Breaking Down Stolen Base Break-Even Points

If you read this site, you’re probably familiar with the rule of thumb that says stolen bases are only beneficial if they are swiped with roughly a 75% success rate. This number stems from looking at a run expectancy chart and comparing the difference in expected runs after a successful stolen base and the difference in expected runs after a failed attempt. At some success rate, the entire benefit a team received from stolen bases is cancelled out by the combined detriment of all of the failed attempts – this is the break-even point. Of course the break-even point is not the same for every situation. Previous studies have shown this required success rate drops as the game moves into the later innings and increases the further a team is down by – but what do these break-even points look like based on the number of outs in the inning?

Stealing second base:

The chart below plots the expected run difference versus stolen base success rate based on the 2009 to 2011 run expectancy charts. This plot was included primarily as a sanity check, but you can see that regardless of the number of outs in the inning, the break-even point, which is the crossing of the x-axis, falls between the 70% to 75% success rates we’d expect to see.

Stealing third base:

This event was the primary reason for this post. Commonly accepted baseball principle is to not make the first or third out at third base. The rationale for this is qualitative in nature. With zero outs in the inning, there will be plenty of opportunities to drive in a runner from second, so there is no need to risk losing the baserunner. Of course with two outs in the inning, making the third out at third removes any chance of that baserunner actually scoring. Therefore, situations with one out are left as the circumstance where a baserunner might as well take a risk. The plot below clearly agrees with the accepted convention as we see success rates of 78%, 69% and 88% for zero outs, one out, and two outs, respectively, for break-even points.

Stealing home:

What some dub as the most exciting play in baseball sees an even bigger split based on the number of outs in the inning. With zero outs in the inning, an 87% success rate is required. With one out, a 70% success rate is required. Finally, with two outs only a 34% success rate is required. The qualitative explanation for why a runner shouldn’t make the first or third out at third base can be applied here, with the difference that if the runner is successful, a run does actually score. This causes the required success rate for stealing home with two outs to drop significantly. Does anyone else think Jacoby Ellsbury, Brett Gardner or Michael Bourn can steal home at a greater than 34% clip?

Double steal of second and third:

Assuming the catcher is attempting to throw the lead runner out, we again see splits in the break-even points for double steals – 64%, 60% and 76% success rates for zero outs, one out, and two outs, respectively. The required success rates for a double steal with less than two outs is lower than “normal” because outside of stealing home, the change in run expectancy we see from a first-and-second situation to a second-and-third situation is the largest increase of any movement on the run expectancy chart when considering the possible outcomes from conventional stolen bases. In other words, a double steal is more valuable than a single steal, which makes plenty of sense. In addition, the increased benefit of a double steal beyond a single steal is proportionally stronger than the more severe penalty of failed attempt.

In no way am I suggesting that the accepted rule of thumb is wrong.  In fact it represents most situations quite well. Just remember that there is quantitative evidence that backs up the adage of not making the first or third out at third base – and that if Coco Crisp gets thrown out trying to steal home with two outs in an inning, it was probably a worthwhile attempt.


Print This Post

36 Responses to “Breaking Down Stolen Base Break-Even Points”

You can follow any responses to this entry through the RSS 2.0 feed.
  1. jake says:

    Articles like this are why I love Fangraphs.

    +55 Vote -1 Vote +1

  2. RMR says:

    Particularly in the case of stealing home, it seems very important to consider the other badmse states as well. The RE of 2O –3 is much less than 2O 123.

    Vote -1 Vote +1

  3. Eminor3rd says:

    Nice post

    Vote -1 Vote +1

  4. KD Hucke says:

    Interesting piece! I assume that you used the generally-accepted run expectancy values, but how do these break-even points change when only a single run is needed to tie a game or win it?

    +5 Vote -1 Vote +1

    • phoenix2042 says:

      make this the next article! how these numbers change when down by a run, up by a run, or even.

      Vote -1 Vote +1

  5. Russ says:
    FanGraphs Supporting Member

    This is an awesome article

    Vote -1 Vote +1

  6. Mingy says:

    this is incredible. ditto to what Jake said up there

    Vote -1 Vote +1

  7. Anthony says:

    Can we start listing guys we know who primarily throw from a windup with a runner on third?

    Only guy I can think of off the top of my head is Verlander after seeing him do it in the playoffs.

    Vote -1 Vote +1

  8. Joe says:

    Great article, what about a triple steal? 1st, 2nd, and third.

    Maybe I should just add those two lines together and divide by 2.

