Let’s just …
Warning! Gory Mathematical Details Ahead
You probably think that The Shift works because you saw a chart that looks something like this:
|Year||GB BABIP vs. The Shift||GB BABIP, no shift|
Those are league-wide numbers and they show that when a ground ball is hit in front of The Shift, it’s turned into an out more often than if it’s hit in front of an un-shifted defense.
Or maybe you saw this chart:
|Year||Overall BABIP vs. The Shift||Overall BABIP, no shift|
You reasonably figured that if BABIP was lower against The Shift, then it must be doing something right. I have bad news for you. We need to talk about selective sampling.
Here’s a list of names: Anthony Rizzo, Kyle Seager, Mike Moustakas, Mitch Moreland, Jay Bruce, Rougned Odor, Carlos Santana, Kendrys Morales, Justin Smoak, Kole Calhoun, Logan Morrison, Matt Carpenter, Carlos Gonzalez, Curtis Granderson, Nomar Mazara, Shin-Soo Choo, Albert Pujols, Matt Joyce, Alex Gordon, Freddie Freeman, Lucas Duda, Brian McCann, Yonder Alonso, Eric Hosmer, Jake Lamb, Chris Davis, Cody Bellinger, Carlos Beltran, Victor Martinez, Brian Dozier, Brandon Belt, Giancarlo Stanton, Scott Schebler, Jackie Bradley, Edwin Encarnacion, Seth Smith, Kyle Schwarber, Kris Bryant, Justin Bour.
What do these 39 men have in common? In 2017, they all hit at least 200 balls in play against a shifted defense. Also, most of them are big, slow power htiters. There’s nothing wrong with being a big, slow power hitter, but those guys tend to have two things working against them in the BABIP department.
Being power hitters, they tend to hit more fly balls than average. Fly balls are easier to catch and turn into outs than grounders, and the ones that they hit over the fence are specifically removed from their BABIPs, despite being a successful outcome. Being slow, they tend to leg out fewer infield hits than average. I think we’ve gotten so used to the DIPS idea that all BABIPs are .300 plus or minus some random error. That’s (kinda, sorta) true for pitchers, but it is most certainly not true for hitters. There are high-BABIP hitters and low-BABIP hitters, and we recognize that as a repeatable skill.
I looked at these 39 men and found that they had a combined BABIP of .279 in 2017. It wasn’t that The Shift was taking away all of their joy. In fact, in plate appearances where these 39 men did not face The Shift, they had a BABIP of .275, while against The Shift they had a higher BABIP of .280. If you take a group of hitters who have a BABIP of .275 and do something to your defense (i.e., The Shift) that raises their BABIP, that’s not an accomplishment. The fact that this number is below the league average is irrelevant.
In the past, I’ve written on this topic and I think that we have our metrics for success on The Shift all wrong. Worse, they’re telling the wrong story. There appears to be little evidence that The Shift, as a defense against balls in play, is actually superior to the two-left-two-right defense. It’s very likely an inferior one. The Shift appears to “work” because the hitters who get shifted against are (on average) lower-BABIP hitters. We need to burn The Shift.
You may be wondering why that is. After all, the idea behind The Shift was that if you look at a spray chart, and if a hitter is going to hit the ball to the right side all the time, you might as well put an extra guy over there. And for what it’s worth, The Shift probably does snare a few extra balls that might have previously gone for hits. The problem is that there are other effects that no one talks about.
Mostly, I think it freaks the pitchers out.
The methodology to look into this involves cross-weighting the data within an individual hitter’s (or pitcher’s) own performance. The short version is that if a hitter has a BABIP of .300 in situations where the defense isn’t shifted, and he has 100 balls in play against The Shift, if the two defenses are equivalent, then he should get 30 hits. (If he has more hits than that, then The Shift isn’t working for him. If fewer, then it works fine.) That gets us beyond the selective sampling problem. Within one hitter, there’s going to be some random noise, but if we sum across the league, we can look at the effects of The Shift in the aggregate. We don’t have to just use BABIP, although when we do use BABIP, consistently, we see higher BABIPs in front of The Shift than we would otherwise expect.
