameFX and we understand the lead that the runner had, no way to really break this out. Not to mention there may be a bit of a continuum on a hit and run as some runners are getting the same lead and jump as a straight steal, while others are being more delayed.
You know, that new fielding system that captures position players would also be a good way to grab the size of the runner's lead... hope they're doing that too.
I wonder further if the handedness of the batter coupled with the throwing arm of the catcher would be a factor? Would a right-handed catcher be more impeded or reticent to jump up and throw if a left-handed hitter is swinging the lumber in his arm path?
There hasn't been a lefty throwing catcher in the majors in 20 years, and only one lefty throwing catcher to play at least 1000 games behind the plate.
I wonder if this is an artifact of the level of play. In high school, little league, ... swinging probably does have a big effect, but by the time a catcher makes it to the major leagues, it doesn't matter anymore. In this way, it could be similar to batter strikeouts or clutch hitting, which most people believe based on their experience playing at a lower level, but might not apply to MLB where only the best of the best (no high school chokers) make it.
I believe another aspect of the success rates for swing vs. no-swing may be the type of pitch that is thrown. Intuitively, it's easier to throw out a baserunner on a letters-high fastball than a breaking ball in the dirt. While a batter may still get fooled and swing on the breaking ball in the dirt, I would hypothesize that some of the difference in success rates for throwing out runners on swing vs. no swing may be attributed to the fact that pitches not resulting in a swing are more likely to be pitches which are more difficult for the catcher to get off a good/quick throw.
Yes, gaucho777. If you look mid-way up the comments, I broke out the percentages after I removed pitches called a ball. Therefore, we are looking at when the batter swings (on any pitch) versus those where the batter didn't swing but called a strike.
As I mention, it would be great to merge this dataset with the Pitch F/X data so it could be an apples to apples of at least pitch location.
Tim,
It would also be interesting to use the Pitch F/X data to look at the flip side of the coin by breaking down what the baserunner does. The runner needs to get a lead, pick a pitch to go on, read the pitcher's motion, and run fast. With the Pitch F/X data, you could isolate at least part of that and see if there is a year to year correlation with runners' ability to choose to run on breaking balls or low pitches, for example.
34 comments have been left for this post.
This was cool enough to be an article in its own right, because the concluding arguments are significant enough to merit shout-outs in a more main-frame venue. Really very outstanding stuff, Tim.