Baseball Therapy: Why The Shift Persists

Image credit: MLB.com

Last month, I suggested that The Shift—the thing where a team puts three or more infielders on one side of the infield—was actually doing more harm than good. And yet, The Shift persists. I still think those findings are correct, but it seemed to me that there must be more to the story, and so I set out to find it.

1. The Shift does have a big effect, just not the one you think.

I don’t know if this question has been fully answered: What is the benefit of The Shift? Sure, we know that it’s basically meant to station an extra fielder, usually on the pull side, who will hopefully snag a few grounders and soft line drives. If a batter hits a bunch of balls that way, it makes sense. We also know intuitively that some hitters make sense to shift against and some do not.

To take an extreme case, if we knew that every time a hitter put the ball in play, he would hit it to the rightward third of the field, then a manager should station all seven of his non-battery fielders in that rightward third. If a hitter is a true spray hitter who sprinkles the ball all over the yard, then shifting against him makes no sense, because while the balls he pokes to the right side will be more easily field-able, the balls that he pokes to the left side will have giant holes that they could find.

I think we’ve all assumed that the size of the hole that’s created on the left side by moving the third baseman over to the right side of the field is equal and opposite to the size of the additional barrier that the third baseman puts up on the right. This turns out not to be correct.

For these numbers, I looked at all cases in which a left-handed hitter (more than two-thirds of all shifts were executed against lefties in 2017) hit a ground ball, and paneled the data by whether the ball was pulled, hit to the middle of the field, or hit to the opposite field, plus whether The Shift was on. (Data is from 2012-2017.)

Grounders:

Line Drives:

Fly Balls:

(Reminder: Does not include home runs.)

We see that ground balls hit into The Shift on the pull side are about 5.5 percent more likely to be turned into an out than if no shift is on, and the effect for liners is about as big. Those are the extra outs that The Shift is supposed to be producing. A ball pulled into The Shift really is easier to field. But then there’s the deep, dark fear that goes along with The Shift, which is that the hitter will go to the opposite field … and they’re a little short-staffed over there at the moment.

The numbers bear that out, too. Grounders hit the opposite way have a whopping 15 percent better chance of becoming a hit when The Shift is on than when teams are playing a two-left-two-right alignment. If we’re going to be opening up a 15-point hole in one side of the infield and only creating a five-point barrier on the other, then we’d better be sure that the batter isn’t much of a threat to actually hit the ball toward that 15-point hole.

BABIP isn’t the only way to do these things. In fact, it’s probably not even the best way to do things. Doubles and triples are worth more than singles, so I’ve also included weighted runs created (wRC) above to account for this, and we have the numbers above, normalized to per plate appearance rate (or technically here, because the publicly available data doesn’t have HR, BB, K, or HBP, per ball in play.)

We can now transform those numbers into estimates for how valuable The Shift is, compared to the two-left-two-right formation, when a batter does each event.

If a batter does pull a ground ball into The Shift, his outcomes are going to be .044 runs worse (or, if you look at it from the defense’s perspective, outcomes are .044 runs better) than if they were playing two-left-two-right. However, if he goes oppo, The Shift costs the defensive team .126 runs. The defensive team needs to believe that a hitter will pull his grounders about three times as often as he will go oppo with them in order for The Shift to break even.

Some guys are extreme pull hitters like that and we can project based on how often a batter hits ground balls and line drives, and how often he hits to the pull, center, and opposite sides. We can figure out what his expected benefit (or loss) would be for a team switching to The Shift on him.

In 2017, there were 318 hitters who logged at least 250 plate appearances. This includes both lefties and righties. When we do a quick multiply out their 2017 rates of pull/center/oppo and their GB/LD/FB rates against the above chart, we find that about half of them (147) have a profile that would make shifting a good idea, although some of them just barely. But that assumes that hitters hit exactly the same in front of The Shift as they do when The Shift isn’t on.

Which brings us to …

2. Hitters change their approach in front of The Shift.

Hitters, of course, aren’t unaware of the fact that there are three infielders on one side of second base, and while they have something about their swing that makes them pull-happy, they are hopefully able to tame some of those impulses when the situation calls for it. The Shift gives them plenty of incentive to do so, and there’s evidence that they’re doing exactly that.

Now, it’s silly to look at rates of pulled balls in front of The Shift and in front of a two-left-two-right defense. There will be more pulled balls in front of The Shift because teams specifically put The Shift on when a pull hitter is up, causing a massive selective sampling problem. But we have a way around that. We can use cross-sampling.