    Vote -1 Vote +1

  9. Anon21 says:

    I really appreciated some vintage Joe Morgan Wrongness popping up in that video of Ellsbury stealing home. Plus immediate internal contradiction. You should never steal home with a left-handed hitter at the plate…but maybe if the pitcher is left-handed…anyway, great job by Ellsbury.

    Truly, Joe was the Mario Mendoza of color commentators.

    Vote -1 Vote +1

    • Greg H says:

      But Joe Morgan was not the Mario Mendoza of base stealers. The guy knows a thing or two about that topic: 689 career steals, 81% success rate during an era in which runs were harder to come by.

      There is nothing contradictory about Morgan’s comment regarding the Ellsbury steal. With a left-handed hitter at the plate, if the pitcher recognizes the runner is stealing home with sufficient time, he can pitch out and the runner is probably a dead duck. The pitch out will literally permit the catcher to step directly into the baseline where he can cut the runner off and tag him out. The only thing the pitcher can do to combat a steal at home with a right-handed hitter is to plunk the batter. A pitchout to the first base side of home would make it even more difficult for the catcher to tag out the runner in time.

      And a left-handed pitcher, especially one going out of a windup like Pettitte did, is also at a disadvantage. He can’t see the runner on third like a righthanded pitcher can. So however inartfully Morgan put it, his statements are true. It probably isn’t worth the risk to steal home with a lefty hitter and a right-handed pitcher because the defense has more options to counter the surprise element of the steal at home.

      Vote -1 Vote +1

      • williams .482 says:
        FanGraphs Supporting Member

        On the flip side, a left handed batter knows that the steal is happening. a right hander does not.

        Vote -1 Vote +1

  10. Slacker George says:

    Great analysis.

    On stealing home, I’m guessing that there is too much SSS to determine what success rate any individual player has. Too much noise in the data. You’d probably have to lump everyone together, calculate by pitcher-handedness and ask yourself: “Is Coco Crisp better or worse than the average stealer of home?”

    Vote -1 Vote +1

    • Josh Goldman says:

      Agreed, the amount of numbers for stealing home just isn’t there. I think the best way to do it would be to model the event. By that I mean, (1) find the amount of time that elapses from when a pitcher begins his windup to when the ball reaches the catcher’s glove, (2) determine the reaction time baserunners have off of the first movement of a pitcher, and (3) find the amount of time it takes a baserunner to run the roughly 85 feet they need to run. Maybe a couple other things I’m missing. It wouldn’t be perfect, but it could get you pretty close to figuring out what you’d want to know depending on how exactly you came up with those numbers.

      Vote -1 Vote +1

      • Jay29 says:

        Great ideas, but no potential stealer of home will have a 5 foot lead. They’ll probably only be stealing because they could get a great lead.

        Vote -1 Vote +1

  11. RobBob says:

    On the double-steal numbers: Clearly a “success” is when both runners advance, but what does a “failure” entail, i.e. which runner is getting thrown out in your analysis? A double-steal attempt with 0 out in which the trail runner is thrown out is not a great result, but it’s not completely negative, whereas if the lead runner is nabbed, then it’s just a disaster.

    Vote -1 Vote +1

    • BenS says:

      It said in the article that the catcher was attempting to throw out the lead runner, so a failure would be the player stealing 3rd thrown out.

      Vote -1 Vote +1

  12. kanosha says:

    Great analysis, though there are a number of variables left out of the equation, the biggest being the hitter at the plate:

    Stealing second with a hitter with a higher ISO would not be as productive as stealing second with a hitter with a lower ISO because the hitter with the higher ISO is more likely to hit a HR

    Likelihood they pitch around the hitter with 1st base open (i.e. should more often steal second with Molina/Jay/Furcal up than with Albert up)

    Great job running the numbers and starting the discussion though.

    Vote -1 Vote +1

  13. Anon says:

    The rationale I’ve always heard for “Don’t make the 1st or 3rd out at 3B” is that with 0 outs a runner on 2nd can score without a the benefit of a hit while a runner cannot score from 3rd on an out when there are already 2 outs. So with 0 outs stay put because you can score with 2 ground balls (or deep flyouts) and with 2 outs stay put because it’s going to take a base hit anyway, plus chances are the runner will score from 2nd on any hit since he can run on contact without worrying about the ball being caught.

    Baseball Prospectus did a larger piece a few years ago on baserunning in general (not just SB) and found teams are MUCH too conservative on the base paths, especially when it comes to sending the runner home. They also found that not making the 1st or 3rd out at 3B held true in every scenario.

