Why look at the data when I agree so vehemently with the headline?

TFA is an interesting close reading of how several bunting situations played out to the bunting team's disadvantage.

Along the lines of Jim's comment, I'd look more favorably on a bunt if the score is 2-1 and you're in a week of 2-1 games where it seems impossible to touch either pitcher. If it's 6-4 (as one of the TFA situations was), and the park and era and pitching matchups conduce to bunches of runs, then yes, you're throwing away an out.

Particularly I hate the bunt in the first inning after the first two batters reach base. Seems like letting the pitcher off the ropes.

I hate the bunt in the first inning after the lead off man doubles. Yeah Dusty I still hate you.

I hate the bunt in the first inning after the lead off man doubles. Yeah Dusty I still hate you.

Gotta get that first run of the game, Dude. That's why we have Jose Macias in that spot. He does the little things.

Why look at the data when I agree so vehemently with the headline?

Indeed: bunt rates are lower this year than last: 0.8% of all ABs vs. 0.9% last year, and is probably close to the all-time low (I checked back about 12 years and couldn't find anything lower).

When I played OOTP I frequently bunted with runners on 1st and 2nd because GIDPs really angered me. I don't think it helped my team's run-scoring, but I got less angry. A sac bunt and a pop-up or strikeout was less annoying than a GIDP.

I would like to see some data on this... you can't look at RE when evaluating sacrifices, you can probably look at WPA, although even that is a little tricky. The primary reason for the sacrifice bunt is not to increase your expected number of runs or even the probability of winning. The goal is to increase the probability of scoring at least one run (which is not the same thing as increasing RE or of increasing WPA) (or equivalently, decrease the probability of scoring zero runs).

When looked at from this perspective, using something like RE to evaluate bunts is horrible. I suppose you could use

I accept that the sac bunt has almost no place in the first 6 innings of a baseball game... there's just too much to play out to concede the possibility of lots of runs for the decreased probability of no runs in a particular inning. I would guess that in innings 7-9, this would be much more important. I would think that WPA would take this into account, but perhaps not because that doesn't take into account the particular attributes of the team. For example, I would conjecture that a team with a great bullpen should probably bunt more often than one with a bad bullpen, because one run is much more likely to be the difference for those teams.

This is a test.

I would never sacrifice in the first inning. Never. Sticking a bad hitter in the 2 spot so he can make an out in the first inning before you know if the pitcher can get anyone out or not. For example, look at the Padres yesterday.

As far as the probability of scoring at least one run per #7 above; there was a book called PerCentage Baseball, many years ago; which had those empirical stats. I believe a runner a successful sacrifice from first to second raised the chance of the runner scoring from 42 to 45 %, obviously while eliminating the chances of the batter scoring. Second to third actually reduced the runners chances.

Every run expectancy table I've seen looks at total runs scored on average. Is there a table that looks at the likelihood of ANY run scoring? Ex., with runners first/second and none out a runner scored say, 40% of the time?

I would not bunt with no one out, except with a pitcher, because the chance of a big inning is so high. After all, all you know about the pitcher is that he hasn't gotten anyone else out yet so far this inning. And I strongly prefer bunting with men on first and second to bunting with just a runner on first. If I'm giving up an out, I want TWO bases for it. This all changes, of course, in the 8th and 9th innings of one-run or tie games, where one run assumes disproportionate value. - Brock Hanke

I didn't read the article as assailing the sacrifice bunt as much as lambasting poor, uninformed managerial judgement. He even opines that he hopes that more managers are hired in the next 5-10 years who use their noggins rather than falling on strategies that are common, but unproductive.

Also, a run in the 1st inning counts exactly the same as a run in the 9th inning mathematically, even though the offense has fewer outs remaining late in the game. Aside from run expectancy, I don't see how a late inning run has more value.

Particularly I hate the bunt in the first inning after the first two batters reach base. Seems like letting the pitcher off the ropes.

Well, bunting with your #3 hitter is stupid in pretty much every circumstance. The way most lineups are structured, he's one of your top-2 hitters, and probably never learned to bunt.

Tango, MGL and Dolphin's Book is full of this stuff. If you're at all interested in this stuff, you need to read it.

Every run expectancy table I've seen looks at total runs scored on average. Is there a table that looks at the likelihood of ANY run scoring? Ex., with runners first/second and none out a runner scored say, 40% of the time?

I think Table 121 (Page 293 in this link) in said book will tell you this, but there is a whole chapter devoted to sacraficing.

For example, it says the odds of at least one run scoring with 0 out, man on first are 44.3%. 1 out, man on second is 41.4%, which seems so wrong that I'm sure I messed something up.

The primary reason for the sacrifice bunt is not to increase your expected number of runs or even the probability of winning.

I agree with the first statement, but the second is crazy. If it doesn't increase your probability of winning, don't do it.

The only reason to sac bunt is b/c the one-run has a much bigger impact on the P(win) that the subsequent runs would, i.e. trading expected runs for certainty of one increases the P(win).