Here, we’re going to take advantage of the fact that hitters have plate appearances both in front of The Shift and without The Shift. If a batter pulls 40 percent of his balls in play when The Shift isn’t on, then in theory, he should pull 40 percent of his balls against The Shift if he’s doing nothing differently. For another hitter, his pull percentage might be 45 percent not against The Shift and we will change his expectations accordingly.

An individual player’s actual results will be subject to a lot of small-sample-size weirdness. If a batter only has 10 plate appearances against The Shift, it’s easy to get the occasional 70 percent pull rate the same way that it’s easy to get the occasional run of seven heads out of 10 coin tosses on a fair coin. However, that weirdness tends to even itself out over an entire sample of players. So, we will sum across the league. This does unfortunately take away our abilities to say anything about whether The Shift should be used against Smith or Jones specifically, but it’ll tell us something about using The Shift against “hitters in general.”

We see that against The Shift, hitters do the obvious and cut down on their pull-happy ways. Clearly, they are still pull-plurality, but in front of The Shift, pulled balls take a bit of a dive. We only see about 93 percent of the pulled balls that we expect. It’s tempting to assume that since “opposite field” is the antonym of “pulled” then the missing pulled balls are being hit to the opposite field. This is not the case. In fact, we see a few missing opposite field balls. Instead, hitters facing The Shift are more often taking the ball to center field. That’s going to be a net positive for our hitters, even if it’s not the huge net positive that we were hoping for.

In the past, I’ve shown that one effect of The Shift is to increase the number of hits that the defense allows. Let’s take a look at what happens in terms of weighted runs created (not the plus, just wRC), using the same method. (The sample consists of only players who have 100 non-shifted balls in play in the season, and the “BIP” column here represents the number of shifted BIP that are in the sample.)

We can see that over time, the linear weight value that hitters who are facing The Shift generate is larger than what we would expect based on their performance when they aren’t being shifted. The effect size is about .0043 runs per BIP. On balls in play. Remember that.

3. Redemption for The Shift?

It’s important to remember that the only (public) data we have for The Shift is on balls that are “in play.” We don’t have numbers on walks, strikeouts, home runs, or hit batters. We can see that, as a mode of defensing balls in play, The Shift is actually a pretty bad way to go about generating outs and on those balls that stay “in play” shifting teams are actually losing on the deal.

How that all happens turns out to be important, though. Here’s isolated power (ISO) for hitters in front of The Shift, with their “expected” totals based on how they performed without The Shift. The numbers are going to look extremely low for what we normally expect from ISO (and in particular given the big slugger population that usually gets shifted) because home runs are not recorded in this sample.

Notice that for the first time, we actually see a situation where The Shift is doing a better job. ISO is (slightly) lower than we would expect when The Shift is on. Digging a tiny bit deeper, we see that The Shift somehow suppresses doubles and triples, although it compensates by adding a larger number of singles. BABIP goes up, and there’s enough value in those extra singles to offset the value of the “lost” doubles and triples (and then some), but it’s worth pointing out that doubles and triples are usually power outcomes.

What’s another power outcome that we can’t be entirely sure about in front of The Shift? Home runs. Do they also go down? (If someone would just release the data, I could answer the question directly.) Let me show you three numbers: 35.2 percent, 9.4 percent, 4.5 percent. In 2017, those were the HR/FB rates for pulled (fly) balls, center (fly) balls, and opposite field (fly) balls, respectively. Most power is pulled power. Remember how we said that hitters were pulling fewer balls in front of The Shift? That might earn them some additional base hits, but it’s also likely that they are settling for fewer homers. It’s not that they never pull the ball. It’s not that they never hit a center field homer. But there are (probably) effects at the margins.

We saw above that without The Shift, we would expect our weighted sample of shifted gentlemen to pull the ball 42.0 percent of the time. It’s probably a bit higher since we were only looking at “in play” balls and that means home runs were specifically excluded (and are more likely to be pulled balls). Along with that, 33.4 percent of balls “should have” gone to center field and 24.6 percent to the opposite field. Instead, we see the actual outcomes are 39.1 percent pull, 37.2 percent center, and 23.7 percent opposite field.

