by Keith Law
Think about the dice game craps. Most bets at the craps table, like most bets in any casino, have an expected negative return on investment (ROI)—if you play the same bet repeatedly over a long enough sample, you can expect to lose money. There is one exception, however: the bet behind the pass line, which is one of the only bets you can place anywhere in a casino that has an ROI of zero, meaning you can expect to neither gain or lose any money on it. If you want to gamble, but don’t want to do something that almost guarantees you’ll walk out with less money than you had when you started, that’s your best bet, pun intended.
The out-versus-hit comparison is similar: The walk-off situations where an out is as valuable as a hit are the equivalent of that pass-line bet. They are the outliers, and can mislead you into thinking that there’s something to the “productive outs” concept. In reality, you would never strictly prefer an out to a runner reaching base safely, even if the latter doesn’t score a run immediately, because you’ve still increased the probability that you will score the run in the inning and/or increased the total number of runs you can expect to score in the inning. Hits and walks are good, and outs are bad. The fact that some outs are marginally less bad than others doesn’t outweigh the fact that making an out cuts your expected output for the inning, no matter the type of out in question.
A specific kind of “productive out” is, of course, the sacrifice bunt.
If you’ve spent any time on Twitter, you’ve probably come across baseball fans and analysts criticizing manager decisions to bunt, from the band Puig Destroyer’s crude but effective song “Stop Fucking Bunting” to my own term #smrtbaseball, an allusion to a Simpsons episode where Homer calls himself smart, spells it S-M-R-T, and then sets the house on fire. Here’s the main reason for all of this vitriol: in the majority of situations when it’s used, the sacrifice bunt is a terribly stupid play.
Defenders of the bunt, often generally advocates of the type of offensive strategy called “small ball” (so called, I think, because it produces smaller numbers on the scoreboard), claim that it helps “manufacture” runs by moving one or two baserunners up on the bases, making it easier for them to score later in the inning. This sounds somewhat logical, but doesn’t hold up to the most cursory scrutiny, as you can see from the run-expectancy table that shows the probability of scoring at least one run from each base-out state from the 2015 season:
The first column shows the runners on base; 0 means a base is empty, while 1, 2, and 3 refer to those specific bases, so “120” means runners on first and second but not third.
Let’s highlight some specific scenarios where you might see the bunt applied. The most common such scenario is a man on first and zero outs, which is a particular favorite of college coaches who find it easier to teach kids to bunt than it is to teach kids to hit. With a man on first base and no outs, a major-league team’s probability of scoring at least one run in the inning in 2015 was 0.499, or roughly 50-50. Pushing that runner up to second base in exchange for an out reduced those odds to 0.447, or just under 45 percent. So not only does the bunt reduce the number of runs the team could expect to score in that inning (from 0.84 to 0.65, as we saw in the earlier run expectancy table), but it reduces the team’s odds of scoring any runs at all. Remind me again what the point of the bunt was?
There are six common situations where you might see a manager call for a sacrifice bunt, so I’ve pulled those out and put them into another table that’s a little easier to read. Each row here describes the starting situation, the probability of scoring at least one run in the starting state, the same probability after a successful sacrifice bunt, and a cell saying whether the team was better or worse off as a result of the bunt. It’s not pretty.
In four of the six situations, a successful bunt makes the team worse off just in terms of pushing one run across in that inning. (Every successful sacrifice bunt reduces the team’s total run expectancy for the inning.) The only situation that is a net positive for the team is men on first and second base, zero outs, where the successful bunt slightly increases the team’s chances of scoring at least one run, albeit not by very much.
