Nov
22

Looking At The Yankees’ Sac Bunts

By

(AP Photo/LM Otero)

Baseball is a game without an official clock. In its stead, the 27 outs each team receives serve as the timekeeper, pushing each game to an inevitable conclusion. Avoiding those outs has become the name of the game over the last ten years, and one of the strategic moves that has come under fire due to this philosophy is the bunt. The sacrifice bunt draws a team one out closer to the end of the game without greatly increasing the chances of a run scoring. A look at run expectancy tables, which tell us how many runs are expected given a particular situation, confirms that bunting usually decreases the number of runs expected to score. While there are a few situations where a bunt is actually the statistically prudent move, on balance it is seen as the misused weapon of weaker, backwards-thinking managers, and is the hobgoblin of sabermetricians everywhere.

All that said, there is at least one study that suggests that managers tend to outperform run expectancy tables when it comes to bunting. This means that on average, managers have a reasonably good sense of the moment and of context, and they bunt in situations where it will produce more runs than one might expect given the post-bunt base/out status. While the numbers still suggest that these bunts decrease run expectancy, it is illuminating and encouraging to see that managers are utilizing the bunt reasonably efficiently.

All of this brings us to the manager of the local nine. One of the most common complaints about Joe Girardi‘s managing is that he bunts too frequently, playing for one run with an offense that can put up a crooked number in a hurry. I thought it would be instructive to look at every Yankee sacrifice bunt in 2011 to see how many runs Girardi actually cost his club with his small ball sensibilities. I broke the bunts down by player and then calculated three numbers:

  1. Expected runs before the bunt. This number tells us how many runs were expected to score given the base/out situation prior to Girardi working his managerial magic.
  2. Expected runs after the bunt. This tells us how many theoretical runs the bunt “cost” the club.
  3. Actual runs. This should tell us how Girardi’s move actually worked out.

Now, a few caveats.

  • Run expectancy is not perfect. It does not account for the score or the quality of offense or opponent, nor does it account for the skills of the hitter at the plate. However, it is a reasonable estimate of how the game has been impacted by a move, and I’ve broken things down by hitter so you can mentally adjust your evaluation based on the quality of the batter.
  • This study does not include the attempted bunts that failed and caused batters to fall behind in the count. However, it also does not include bunt singles or bunts in which the batter reached on a fielder’s choice or error, which help to greatly increase run expectancy (I also excluded Nick Swisher‘s bunt against Boston where he lost track of the number of outs and bunted on his own). The analysis is limited to successful sacrifice bunts. I’ve also removed all bunts by pitchers, as I think most of us can agree that bunting with an American League pitcher is almost always the correct move.
  • We cannot calculate what would have happened if Girardi had chosen not to bunt. To provide an example of why this is an issue, imagine an inning where Brett Gardner bunts a runner over and then Curtis Granderson homers. While we can figure out the run expectancy before and after the bunt and can observe actual runs scored, we can’t know what would have happened if Gardner had not bunted. So if one run was expected and two actual runs were scored, there is still the possibility that without the bunt, three runs would have scored (because Gardner could have reached prior to the home run). If we assume that everything would have been different and Granderson may not have homered had Gardner reached, the expected runs v. actual runs analysis is relevant. As such, this study is making the assumption that the bunt changes the entire inning, such that whatever happened afterward is connected to (but not necessarily caused by) the base/out state created by the bunt. Discarding that assumption does not make the conclusions irrelevant, but it does sap them of some of their power.

Keeping all that in mind, let’s take a look at the sac bunts Girardi called for in 2011.

Brett Gardner

# of sac bunts: 8

Expected runs, before the bunts: 7.0173

Expected runs, after the bunts: 5.4602

Actual runs: 11

Loss of run expectancy: 1.5571

Actual impact: Gain of 3.9827 runs over expected runs

(To be fair to Girardi and his predilection for bunting with Gardner, it is important to note that all of Gardner’s bunts but one came in the late innings of a tight game, when playing for one run is acceptable. The lone exception came against Justin Verlander, which represents another understandable, if not entirely defensible, use of the bunt.)