I seem to remember a study that looked at actual cases where teams bunted (rather than raw run expectancy tables), and found that bunting led to more runs than the run expectancy tables would have predicted -- i.e., that managers were at least somewhat selective in when they had their teams bunt, and that this led to an improvement in results. Anyone know what I am talking about?

Also, a run in the 1st inning counts exactly the same as a run in the 9th inning mathematically, even though the offense has fewer outs remaining late in the game. Aside from run expectancy, I don't see how a late inning run has more value.

Because in the late innings you've got a pretty good handle on the run context of that game. In the early innings you don't know if the game's going to be a pitchers' duel or a slugfest.

If you're accepting a lower expected numbers of runs, for a greater prob. of getting just one, you have to be fairly confident that one run meaningfully changes your win prob. In the 8th inning of a 1-1 game, you know that one run is very, very valuable in terms of winning the game. In the 2nd inning of a game that may turn out to be 11-8, you can't be confident at all that one run is going to do it. You may need three.

I seem to remember a study that looked at actual cases where teams bunted (rather than raw run expectancy tables), and found that bunting led to more runs than the run expectancy tables would have predicted -- i.e., that managers were at least somewhat selective in when they had their teams bunt, and that this led to an improvement in results. Anyone know what I am talking about?

Yes. Run-expectancy tables show you the results of a successful bunt (e.g. man on 2nd one-out, vs. man on 1st, zero out). In reality, the range of outcomes is much larger (hit, errors, etc.).

Table-based analysis will also assume average bunt success, whereas actual teams are more likely to bunt with good bunters, and defenses that are not playing for the bunt.

Just wondering how they account for unsuccessful bunts. If they only look at cases where the teams bunt successfully doesn't that skew the results?

Yes. Run-expectancy tables show you the results of a successful bunt (e.g. man on 2nd one-out, vs. man on 1st, zero out). In reality, the range of outcomes is much larger (hit, errors, etc.).

And it also doesn't show the results of falling in the hole because the incompetent bunter fouled off the first two pitches.

I'm generally in line with Brock's position on the bunt (with the occasional deployment of the early sacrifice to keep the defense honest), though I wouldn't mind seeing more speedy guys pseudo-sacrifice with men on base (drop on down with the hopes of getting a base hit, but settling for the advancement).

If you're accepting a lower expected numbers of runs, for a greater prob. of getting just one, you have to be fairly confident that one run meaningfully changes your win prob. In the 8th inning of a 1-1 game, you know that one run is very, very valuable in terms of winning the game.

This assumes that your bullpen is capable of holding that thin lead or at least not letting the game turn into a laugher.

This assumes that your bullpen is capable of holding that thin lead or at least not letting the game turn into a laugher.

Assuming the quality of you bullpen as a constant, the one run still has a huge impact on the P(win).

If your bullpen is a total clown show, you're f-ed at 1-1 anyway.

I wonder how much Wedge was influenced in the example above by the successful sacrifice bunt in the ninth inning of the previous night's game. In that situation, the bunt made sense. Tie game, bottom of ninth, runner on second, none out. Kyle Seager (the M's hottest hitter) bunted Kawasaki over to third, where he could score the game-winning run on a fly ball - and that's exactly what happened, with Jaso hitting the sacrifice fly. But that's a heck of a different situation than being down by two runs and bunting, which is what happened on Tuesday when Wedge called for the bunt and Ackley couldn't get it down.

I seem to remember a study that looked at actual cases where teams bunted (rather than raw run expectancy tables), and found that bunting led to more runs than the run expectancy tables would have predicted -- i.e., that managers were at least somewhat selective in when they had their teams bunt, and that this led to an improvement in results. Anyone know what I am talking about?

I suspect you thinking of Empirical Analysis of Bunting, by Dan Levitt. He actually breaks it down by lineup position, and includes the probability of one run tables that #7 and #9 were looking for. IE,

Table 5 indicates a number of cases in which a successful bunt increases the probability of scoring a run. In the AL, when the ninth place batter bunts with a runner on first and no outs, the probability of scoring at least one run moves from .423 up to .441.

I've never found the bunt studies very convincing in that they never do a good job of accounting for all the possibilities. The bad: popping up for an easy out, fouling pitches off and getting behind in the count. The good: the bunt becoming a hit, an error made fielding the bunt (happens more than you would think, especially since the pitcher is usually involved), other defensive miscues like unsuccessfully trying to get the lead runner, the pitcher altering his pitch selection to throw hard-to-bunt but hittable high fastballs.