These tables are aggregate figures covering all situations for all teams, all hitters, all ballparks, and so on. A manager’s decision at a specific moment to call for a bunt should include those other factors, especially the biggest one of all: who’s at the plate. If we’re playing in a National League park and the pitcher is batting, his odds of getting a hit or drawing a walk are not that high, so a sacrifice bunt attempt is mathematically sound. (It’s also a damn good argument for putting the DH in the National League, because I have yet to meet the fan who bought a ticket to a major-league game because she really wanted to see guys drop some sac bunts.) If we’re in a high-offense environment, like Coors Field or minor-league parks like those in Albuquerque or Lancaster, California, the bunt makes almost no sense for nonpitchers because home runs are easier to come by and merely putting the ball in play is usually favorable for the hitter.
Bunting with your number-two hitter, as college teams are so fond of doing, ahead of your number-three hitter, who is usually the best bat in the lineup the way managers and coaches typically construct them, is incredibly counterproductive, especially when the number-three hitter has power. In the 2001 postseason, Arizona manager Bob Brenly repeatedly had number-two hitter Craig Counsell lay down bunts in front of number-three hitter Luis Gonzalez, who hit 57 home runs that season. In the third inning of game one of that year’s World Series, against the Yankees, Counsell bunted Tony Womack over from first to second; four pitches later, Gonzalez homered, scoring Womack from second base as easily as it would have scored him from first. Brenly gave away an out for no gain whatsoever, and surrendered the chance to put Counsell on base ahead of the home run. In game four, Counsell bunted Womack over three times for Gonzalez, with one of those leading to a semi-intentional walk of Gonzalez; those three bunts led to zero runs and Arizona lost that particular game by one run. If Counsell wasn’t a good enough hitter to try to reach base in front of Gonzalez, why was he hitting ahead of Gonzalez at all?
Aside from the pitcher batting, there are two other major exceptions to the whole “stop bunting” thing. One is the bunt for a hit, which is entirely separate from a sacrifice bunt—that is, a bunt for an out. If the batter is quick and/or a good bunter, and has a reasonable chance (somewhere above one-third) at a hit from a bunt attempt, then the attempt itself is often a smart move, especially since a good attempt can yield a wild throw or other misplay that might result in further advancement. The other exception is when a hitter attempts to take advantage of a poor fielder, such as when Miguel Cabrera played third base for the Detroit Tigers to accommodate Prince Fielder at first. Bunting on a third baseman who’s slow or who has an erratic arm has a higher rate of success for the team—reaching on an error won’t reflect positively in the hitter’s individual statistics, but it would help the team improve its outlook for the rest of the inning, since an errant throw would advance the runner(s) farther and perhaps allow a run to score.
These probabilities and expected outcomes make some manager and player tendencies particularly maddening because they are just so stupid. Cleveland shortstop Francisco Lindor, the runner-up in the AL Rookie of the Year voting in 2015, led the American League with 13 “successful” sacrifice bunts despite appearing in only 99 games. His OBP, which excludes those sac bunts entirely, was .353, so in 35 percent of his other plate appearances he reached base safely, as opposed to none of those 13 times he bunted a runner over. He did this while typically batting second in the order, one spot ahead of the team’s best hitter, Michael Brantley. And those bunts, on the whole, did not help the team score more runs.
6/22, 3rd inning, 0 outs, Jason Kipnis on 2nd, runner scored on single, no further runs scored, team lost by 3
7/5, 3rd inning, 0 outs, Kipnis on 2nd, runner scored on single, no further runs scored, team lost by 2
8/3, 1st inning, 0 outs, Jose
Ramirez on 1st, runner scored on single, HR later in inning, team lost by 1
8/5, 8th inning, 0 outs, Ramirez on 1st, runner did not score, team lost by 1
8/7, 1st inning, 0 outs, Ramirez on 1st, runner did not score, team lost by 1
8/8, 1st inning, 0 outs, Ramirez on 1st, runner scored on single, team won by 13
8/12, 1st inning, 0 outs, Ramirez on 1st, runner did not score, team won by 1
8/13, 1st inning, 0 outs, men on 1st and 2nd, runners both scored (SF, single, single), team lost by 2
8/17, 1st inning, 0 outs, Ramirez on 2nd, Lindor reaches on error, no runners score, team won by 6
8/25, 1st inning, 0 outs, Ramirez on 2nd, runner scored on double, team won by 5
8/30, 1st inning, 0 outs, Ramirez on 2nd, runner scored on double, team won by 7
9/10, 3rd inning, 0 outs, men on 1st and 2nd, runners did not score, team won by 2
9/19, 1st inning, 0 outs, Ramirez on 2nd, runner did not score, team lost by 1
In six of the thirteen instances here, the runner advanced on the bunt never scored, and in three of those six instances Cleveland lost the game in question by just one run. In five of the remaining seven instances, the runner advanced by the bunt would have scored anyway if subsequent events were unchanged:
• Kipnis is a fast enough runner to score from second base on most singles (6/22 and 7/5).