We can do any stat, as long as we have data, and it turns out that the story is told in more numbers than just BABIP. When we look a little deeper, we see two concerning problems, which I showed earlier this year. One is that hitters (as a group) hit more line drives in front of The Shift than we might expect based on their non-shifted performance (and pitchers give up more than we would expect). The other is that we see evidence that pitchers throw more balls in front of The Shift (and batters receive more balls) than we might expect of either of them. Those two facts are probably related. If a pitcher is throwing extra balls to a batter, then the batter is going to be in a hitter’s count more often, and we know that BABIP is better when the batter is in a better count.
We can even put a run value on that extra ball, thanks to the work of researcher Dan Meyer. It checks in around .068 runs. (That’s lower than the oft-quoted figure for a successful catcher frame, as the frame both takes a ball away and adds a strike.)
We have data on how many balls were involved in the plate appearances where a batter saw The Shift. Or at least some of them. Only plate appearances that end “in play” (i.e., not home runs, strikeouts, hit batsmen, or inconveniently enough, walks) are included. So, we will compare how many balls per plate appearance a hitter saw in his non-shifted (and in-play) plate appearances and compare it to his shifted (and in-play) plate appearances, and sum across the league. (We can do the same for pitchers.)
If our reference group is hitters (using 2017 data), they see 0.044 more balls per PA. For pitchers, it’s 0.064. (The difference is because I had to set some minimum PA cutoffs.) We might be undershooting that a little, because we’ve specifically excluded the type of PA that by definition has the most balls in it, the walk, but let’s split the difference of what we have and say that The Shift increases balls by 0.054 per plate appearance. That increase is not going to be visible to the naked eye, because we’re talking about an extra ball once every 20 PA or so. It might not even be apparent to the pitcher himself, and it’s not like the extra ball comes with a tag on it that says, “I’m the extra one!” in the same way that it’s pretty obvious if a ball scoots through the hole in the right side where you were thinking about putting an extra infielder. But the data set can see those extra balls, even if you can’t.
If there’s something that we learned from catcher framing, it’s not to ignore a small effect that impacts a high-frequency event. And as has been pointed out several times, The Shift has become a very frequent event. In 2017, there were 26,700 balls in play hit into The Shift. We don’t have numbers on not-in-play events, but of course, there were times when the defense shifted, but the batter politely struck out. There were a total of 121,494 balls in play in 2017, meaning that 22 percent of all balls in play were hit into a shifted defense.
Let’s assume that 22 percent of all plate appearances (and in 2017, there were 185,295 of them, league-wide) involved a shift. That’s 40,764 shifts league-wide or 1,359 per team. If every shift is adding .054 balls per PA, and each added ball is worth .068 runs, then The Shift is costing the average team about five runs per year. There are some teams that are shifting a lot more than that, meaning that there are teams that could literally make themselves a win better by deleting their shift key. And we all know that a win is worth $70 gazillion (give or take).
That effect is mediated through a mechanism that’s hard to see. Maybe pitchers nibble a little more with The Shift on. Maybe they don’t feel comfortable with one of their pitch-location combos because they’re worried about the batter getting a hold of it and poking it the opposite way … and no one’s there right now. Maybe it makes him a tiny bit more predictable. It probably wouldn’t even register to him as something that he’s not doing. It’s just that he pushes that option down in the preference order a little further.
Sure, when the batter rolls one over to the right side, it’s easier for that extra fielder to pick it up, and that part is really easy to see. But if that extra fielder is causing something to happen in the pitcher’s mind that more often puts the batter into an advantageous count where he can make better contact and hit a line drive instead of a ground ball (or he can walk to first, since you can’t throw him out then), then that extra infielder on the right side is doing more harm than good.
All Hands Abandon Shift!