Eduardo Nunez

# of sac bunts: 6

Expected runs, before the bunts: 5.5296

Expected runs, after the bunts: 4.4064

Actual runs: 3

Loss of run expectancy: 1.1232

Actual impact: Loss of 2.5296 runs under expected runs

Derek Jeter

# of sac bunts: 4

Expected runs, before the bunts: 4.1974

Expected runs, after the bunts: 3.4954

Actual runs: 2

Loss of run expectancy: 0.702

Actual impact: Loss of 2.1974 runs under expected runs

Curtis Granderson

# of sac bunts: 3

Expected runs, before the bunts: 3.7157

Expected runs, after the bunts: 3.2358

Actual runs: 6

Loss of run expectancy: .4799

Actual impact: Gain of 2.2843 runs over expected runs

Ramiro Pena

# of sac bunts: 2

Expected runs, before the bunts: 1.701

Expected runs, after the bunts: 1.3028

Actual runs: 1

Loss of run expectancy: 0.3982

Actual impact: Loss of 0.701 runs under expected runs

One each for Russell Martin, Frankie Cervelli, Chris Dickerson, and Brandon Laird

# of sac bunts: 4

Expected runs, before the bunts: 3.402

Expected runs, after the bunts: 2.6056

Actual runs: 1

Loss of run expectancy: 0.7964

Actual impact: Loss of 2.402 runs under expected runs

Conclusion

# of sac bunts: 27

Expected runs, before the bunts: 25.563

Expected runs, after the bunts: 20.5062

Actual runs: 24

Loss of run expectancy: 5.0568

Actual impact: Loss of 1.563 runs under expected runs

Regarding that actual impact number, I am uncomfortable concluding that the bunts were always directly responsible for what happened after them. For example, I do not think Granderson’s lone “successful” bunt actually caused all 6 runs that subsequently scored in the inning. That said, I think it is fair to conclude that Girardi’s proclivity for bunting did not hurt the Yankees much in 2011. In terms of run expectancy, all of the bunts over the course of the season only cost the Yankees five runs, and that ignores the fact that many of them came in situations where playing for one run at the expense of a big inning is actually the right thing to do. Furthermore, the team outperformed the “runs expected after the bunts,” suggesting that Girardi may have utilized the strategy in optimal situations. Taking into account the fact that the actual runs scored was about the same as the number of runs expected, it seems clear that Joe Girardi’s bunting problem was not much of an detriment to the Yankees in 2011.

Update (12:28 p.m.): I am new to play index, but I just figured out how to get bunt singles and bunt outs listed properly(still no foul bunts, however). Here are the results for the 18 sac bunt attempts that ended without a sac bunt:

10 runners reached base
8 made force outs or popouts

On the outs:

RE before the bunts: 7.5994
RE after the bunts: 4.431
Actual: 7
Loss of RE: 3.1684
Actual impact: 0.5994

On the hits:

RE before the bunts: 9.9545
RE after the bunts: 16.1122
Actual: 17
Loss of RE: Gain of 6.1577
Actual impact: Gain of 7.0455

New total:

RE before the bunts: 43.1169
RE after the bunts: 41.0494
Actual: 48
Loss of RE: Loss of 2.0675 runs
Actual impact: Gain of 4.8831

Categories : Analysis

67 Comments»

  1. “However, it also does not include bunt singles or bunts in which the batter reached on a fielder’s choice or error, which help to greatly increase run expectancy…”

    Shouldn’t those results be included, if they occurred during a sac bunt situation?

  2. Plank says:

    I feel like Jeter bunted a lot more than 4 times. Maybe I just remember the decade and a half of infuriating bunts.

    • Moshe Mandel says:

      Hey may have attempted more bunts and failed.