• On 8/3, Ramirez would have scored anyway on the home run later in the inning.
• On 8/25 and on 8/30, Ramirez would have scored anyway on the next hitter’s double.
That’s 11 of 13 so-called successful bunts that went for naught, and in all 13 situations, Lindor didn’t give himself a chance to get on base safely and create another run-scoring opportunity. He reached once on a fielding error, instead of the four times in 13 we’d expect to see him reach base, perhaps even driving in a run or two himself. Lindor apparently liked to do this on his own, which is somewhat appalling for a player who’s otherwise praised for his baseball acumen—he may really think he’s doing something positive. Someone needs to show Francisco the numbers and ask him the question every GM should ask his bunt-happy manager: “If what you’re doing makes us worse off, why are you doing it?”
The hardest myth for fans and even baseball folks to give up, in my own experience, is the myth of the “hot hand.” The term itself derives from basketball, but there are many equivalents you’ll hear in baseball, from “in the zone” to “locked in” (an unfortunate term given its second meaning in medicine) to “feeling it.” The story goes that a player, nearly always a hitter, has somehow captured some extra magic and is on a run at the plate in which his results are consistently better; he’s hitting the ball harder, or balls are falling into play. You’ll hear announcers say a pitch must have looked like a beach ball to him, or that his confidence is through the roof. The problem with this myth, as with the others, is that the evidence from reality shows that this effect either barely exists or doesn’t exist at all. It’s merely our brains’ attempts to find patterns in data that are pretty close to random.
Hitter performances throughout a season include a lot of randomness; there certainly isn’t any predictability in when a hitter will be more or less productive—like, say, Joey Bagodonuts is a really great July hitter, or Twerpy McSlapperson is much better on Tuesday afternoons. You expect that a hitter’s line at the end of a full season will probably look a lot like his lines from the season or two before that; the more at bats a guy has, the more we expect his production to look like whatever his underlying true talent level might be. A hitter whose “true” on-base percentage, meaning his natural ability to get on base, is around .400 could easily have a month where he only gets on base at a .300 clip. You might think he’s slumping, but in reality it’s just a normal turn of the baseball screw: every hitter has periods of highs and lows, but barring some other explanation like an injury, players will eventually return to form—the baseball version of what statisticians refer to as “regression to the mean.”
However, our brains are not well wired to handle randomness; we tend to think random distributions look more like uniform distributions, when random series tend to be clumpy and uneven, giving the impression of pattern where there’s none. Alex Bellos, who’s written several wonderful books on mathematics for the popular audience, wrote in London’s Daily Mail in December 2010, “As humans, when we come across random clusters we naturally superimpose a pattern. We instinctively project an order on the chaos.” He brought up the example of the random song shuffle feature on the original iPods, which produced random results like clusters of songs by the same artist, which, of course, confused listeners who thought the system was therefore not random. Apple had to make its randomizer less random to convince people it was random enough.
This is what happens in baseball—and other sports, as far as I can tell, although it’s heretical for me to even admit the existence of other sports—when a hitter has some sort of “hot streak,” “slump,” or anything that doesn’t seem random to our pattern-seeking minds. (In fact, you hear a lot that a certain hitter is “streaky”; all hitters are streaky, because random distributions are not uniform. If a hitter never had a streak, I’d suggest we all bow to our new robot player overlords.)
