2023 SABR Analytics Conference Research Awards: Voting Open Now!

With the end of the playoffs last week, we’ve reached the end of designated pinch runner season. Quintin Berry had a good run on the Red Sox postseason roster, appearing three times (once in each of Boston’s three postseason series) as a pinch runner and stealing a base each time. Over the past few years, it seems that teams have been more willing to use that strategy in the playoffs. The nice thing about a playoff roster is that with plenty of off-days, having guys on the bench who can cover for tired regulars isn’t as important. A lot of times it frees up a spot for a designated runner, a guy who is really, really fast, but has very little value in any other part of the game.

A good designated runner can add real value to a team (although let’s not overstate exactly how much), and with the 25th man not needing to do as much on a postseason roster, maybe the best use of that roster spot is to have a DR sitting there waiting for an opportunity to pinch run. It might even be worth it to sign such a player mid-season and stash him somewhere in the event that your team makes a postseason run.

There might be a little problem. It’s widely known that when a hitter is asked to pinch hit, his performance is somewhat less than we might expect of him based on his usual stats. The usual explanation here is that a hitter suffers from coming off the bench “cold.” It’s not that he’s reduced to complete impotence, but he loses a few points off his triple-slash line. Some time ago, I looked at the issue from the perspective of a fielder who enters the game to see whether he loses any of his ability to track balls down. I found little evidence that a pinch-fielding penalty exists. What about a pinch-running penalty?

There are some big methodological problems that come with studying pinch running. For one, we’re most interested in what pinch runners do on the bases, but the reason that they’ve been selected to pinch run is usually because they’re pretty good at running the bases. Then, there’s the fact that pinch runners are usually inserted into games in situations where attempting a stolen base makes sense. Half the time, everyone in the ballpark knows that they’re going to break for second at some point during the next at-bat. The defense will be prepared, and that will probably factor into his success rate. We’re going to have to be careful about how we construct our sample.

Warning! Gory Mathematical Details Ahead!
Here’s the easy part. We’re working with all games from 2003-2012. I also used a speed score that I developed a few years ago. The speed scale is based on the normal distribution, so an average runner has a speed score of zero, and a score of one indicates that the runner is a standard deviation above the mean.

If we’re going to look at pinch runners trying to steal bases, we need to know what situations usually call for a pinch runner to try to steal. I selected for all situations in which a pinch runner was on first and he made an attempt on second base. The results shouldn’t come as a shock:

  • The median speed score was 0.85. This isn’t surprising. There are some pinch runners who are inserted not because they are particularly fast, but because the guy they are replacing is glacially slow. Those guys aren’t stolen base threats. The speed demon guys are, and that’s probably why they’re in there.
  • Again, not surprisingly, almost all the attempts (94.6 percent) came in the seventh inning or later. Surprisingly, 30 percent came with two outs!
  • 39.2 percent of attempts came with the batting team behind, 25.0 with the game tied, and 35.8 with the batting team ahead.
  • 81 percent came with the game within two runs (in either direction), and only 0.9 percent came with the batting team down by three runs or more.

So, we’re looking for the late innings of close games with second base open. We’ll say, for our sample, the seventh inning or later, with the game within two runs. The nice thing is that there are often non-pinch runners on base during those times. Guys who got on base the old-fashioned way, by getting a hit or a walk. For what it's worth, because we're limiting ourselves to the late innings in close games, we're also likely dealing with relievers on the mound, who are generally not as good at holding runners on. There may also be defensive replacements with better arms in the outfield (or behind the plate) in the game. This may affect the results overall, but it's not likely that it will bias them toward either pinch runners or batters-turned-runners. Plus, we have a couple thousand cases to work with, so most of the variance should wash out.

Let’s look first at whether a manager might as well send out a pinch runner with a giant sign saying “I’m going to try to steal here!” Looking at runners who were on first in these close/late situations, let’s see whether they tried for second as a function of whether they earned their way on first base or were placed there in a pinch.

Speed Score of

Former batters who tried to steal

PR who tried to steal

Greater than 0



Greater than 0.5



Greater than 1.0



Managers do try to steal with pinch runners more often than they do their regular hitters. But, we also see evidence that pitching teams are aware of this. One indicator that we might look at is whether the pitching team throws over to first to “check on” the runner. In the past, I’ve found that a throw to first doesn’t necessarily stop runners from going, but it does cut down their chances of success.

Again, this is close/late situations

Speed Score of

Former batters who drew a throw

PR who drew a throw

Greater than 0



Greater than 0.5



Greater than 1.0



Opposing managers appear to be hip to the fact that pinch runners are more likely to go and are checking up on them at a greater rate.

Now, is the pinch runner more or less likely to make it if he tries than a runner would be if he had earned first base? Again, I’m looking at close/late situations (seventh inning or later, game within two runs). I ran a binary logistic regression looking at success on steal attempts of second base, controlling for the speed of the runner and whether a throw had been made to first. I coded for whether the runner was a pinch runner or not, and let it fly. I initially looked only at instances where the runner had a speed score of 1.0 or greater, but then lowered it to 0.5 or greater, and then further to 0.0 or greater.