      • Plank says:

        Doesn’t a failed bunt attempt still count as a bunt attempt? Especially when the purpose is to show the impact on the runs scored.

        • Moshe Mandel says:

          As I said in the post, I had to exclude those due to some data issues. On the flip side, I also excluded bunt singles. This is solely to examine whether the Yankees were hurt when the team attempted to give away an out and succeeded at doing so.

          • Plank says:

            Looking at only successful attempts is very limited in usefulness though. It is like looking at the impact of singles. The chance of getting a successful bunt in a bunt attempt is much higher than getting a single in an at bat, but they are both equally impossible to know the outcome of them before the attempt.

            I think the end result of the study mischaracterizes the positive impact of bunts on runs scored since you are only looking at successful attempts.

            • Moshe Mandel says:

              I’m also not looking at bunt hits, though. I’m pretty certain that the impact of bunt hits would outstrip the impact of foul bunts and even bunt pop-ups, simply because hits are rarer than fouls and outs.

              • Plank says:

                I’m just saying looking at one specific bunting outcome post hoc doesn’t tell about bunting and how it effects the runs scored, which it seems is the conclusion this article makes.

                When someone goes up to bunt, no one knows what the outcome will be, so looking at only one specific outcome and drawing a conclusion to the thousandths decimal about sac bunting as a whole seems flawed.

                • Moshe Mandel says:

                  None of the other outcomes are sac bunts, though, which is what I’m trying to measure here. Yes, there are things that are not measured in this post that relate directly to sac bunting. But the sac bunts themselves only impacted the Yankees by X runs. That’s all I’m saying.

              • Mickey Mantle's Outstanding Experience says:

                The Yankees only had 8 bunt hits with men on base this year. Assuming all of them were attempted sac bunts, they’d only need 12-14 fouls/popups to cancel out the positives of the hits.

    • Thomas says:

      I think with Jeter and quite a few other players, they attempted to bunt, but failed to get the bunt down. Then the batter is down in the count and usually this leads to outs made by the offense. Unfortunately, these weren’t taken into account in the analysis, because they aren’t listed as sacrifices in the box score.

      I personally feel like these are the most common Yankee bunt attempts and ultimately bunt failures, but they just were able to be accounted for.

  3. Plank says:

    Did any players bunt foul for strike three this year?

  4. MattG says:

    I love WAR. It takes a player’s total contribution, and does a fine job of approximating how that impacts wins. It’s very handy.

    But in certain instances, it is stupid to look at WAR in the accumulative. You can’t actually win 3.4 games, anymore so that you can score 1.6 runs.

    The idea of taking all these parts and summing them up as though they mean something has gotten way overdone. The only useful way to look at a decision to bunt is on a single, case-by-case basis. Then, you would see you would probably never use “run expectancy” as an input; instead, you would want the “expectancy of one run,” because in most situations, you believe one run is what you want.

    • Moshe Mandel says:

      Yes, going bunt-by-bunt would be more useful, and the post I linked to has the expectancy of one run chart that you would need for those situations you referred to. That said, I think this is a useful way for looking at the overall bunting tendencies of Girardi and how much they might have impacted the Yankees. If I did it bunt by bunt, he would likely come off looking even better.

      • MattG says:

        Girardi might look better. I am willing to entertain that thought, but I do know that I never seem to agree with it when I am watching the batter square.

        The expectancy of one run is very important, because a big part of the decision to bunt is the need to avoid the double play. This would show up better if we understood the difference in likelihood of scoring just one run from second with one out, as opposed to one run from first with no outs.

        The rest of my comment was an observation I am sharing, in that while we can now easily divide wins and runs into fractions, many have come to decide this is relevant in places where it really isn’t. Like all things, analysis can be distorted, and often is, in the desire to publish new content. I really can’t trust these conclusions, because I don’t think this is the proper way to analyze the decision to bunt. I think you know this too, as you placed many qualifiers into the article itself.