A landmark 1985 study of streakiness in basketball by Cornell psychology professor Thomas Gilovich and Stanford professors Robert Vallone and Amos Tversky looked at data from several players and teams in the NBA and found no evidence of the so-called hot hand effect: a player who has made several recent shots is, in fact, no more likely to make his next shot. If you like semantic games, you would say that he has been “hot,” but he is not “hot.” That is, he’s made a few shots in a row, but he is still the same shooter he was before that streak of success.
Indeed, many researchers continue to try to find evidence of the hot hand effect in various endeavors, as if we just don’t want to accept that our sports narratives are untrue. (Cf. “On the possibilities of self-propelled aerial locomotion,” by Kelly, R., 1996, Atlantic/Jive Records.) Many attempts to identify and quantify the hot hand effect have appeared, with a recent paper from two economists, Jeffrey Zwiebel and Brett Green, claiming to have found evidence of this streakiness in baseball at a magnitude previously unheard of in the field. It didn’t hold up under scrutiny: they used small samples to establish the players’ base ability levels and thus tended to confuse changes in true talent levels with streakiness. The resulting consensus among sabermetricians who examined this study was that there might be some “hot hand” effect in baseball, maybe, but if it’s there, it’s minuscule, to the point where managers could safely ignore it.
There are real changes in player performance that teams need to monitor. Players get hurt and see temporary changes in skill levels. Hitters who suffer hand or wrist injuries lose some of their grip strength and thus may see a decline in power or in the quality of contact, while pitchers fighting just about any kind of malady can see a change in their mechanics and lose velocity, the ability to locate pitches, or both. Mechanics can change for the better, typically between seasons but sometimes during the year. Pitchers can add new pitches or change grips. Hitting and pitching coaches work with players on all of these things as part of their jobs, often reviewing video after each game and sometimes between at bats or innings to identify tiny yet tangible differences that might prevent the player from performing at his optimal level. (I was in the clubhouse occasionally for some of these conversations while I was working for the Blue Jays. Some were beneficial, but others involved a lot of complaining about the umpires, too.) This is why teams long employed advance scouts, watching the next series’ opponent to see who might be healthy, who didn’t seem to have a good grip on his breaking ball, who wasn’t running as well as he normally did, and so on. Today, some teams have dispensed with advanced scouts in favor of sophisticated video and computer systems that can track plays and players more thoroughly, often with young front office employees revie
wing the results to provide detailed notes for the major-league coaching staff.
But none of this is “hot hand” nonsense—it’s looking for physical, verifiable changes, not arguing that some player has been temporarily possessed by the spirit of Honus Wagner at the plate. So the next time you hear that a hitter is locked in or that a starting pitcher is cruising, just remember, it’s probably a narrative sitting awkwardly on a pile of random data.
This appeal of the narrative says a lot about why all of the myths here are able to persist in the absence of data. We want baseball to be the sepia-toned sport of our constructed memories. In a game that goes back at least 150 years, there is no shortage of stories for the sharpshooter to pick to try to support his preconceived notions about the game or his desire to find a narrative in the noise. We romanticize the past, we canonize the players we like (while demonizing those we don’t), and we pick and choose what stories to remember or to believe. This irrationality drives so much of baseball coverage and discussion because of what we want baseball to be, when the game provides plenty of real, fact-based stories that are rooted in truth, not mythos.
For a long time, we lacked the information or the tools to adequately dispel these ideas about baseball—and other sports as well—to the point where we might see behavior change, but the last fifteen years have seen a revolution in baseball thinking. This upheaval in baseball thinking, from using better metrics based on the traditional stats to the incoming wave of new data that’s changing the way front offices build their teams, won’t get rid of great stories. We’ll get new great stories now, stories about underdogs who were overlooked by traditional methods of evaluation but whose hidden talents came out because of the new data.