Reader, if you might indulge me for a moment while I break the fourth wall, because there are people in front offices that are going to read this one. (Hi!) You can interpret the above findings in one of two ways. One is that you should stop shifting, or at least significantly cut back. The other is that the effect could be countered by “working with” your pitchers. And in the sense that this is possible (and maybe some of you have already had “The Talk” with your pitchers) you’re correct. I would caution you against relying on that second option as a way to say, “We’ve got this covered.”
We are talking about a very subtle effect, one that really only has the power that it does because of the volume of shifts. It would be a lot easier to detect if the active ingredient were that pitchers were doing something that we could all point to and that immediately led to 500-foot home runs. Instead, I’m saying that perhaps pitchers are doing something a little more often or shying away from doing something that they might not have done anyway and it’s leading to an extra ball. And it’s not like we are trained to yell at a pitcher for giving up a ball. They happen and we’re forgiving of the fact that they sometimes happen, but they do have consequences.
The choices that pitchers make are the result of learning that’s probably taken place since their days on a Little League field. It’s not that anyone sat down with a textbook and taught them these things. These are the kinds of behavior patterns that show up after a couple of decades of experiential learning. Pitchers learned to pitch in front of a two-left-two-right defense, and still take most of their turns in front of a traditional alignment. It’s silly to believe that putting them in front of a shifted defense doesn’t change that equilibrium, even if only a tiny bit. Maybe it is only a little bit, but math is sneakily not your friend on this one.
You can chat with your pitchers about “pitching the same way you always have” and they may even nod and swear that they are doing it. That would be the Full House model of behavior change, where change can be affected with a two-minute heartfelt conversation, some sappy music, and a hug from Uncle Joey. Maybe you guys hug your pitchers all the time. That’s not how behavioral change works, especially when you’re dealing with something that has been reinforced constantly over time and since well before they came into your system.
We all have behavioral patterns which, if we thought about them, we would realize that we have no clue why we do things that way. However, if we tried to break that habit, it’s a rather big cognitive lift. People have all sorts of bad habits that they realize they need to get rid of, but can’t. Take it from a former therapist, or just someone who has his share of bad habits.
And even if your pitchers are aware of what’s going on, everything lines up for them to minimize it. “Talking to them” probably involves a conversation that starts with: “Hey, we’ve noticed that when we have The Shift on, you change the way that you pitch a little bit. You get a little more tentative and nibble a bit more.” Hyper-competitive, 28-year-old men are not widely known for their abilities of understanding and responding to their own emotional states, particularly when you’re basically accusing them of being afraid. And it’s easy to say, “Well, if it’s just a little thing, I can fix that!”
On top of that, I’m sure there’s some Lake Wobegon Effect going on here as well. While all of these things might be problems league-wide, you guys have a “special” way of working with your pitchers, and it’s the other 29 teams that are driving these findings. Right? You’d better be right about that. Someone’s clearly still screwing up.
Many of you have been using The Shift for years. It’s become part of your organizational philosophy and we here at BP have spent so much time giving out gold stars to teams who were at the top of The Shift list. I think we’ve had it backwards this whole time. The Shift was new and shiny and counter-intuitive and made mom and dad mad. That isn’t the same thing as being a good idea.
Maybe there’s room to say that in a specific case where a hitter who was so pronounced in his pull tendencies or a pitcher who was particularly aware of his own emotional state and how it affects him was on the mound, a shift might be warranted. But it’s a lot harder to make the determination on the latter part than you might think, and on the former piece, when I ran the analyses with the 39 hitters who were most shifted against (presumably, the ones with the most “shiftable” profiles), all of the effects that I’ve been talking about were still there.
If I were given a choice between using The Shift as it is now and never shifting, based on the available data, I would recommend the latter. It doesn’t matter what the spray chart says. It matters that putting a third infielder on the right side causes other problems that outweigh the benefit you get from the third infielder on the right side.
Burn The Shifts.