The “throw to first” variable did make a significant difference (runners who drew a throw were less likely to succeed, and speed score made a difference in two of the three regressions, likely failing in the third one—the one where speed score had to be above 1.0—due to colinearity problems). But being a pinch runner was not a significant predictor of SB success. What coefficients there were actually pointed in a positive direction, but the significance levels were really high (p > .70!) If there’s any effect, it’s that the pinch runner has a bit of an edge over a batter-turned-runner.

I ran similar regressions on another type of “stolen base,” going from first to third on a single. Again, controlling for having drawn a throw, speed score, and this time for the number of outs (entered categorically), I checked to see, again within close-and-late situations, whether a runner’s status as a pinch runner predicted success in reaching third. Again, faster runners were better, and those who had drawn a throw were less likely to take the extra base, but being a pinch runner had a positive (albeit very much non-significant) effect on success rates. I looked also at scoring on a double from first. Same basic results.

There is one way, though, in which pinch runners do appear to fall behind their former-hitter counterparts. Using the same set of parameters (late in a close game), I looked to see what the chances were that the runner would be picked off first.

Speed Score of

Former batters who were picked off

PR who were picked off

Greater than 0



Greater than 0.5



Greater than 1.0



For the curious, the differences between the first and third lines in that chart approach significance (p = .14 and p = .08, respectively). There’s some evidence that pinch runners are more likely to be caught napping. Part of that, as we saw above, is that they received more attention from the pitcher, but even when I limited the sample to those runners who drew a throw, the same pattern appeared.

Hittin’, Runnin’, and Thinkin’
From these data, it looks like there is no pinch running penalty in terms of speed. Pinch runner have success rates comparable to what we would normally expect of them had they been hitters who made their way on base through the virtue of their bat or eye or being willing to take a bruise on their left forearm. It would also be proper to point out that there is no advantage to being a pinch runner, either. Pinch runners don’t seem to benefit from the fact that they’ve (in theory) been resting most of the game to that point. There is, however, some evidence that pinch runners are a little more likely to be caught napping.

I find it most useful to interpret these findings in the context of the findings concerning the pinch hitting penalty and the lack of a pinch fielding penalty. Baserunning and fielding, while not completely mindless, are primarily physical activities. Hitting, on the other hand, is a different animal. Sure, there’s a huge physical component, but it requires much more than brute strength. For one, there are more moving parts to a swing that all have to come together to work. That sort of motor coordination is actually a higher-order function in the brain (there’s a fully differentiated motor coordination cortex up there), and the development of this brain structure is one of the reasons why amazing athletes don’t always make good hitters.

Then there’s the eternal chess match between pitcher and hitter in trying to predict what pitch might be coming next, which—whether players verbalize it as such or not—is a fantastic neurological undertaking. I find it interesting that while the base physical requirements of being a pinch runner (i.e., run, Forrest, run!) don’t suffer upon initial entry into the game, the element that requires higher-order functioning in the form of vigilance and attention (our friend the pre-frontal cortex) seems to suffer a bit, and pinch runners endure the indignity of more pickoffs as a result. We’ve seen that it sometimes takes relievers longer to get their heads in the game when they have limited time to prepare for a save. Seems that Yogi Berra was right when he said that 90 percent of the game is half-mental. Baseball has a massive mental/neurological component, and we know so little about it. One thing we do know is that it seems to take those neurons a little bit to warm up and get firing.

Thanks to Ben Lindbergh for, on more than one occasion, wondering aloud within range of my ears what the answer was to this question. I can take a hint, Ben.

Thank you for reading

This is a free article. If you enjoyed it, consider subscribing to Baseball Prospectus. Subscriptions support ongoing public baseball research and analysis in an increasingly proprietary environment.

Subscribe now
You need to be logged in to comment. Login or Subscribe
Really interesting stuff. Thanks.
Don't you have to consider whether there are more pickoffs w/pinch runners because the defense is trying harder? This may not be fully captured by a simple yes/no as to whether the pitcher threw over to first. The number of throws and the quality of the move also matter. Even though you've limited your analysis to late and close situations, the leverage could still be higher with pinch runners.
If we assume all of effects are real, whether they were big enough to be significant or not, could it be that the pitch runner has a bit of an edge in success because they are being more aggressive? Because they are looking to be more aggressive, they are also more likely to get picked off (so you wouldn't call that "caught napping" but "leaning the wrong way").

You didn't report the (non-significant) increase in success rate on SB or first-to-third. Is it some small fraction of 1%? The increase in pickoff rate isn't much in raw terms. Which also has me wondering whether the stats are just looking at absolute differences or if they are relative in some way (a 0.5% increase in pickoff rate looks bigger because the pickoff rate is so low).

And at the risk of making this comment too long, are your stats reliable near 0? Technically, something based on a normal distribution doesn't work well near 0% or 100%. In my field, people use a "rationalized arcsin transform" to do parametric stats near the ceiling or floor. Just wondering if binary logistic regression has the same issue.
I suppose you could interpret it that way. Or perhaps as a failure to maintain a good balance between aggression and caution.

Binary logit does freak out a little bit near 0 and 100, but with roughly 2-3% of the cases being positive, it should be OK.