Stealing a base after a disrupted attempt

During one of the games I was watching this past weekend, one of the broadcasters made an interesting comment about baserunning. The comment followed a situation in which a baserunner attempted to steal second base but the batter swung and fouled the pitch off, thus disrupting the steal attempt. The play might have been a botched hit and run. It is difficult to be certain. Regardless, it was at this point that the broadcaster suggested the batting team should be wary of sending the runner again, perhaps owing to the lost element of surprise or that the runner would now be slightly fatigued after having dashed for second base.

Now I should note my being vague is not to protect the identity of said broadcaster (or the players involved). It is because I cannot remember the game I was watching when the comment was made. It was likely a Tigers-Jays, Red Sox-Angels or Cardinals-Orioles game, but I cannot say for certain and really it does not matter. What does matter is if the comment is correct and as such I evaluated this idea of stolen base success rate changing after a previous attempt was disrupted by a foul ball.

Broadcasters say things that don't reflect what actually happens on the field all the time. Harold Reynolds has made a career out of doing so. This proposed baserunning idea may be another example, or perhaps it taps into something real about the game and represents something to be aware of when deciding to send a runner. The hypotheses stated above for not sending a runner after an attempt has been disrupted by a foul ball are interesting.

With that said, I doubt that losing the element of surprise plays a large role. Teams are very aware of their opponents' base stealing threats. It is likely a safer strategy to assume that a runner could be going on any pitch. While the runner going on a pitch that is then fouled off removes some uncertainty for the defense, it does not guarantee that they will attempt a steal later in the plate appearance. From my perspective, fatigue is more likely to play a role. I realize that we are talking about professional athletes, but many plays on stolen base attempts are very close. Having burst off for a stolen base on a previous pitch could be enough to slow a runner down and give the defense a better chance at throwing him out. Moreover, fatigue could be additive after multiple attempts are thwarted by foul balls. Knowing the effect on stolen base success rate could affect teams' running game decisions.

Using the great BaseballSavant.com PITCHf/x search I compiled data that can respond to this question. First, I gathered all pitches on which there was a stolen base attempt for the 2010-2014 seasons. An attempt happened in roughly 2.2% of all plate appearances. The number of attempts did not match perfectly with those found on the Fangraphs league stats leaderboard, which is likely a result of some of the plays being incorrectly coded within the PITCHf/x database. Despite this the stolen base success rates were very close (average difference: 0.21%). Second, I gathered all pitches that were coded as Foul (runner going) within the PITCHf/x database. By comparing these with the stolen base attempt data I was able to isolate the plate appearances in which a stolen base was attempted after a batter fouling off a pitch had disrupted a previous attempt.

Here are the overall stolen base success rates across seasons:

Within my records there were 10,057 plate appearances that involved a Foul (runner going) pitch. Of these, 1,181 involved an undisrupted stolen base attempt later in the plate appearance (11.74%). This is critical to note for a few reasons. First, our sample of 1,181 plate appearances is not very large. Second, it appears that teams change their approach to sending a runner after an attempt is disrupted by a foul ball. They only carry on with a stolen base attempt 12% of the time, so perhaps the broadcaster was making an inference based on previous experience. The reasons for this change in strategy could be those suggested above (i.e., lack of surprise, fatigue), changes in the ball-strike count that make stealing less advantageous, a successful hit-and-run, and/or a survivorship bias.

What I mean by survivorship bias is that there might be a bias for teams to only elect to continue on with a stolen base attempt (after a foul ball ruins a previous attempt) with their better base runners (i.e., those with higher career success rates). Unfortunately, the PITCHf/x database from which these data are drawn does not include the name of the player attempting the stolen base, so I cannot speak directly to the survivorship bias issue. A future investigation could match these instances with play-by-play data to examine the players that are involved. Regardless, this must be considered when interpreting the findings below, as it might affect the comparison that is made.

Ok. What is the success rate of those 1,181 stolen base attempts after an attempt was thwarted by a foul ball? Well, it is quite similar to the overall season rates:

There are really two different ways this finding can be interpreted. The first is to suggest that there is no effect of a previously disrupted stolen base attempt on another stolen base attempt within a plate appearance. Comparing the rates in these situations with the overall rates reveals little difference beyond random variation. An alternative interpretation is that there is actually a negative effect. However, this depends critically on the presence of a survivorship bias. If there is a bias then this set of stolen base attempts includes runners with higher than typical success rates (e.g., >85%), so the observed rates are quite a bit lower than what would be expected.

For interest's sake we can look into the hypothesis that fatigue plays a role. This can be done by looking at stolen base attempts that were disrupted multiple times by foul balls within a plate appearance. Here are success rates for those 1,181 attempts, this time given as a function of the number of foul ball disruptions in the plate appearance:

The sample size issue here significantly limits any confident interpretation. Players do make fewer attempts to steal the more they are disrupted and the rates decline with more foul balls (even from 1 to 2 we see a decline) but the decrease in success is not all that much. For those interested (and still with me), Michael Bourn is the player that made the successful attempt after 5 disrupted attempts. He did so in the bottom of the 2^nd against Justin Verlander and the Tigers on May 22^nd, 2013.

Conclusion

To conclude, this work is difficult to reconcile. While it seems clear that teams limit their stolen base attempts after a disrupted attempt (something for which teams can take note), the reason for this change is less clear. When they persevere in stealing a base the success rate is similar to the typical rate. So they might feel confident in their ability to steal the base. However, as stated a number of times this similarity may actually be representative of a decline for a group of more effective base stealers. This aspect of the analysis needs further examination. Finally, thanks to the mystery broadcaster for sending me down this little path.

. . .

All statistics courtesy of FanGraphs and Baseball-Savant.

Chris Teeter is a Featured Writer at Beyond the Box Score. You can follow him on Twitter at @c_mcgeets.