        (BTW, I do not see the one run expectancy at the BP table, but maybe because I don’t have an account)

        • Moshe Mandel says:

          The table is at the other link.

          As for analyzing how to bunt, I think this provides a reasonable estimate. I’ve done bunt-by-bunt analyses in the past, and the results have not been significantly different than what I found here. That’s why I was comfortable using this method as a proxy, with caveats of course.

          • MattG says:

            And there it is! From 1977-1992, the likelihood of scoring one run in bunting situations:

            Runner on first, no-one out: 43.2%
            Runner on second, one out: 41.4%

            So, all things being equal, we’re talking a 1.8% difference. That can easily be overcome if the hitter is below average, or the pitcher is above average at inducing ground balls or strikeouts. If you really only need one run, this isn’t as bad of a strategy as I thought.

            But if you would like more, its still pretty inappropriate.

          • MattG says:

            Couple more:

            I always sort of liked the bunt when my team needs one run and has first and second with no one out. Turns out I was right:

            1st and 2nd, no one out: 63.7%
            2nd and 3rd, one out: 68.9%

            Bunting a runner to third surprisingly helps, too:

            2nd, no one out: 63.4%
            3rd, one out: 67.0%

            (Isn’t it odd that first and second has a better chance of scoring one run than second alone? That means the IBB to setup the double play is a fallacy, albeit by .4%)

            The big problem here is this: the bunt does not always succeed…or sometimes it succeeds better than expected. I doubt these things actually even out though, so who’s already done the definitive study on the results of bunts?

            Finally, there’s this (nothing to do with bunting, but instead IBBs):

            2nd and 3rd, one out: 68.9%
            bases loaded, one out: 67.9%

            This shocks me. I always believed that the threat of walking in a run was much more severe than the likelihood of getting the inning ending double play. It’s actually a good idea to walk the bases loaded when you have to avoid a run? I am floored.

            • Plank says:

              Bunting a runner to third surprisingly helps, too

              I think it makes sense. Any ball hit to the outfield will score the run with a runner on third.

            • Moshe Mandel says:

              Yeah, it actually seems somewhat hard to reconcile this point:

              “Isn’t it odd that first and second has a better chance of scoring one run than second alone? That means the IBB to setup the double play is a fallacy, albeit by .4%”

              with this point:
              “I always believed that the threat of walking in a run was much more severe than the likelihood of getting the inning ending double play. It’s actually a good idea to walk the bases loaded when you have to avoid a run?”

              I mean, isn’t the first one basically telling us that the walk is more likely than the DP, and the latter telling us the opposite?

              • Mickey Mantle's Outstanding Experience says:

                With how close the percentages are there’s a lot more to it than walks vs dps.

                In the 1st example, a single might be more likely to score the runner, since he won’t have to wait to see if a ground ball gets through.

                In the 2nd, teams may bring in the infield in some situations, which makes a hit more likely without a man on 1st.

            • KeithK says:

              I think the last pair of stats points out one of the limitations of this type of analysis. Yes, averaged over all such situations having the runner on first makes it less likely to score a run. But in a specific situation the “actual” probabilities will be spread depending on so many factors (e.g. walk rate of pitcher and batter, ground ball rates).

  5. PaulF says:

    Wouldn’t the “actual impact” numbers be inaccurate because they assume average run expectancy? Shouldn’t the Yankees, having an above average offense be expected to outperform their run expectancy? On the other hand, if the run expectancy numbers come from 1997-2003, they’ll be quite a bit higher than the numbers now, so these two facts might cancel out.

    • Moshe Mandel says:

      It is based on 2011 numbers. As for the first point, certainly, that was actually the first of the 3 caveats I listed. I think this is an estimate more than anything. To me, it shows how little these decisions add up to be worth. 5 runs over the course of a season is practically nothing. I mean, that’s the margin of error for WAR.