    Vote -1 Vote +1

  14. Loren says:

    Another thing I’ve wondered about is the effect on the break even point of the count on a batter with two outs. With an 0-2 count if the runner gets caught, at least the batter will start out with a clean slate in the next inning. On the flip side, the break even point would have to be really high on a 3-0 count. The count should factor into the break even point even if there aren’t two outs, but probably less so.

    Vote -1 Vote +1

  15. Bip says:

    It’s rare that an article has data that is clear enough that a conclusion can be drawn that actually could affect the way baseball strategy is implemented. This article has that, which is awesome.

    Vote -1 Vote +1

    • Bip says:

      I say that because a lot of articles, especially Pf/x ones, are analyzing results that are influenced by so many factors and using data that is so hard to interpret that it seems hard to draw any real conclusions.

      Vote -1 Vote +1

  16. adohaj says:

    So with the pitcher batting and 2 outs you should attempt a steal of home almost every time someone who has speed is on third?

    Vote -1 Vote +1

    • Aaron (UK) says:

      It’s not clear with the run expectancy calculations are with respect to the innings or the game as a whole, which would make a big difference here.

      And, as others have said, it would be nice to do something factoring in the ISO (and other batting stats) of the batter. It might even persuade teams to stop putting base-stealers ahead of sluggers in the lineup :-)

      Vote -1 Vote +1

    • Jay29 says:

      The problem is nobody knows the expected success rate of a steal of home for the best base-stealers. If no one can sniff 34% then no one should ever steal home under normal circumstances (barring a weird mistake by the pitcher or defense).

      Vote -1 Vote +1

  17. john says:

    Mind = blown

    Vote -1 Vote +1

  18. Anon says:

    Actually another interesting aspect to these charts is the slope of the lines. Other than stealing home, stealing with 2 outs has comparatively little impact on run-scoring while stealing with 0 outs has quite a bit more impact. IOW, neither safe nor out has much impact with 2 outs while it has a lot of impact with 0 outs.

    It would be interesting to see when teams try steals the most often. I’m betting they steal more often with 2 outs when it has the least impact and least often with 0 outs when it has the most. If so, that would tell me that teams focus on the negative much more than the positive since the the rationale behind not stealing with 0 outs is not to run yourself out of an inning and for stealing with 2 outs is that it doesn’t cost much if you do get thrown out.

    Vote -1 Vote +1

  19. Anon says:

    BTW, I’m trying to figure out how a 100% success rate on stealing home doesn’t equal 1 on the vertical scale. Wouldn’t that mean a run every time?

    Vote -1 Vote +1

    • Josh Goldman says:

      It doesn’t equal one because the vertical axis is the difference in expected runs. For example, with a guy on third and two outs, the expected runs scored is about 0.35. With the bases empty and two outs, the expected runs scored is about 0.1. So if he’s successful 100% of the time, he gets credited with the difference in expected runs given the situation before and after the stolen base, plus one – because a run scored (ie 0.1-0.35+1=0.75 runs).

      Vote -1 Vote +1

  20. Bill Deane says:

    I’ve long thought that the optimal time to steal home would be while a right-handed batter is being intentionally walked (ideally, with two out and a 3-0 count). The pitcher is lobbing the ball in, the catcher is moving several feet toward first base, and his view of the runner is blocked by the batter.

    Vote -1 Vote +1

    • Curtis says:

      Oohh! Damn, that’s good.

      Vote -1 Vote +1

    • Anon says:

      I thought in that situation the 3B covered the bag expressly to stop that possibility? He doesn’t really need to be off the base because the hitter isn’t hitting.

      I believe if you watch a pitcher in that situation, you will see them checking the runner at 3B. Actually I’m pretty sure it’s true at 2B as well when 3B is open.

      Vote -1 Vote +1

      • joser says:

        Well, maybe mostly. But IBBs are usually part of a high-stress, high leverage situation where the pitcher already has been struggling, so an attentive runner at 3rd will probably frequently see a distracted, inattentive pitcher.

        Vote -1 Vote +1

  21. Jake says:

    I am currently trying to work on a project for a statistics class and my group and I were interested in studying stolen bases also. Is there a way we could get access to the statistics used in this study?

    Vote -1 Vote +1

  22. John Cole says:

    Great article. We rarely steal 3rd or home with a left hand hitter up. Are the odds of stealing second better with a left handed batter at the plate?

    The best situations for stealing home are on mistakes where the pitcher uses the windup,the third baseman is off the bag, you have a fast base runner with a right handed hitter. Given this scenario, we were 6 for 6 in stealing home this summer. Right or left handed pitcher didn’t matter.

    Coach Cole
    East Marietta Mets

    Vote -1 Vote +1