That said, only replacement level hitters and below should ever be sac bunting.

zack (#13) has, "For example, it says the odds of at least one run scoring with 0 out, man on first are 44.3%. 1 out, man on second is 41.4%, which seems so wrong that I'm sure I messed something up." You did the math right. What the numbers are telling you that the one base you gained to get to man on 2nd, 1 out, was worth less than the out you gave up to gain the base.

snapper (#16) ha a comment that gets close to my reasoning. Mine goes like this: If you're down one run in the 1st, you've got 8 or 9 innings to get that run back. But, If you're down one run in the 9th, you have to get that one run now, or you lose the game. That's what mean by saying that the run has more importance in the later scenario. Among other things, it explains why stolen bases were high in the dead ball era and in the 1960s. In both periods, there were very few runs scored per game. Therefore, runs were expensive and outs were cheap. When that happens, stolen bases have more value, and the outs they cost, cost less. - Brock

i really REALLY hate the sac bunt when there are 2 on, zero or 1 out, and a really good hitter like jose altuve or jed lowrie, ESPECIALLY WHEN THEY HAVE ALREDY GOTTEN A HIT OFFN THAT PITCHER, are sent up to sac bunt. why make an out on purpose with one of your 2 best hitters????!!!!

sac bunting is fine if it is chris snyder/jason castro. you notice that chris snyder and his 38 OPS+ is LUVVVVVVED by this Organization and has like zero chance of being sent down/DFAd no matter HOW many PA, unlike, say JR towles, who could at least hit for power

Managers need a SBA (successful bunt attempted) statistic to help them decide whether they should try it or not. It's mostly a borderline call anyway, but all the time I see it attempted with guys that obviously have never tried it before or can't do it. Why not know what your odds of success are before you decide?

#18 addresses this too..also, when the batter goes to two strikes while failing to bunt, and then hits away, it reduces the likelihood of success considerably..but this isn't factored into any success ratio with regards to bunting.

Why should we assume that managers are capable of perceiving small differences in win expectancy between bunting and swinging away with certainty, and, furthermore, able to adjust these small differences to consider the players involved? Instead, perhaps we should assume that they use heuristics—information shortcuts—to help make tactical decisions, like sacrifice bunting. Additionally, the types of information shortcuts that managers regularly use in tactical decisions should vary based on how they tend to frame the outcome of a game situation. This comes from prospect theory, which argues in part that people act irrationally to avoid losses because a loss would feel worse than a gain of the same amount would feel good. For instance, the negative feeling one gets from losing $100 is greater than the positive feeling of gaining $100. Thus, managers who bunt more often (the Bob Brenlys of the world) may have a higher degree of loss aversion toward game situations than those who do not. In other words, the decision to bunt is a decision to forgo a chance at the gain of scoring multiple runs because of the fear of scoring zero runs. The tendency to wait for the three-run homer, other things being equal, is the less expected tactical choice. Late in close games, this loss aversion becomes even more salient, so managers might become even more inclined to sacrifice despite the possibility of greater expected utility from swinging away. This might explain a key element in what makes a successful tactical manager: framing game situations in terms of gains more often than their competitors, true talent of personnel being equal.

I like the prospect theory angle.

But in general, managers should know the run expectancy values, but they should not be so predictable as to adhere to them. Tactical and strategic unpredictablity itself has value.

Bunts are awesome to watch unfold. They should make a rule that says a runner reaching on a bunt is awarded second base to encourage more bunts.

Until then, someone show this link to Mike Matheny, STAT!

The use above of the phrase "a really good hitter like jose altuve or jed lowrie" makes me wonder, how many teams have had their 2B and SS be their two best hitters? (by OPS I guess)

That's a fun question to ponder. Without looking anything up, I can't think of much. Some random and very wild guesses:

1948 Dodgers
1997 Red Sox
1870s Red Stockings
Giants and Cardinals of the 1920s
1980s Tigers
Was there a war year when Doerr/Pesky played but Williams didn't?
Kent and Aurilia were really great one year but I think some other guy was on the team
Rangers with Arod and Young
2006 Yankees, or whatever year that was when Morneau screwed Jeter out of the MVP

The use above of the phrase "a really good hitter like jose altuve or jed lowrie" makes me wonder, how many teams have had their 2B and SS be their two best hitters? (by OPS I guess)

Did ARod play with Boone druing any of Boone's great seasons?

Edit: Nope. Missed by one year.

Most of my guesses were off but the 1871-73 Red Stockings, 1983 Tigers and 1988 Tigers turned out to be correct.

how many teams have had their 2B and SS be their two best hitters? (by OPS I guess)

Honus Wagner tied for the second highest OPS+ on the 1913 Pirates, who were led by second baseman Jim Viox. Wagner led the 1915 Pirates in OPS+, and Viox was tied for third. Wagner would have done it with two other second basemen in 1906 and 1909, if it hadn't been for Fred Clarke getting in the way.

Did ARod play with Boone druing any of Boone's great seasons?

I also thought about A-Rod with the Rangers. Didn't remember Michael Young being such a lousy hitter early in his career.

Any years where Sandberg and Dunston were both team leaders?

Any years where Sandberg and Dunston were both team leaders?

Not even close. Dunston's career high OPS+ (for anything close to a full season) was 108. I'm not sure he was ever better that the fifth or sixth best hitter on his team.

Whitaker and Trammell 1983. Maybe other years too, but I've only looked up to '83.

EDIT: Also '88 (Whitaker a bit short of 3.1 PA/game) and '93 (both short of PA)
EDIT 2: Missed God's guesses in #32/34. But even He didn't get 1993.

Also in the not-quite category: 1918 Cardinals (Hornsby SS, Bob Fisher 2B: short on PT but 3x the games at 2B as anyone else)