  6. Monteroisdinero says:

    Another factor in a sac bunt situation is WHO is on base. If Gardner is trying to bunt Jorge over…..dumb And no one should be sac bunting if Gardner is the lead runner. These are frustrating situations that occurred as well.

    PS: I am at a Mac store at Covent Garden in London. Yankee fans everywhere!

  7. Bronx Byte says:

    In short, the idea is to keep the runners moving and lengthen the innings. Get to the opponent’s bullpen early and often.
    Consistently waiting for the long ball is not the answer.

  8. Professor Longnose says:

    What was the Yankees’ record in games in which they sac bunted?

  9. Total Dominication says:

    You want to lengthen the inning by sacrificing a third of it?

  10. Jose M. Vazquez.. says:

    I like bunting, but only in certain situations such as bottom of the 9th or extra inning with man on first and no outs with your third or fourth batter on deck. I would not bunt with the eighth or ninth man in the lineup coming up. In 2011 there bunts in the early innings that may have killed rallies and led to no runs being scored.

  11. Mike HC says:

    So the Yanks bunts in 2011 weren’t really stupid, they were just a little stupid.

  12. Jose M. Vazquez.. says:

    I omitted saying that the score would be tied or us being one run behind in these circumstances.

  13. Monteroisdinero says:

    Don’t sac bunt with speedy (Gardner/Nunez/Grandy) guys on first. Fake it and steal or hit and run. ARod, Tex, Cano, Montero, Swish can’t bunt anyway. Only a guy like Martin or Jeter should bunt to stay out of the gidp.

  14. Ted Nelson says:

    Good analysis.

  15. KPOcala says:

    If you guys can find it, check on Earl Weaver’s “On Strategy” w/ Terry Pluto. Besides being years ahead of his time (maybe the first “saber-manager) more about understanding baseball can be gleaned there than any place I know. Weaver hated the sac bunt, the hit and run (but not, the “run and hit), and believe that it easier to find four good starting pitchers than 5… If I were a GM, that would be a book I’d memorize, same goes if I were a manager. I believe it’s back in print, here’s some links http://www.google.com/search?s.....l0.2.7l9l0

  16. Jonathan says:

    I enjoyed this post and I’m sure it took a lot of research. I’m no proponent of the sac bunt but I think a lot have people have heard Mike say how much he hates them over and over that they’ve adopted it as something that should never be done.

    One issue that I have with Girardi’s bunting that isn’t in this info, and i’m not sure exactly how you’d get it in, is that a lot of times there’s somebody on 1st who’s more than capable of stealing 2nd but we bunt him over anyway. I know there’s a chance to be caught but when guys like Gardner and Nunez and other fast players are on first I’d like to see more of sending them and then possibly moving them over if the situation calls for it. Or you could then just have 3 chances to drive them in or ask the next batter to hit behind the runner to move him up while still giving the chance to get a hit himself.

    Here’s a slightly off topic point on bunting but it frustrates the hell out of me that Teixeira has seen his AVG drop 40-50 points over the last 2 seasons which makes his OBP and SLG drop too even though his power and on base abilities haven’t eroded. And basically the issue is the league caught on and put the extreme shift on him. I’m not 100% sure but I think the same thing happened to Giambi who went from a 300+ hitter to a 250 guy with the shift. Why on earth doesn’t Tex just bunt down to third every time he’s up lefthanded with the shift? It is almost an automatic hit and if you committed to doing it they would have to either let you get a free hit every time or modify the switch. Either way, the AVG and therefore, the real issue, his OBP climbs. Tex pops up too much and gets homer happy but I’d love to see what his triple slash line would have been on balls hit into the shift that would have been hits with a normal infield setup.

    You see guys do it like David Ortiz or Pena in big situations in the game like they’ve been holding onto it like a special trick play in football that can only work sometimes. But it can work every time. Either you get an auto trip to 1st and raise your AVG/OBP considerably or they have to change the shift which will have the same desired effect. We signed a .285 hitter with a very high 300′s OBP and consistently slugged over .500 with out of this world defense and if something as simple as bunting to an empty side of the field could bring that guy back I see no downside to trying it as I think we’re past teaching him to just hit the other way with success. Has anyone ever done that to beat the shift or are they all so stubborn they’d rather hit rockets into short right for outs than bunt for singles?

    • TedK says:

      Jonathan, I couldn’t agree more. Also, thank you Moshe for taking the time to research the data, including the update on bunt singles and bunt outs (are there additional PAs where the defense made an error?). Even if you nitpick some of the assumptions, the analysis seems to show bunting wasn’t a travesty.

      One point which has come up in other bunt discussions is how the threat of the bunt affects the defense. I think someone fast like Gardner and someone batting against the shift like Tex are two cases where bunting often enough to force the defense to shift is a clear win for the batter. (Gardner wants the corners in so more of his weak hits are singles, Tex wants to force them out of the extreme shift.) Maybe in situations such as if there are two outs and nobody on, you’d rather Tex swing away and hope for the home run, but with bases empty and no outs, why not drop the ball down the third base line and let somebody else drive you in?

      In this case, a big part of the goal is to increase production in the ABs when you are not bunting. I have no idea how to measure it precisely. I have seen analytical articles showing that the shift does in fact work, so it would make sense forcing the D to abandon the extreme shift should be a clear benefit to Tex.

      • Ted Nelson says:

        “I have no idea how to measure it precisely.”

        That’s a bit of a problem. You have a theory that makes some sense, but haven’t proven it.

    • Ted Nelson says:

      Stealing is not risk free, and an out is going to leave no men on.
      I think it’s worth looking into the odds vs. bunting, but I am also willing to bet a lot of money that the Yankees organization has done exactly that. They employ statisticians. I am willing the binder isn’t just nudy pics and a big piece of paper that says “Offense: Righty-Lefty and Lefty-Righty; Defense: Righty-Righty and Lefty-Lefty.”

      Teixiera’s problems are pretty well known, and he’s publicly said he’s working to fix them. If he can’t fix it bunting’s not the worst idea, but given that he’s slow and the P/C can still field the bunt if it’s not laid down very well his odds of success might still be higher by just taking a regular PA.

  17. the Other Steve S. says:

    So, if I’m reading this right, all the howling all year long after every game where there was a bunt was because of the “projected” loss of 1.5 runs? Geez, guys….

    • JAG says:

      While I agree that the degree may be blown out of proportion, it’s still justified to be irritated when Girardi or any other manager makes a decision that is an outright mistake. “Only a little wrong” is still “wrong”.

      • Tim says:

        But the analysis above shows that Girardi’s bunting decisions this year have NOT been an “outright mistake”. The fact is, the team outscored their “expectancy” in these situations after the bunt, and they won those games at a greater than 72% clip as well, so they outperformed their “win expectancy”, too.

        It appears to me that this analysis is strong evidence that the “anti-bunting” crowd needs to shut up.

        • Mickey Mantle's Outstanding Experience says:

          1. This analysis understates the Yankees run expectancy (before and after the bunts) because they have an above average offense.

          2. The reason the loss is so small is because it’s only 27 PAs. For comparisons sake, Girardi could have had Ivan Nova pinch hit for Brett Gardner 27 times and it would cost the Yankees about the same amount of runs as the bunts did. (This assumes Nova would hit like an average pitcher does) Would that be not be a mistake because it only cost the Yankees a few runs or even if they won 72% of their games?

  18. David in Cal says:

    Great analysis, Moshe. Of course, more could always be done. Presumably most of the bunts happened when either a single run was very important or when the batter had a poor chance against the pitcher. If these two items were somehow factored in, I think the bunts would show to greater advantage.

Leave a Reply

You may use <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <strike> <strong> in your comment.

If this is your first time commenting on River Ave. Blues, please review the RAB Commenter Guidelines. Login for commenting features. Register for RAB.