If we just go with those numbers, hitters lose about .007 home runs per fly ball, and if fly balls are about 40 percent of what these guys do (the league has about a 36 percent fly-ball rate, but these guys are usually shifted on specifically because they are power hitters) that means they’re losing about .0028 home runs. Your average homer is worth about two runs, putting the run value lost to home runs at about .0056. Above, we saw that the value lost via the increase in BABIP was .0043 runs per BIP.

The power outage (specifically, the home run loss) from The Shift exceeds the loss on balls in play. The Shift works, although not in the way that we think it does. It doesn’t make fielding the ball easier. It scares the batter from pulling the ball so much and saps his home run power as a result!

4. Not so fast. In fact, use your walking feet.

In the previous article, I found that one other thing The Shift does is increase the number of balls that we see in a plate appearance. The effect is small, adding one extra ball every 20 PA or so. Previously, I calculated the effects of The Shift in terms of these extra balls, which are more likely to put hitters into more advantageous counts and perhaps lead to better contact. Well, now we’ve accounted for that contact, but we still need to figure out the walks. Again, this would be easier if walk data in front of The Shift were available, but for now, we will make do with what we have.

We previously estimated that The Shift adds about 0.054 balls per PA to the count. In 2017, there were 260,184 non-intentional balls called. They broke down as follows:

That extra ball is the one that produces a walk 5.7 percent of the time. Some of those other extra balls may have turned the count from 1-1 to 2-1, and that is going to be an advantage to the hitter, but that’s going to be in the form of better results on balls hit into play (which we covered previously). If an “extra” ball happens 5.4 percent of the time, and 5.7 percent of those balls “cause” a walk, then The Shift “causes” an extra walk in an extra 0.3 percent of plate appearances. That increased chance of a walk is worth 0.021 runs per PA.

Let’s take stock of what we’ve estimated so far.

I’m not going to defend those values down to the third decimal place. Because I’m working in a surprisingly low information environment, I’ve had to make a lot of assumptions, and better data might render this all obsolete. But this is probably the right story and it’s an important story.

The Shift is still a net negative for the team that is doing the shifting, although not quite as negative as I had originally thought. The reasons why are really important, though. Insofar as The Shift “works,” it’s because it’s a psychological ploy. The evidence shows pretty conclusively that The Shift is actually an inferior defense to two-left-two-right, at least as “defense” has been traditionally defined as “balls in play.” It’s telling that no one thought it would be helpful to release the data on how many home runs are hit in front of The Shift. Apparently, no one figured that The Shift would have any effect on that. That’s a rather short-sighted point of view.

The Shift causes a behavioral change in the hitters that face it, and it turns out to be an irrational one. The hitters, faced with a tactic that appears to be a direct challenge to their pull-hitting ways, ever so subtly begin to pull the ball less. They don’t hit the ball to the opposite field. They start taking the ball up the middle. It’s only a few percentage points, but it means that they’re getting away from the one thing that got them shifted on to begin with. They are pull power hitters and they are moving away from their power, and in doing so they sacrifice some home runs (and doubles and triples).

They do replace them with a few more singles, but those aren’t as valuable as the lost extra-base hits. Hitters in front of The Shift are clearly trying to beat the challenge in front of them, rather than the entire system. They would actually do better to simply swing as they usually do and accept the fact that they will have a few balls that go for 5-3 putouts, despite the fact that the third baseman is playing in short right field. It’s a bad idea to go away from your power.

That means teams looking to shift should pay less attention to spray charts (OK, you have to do some of that), and instead find the hitters who have a lot of power to the pull field and who will change their behavior. That’s what really drives The Shift: it tricks hitters into giving up their power swing.

The findings concerning The Shift likely increasing the number of walks are still valid, and they turn The Shift from something that is actually a benefit for the defense to something that they should re-think, but I think there’s a bigger message here. Last time, when I talked about The Shift increasing walks, I suggested that teams might talk to their pitchers about how they were nibbling a little more and the problems that was causing. Maybe they could have a similar conversation about their shifted-on sluggers. “Don’t worry about The Shift. Pulling the ball is more efficient.”

The net total would be a few runs–perhaps half a win–returned to a team’s bottom line for having made the effort, but it’s something that could theoretically be done by just presenting the information to the affected people. But that’s the new challenge in a front office: figuring out why something works, and in some cases figuring out how to use that to shape players’ behavior. It turns out that The Shift is such an opportunity, although we rarely consider it that way.

But that’s the surprising story of The Shift. It’s a psychological ploy for which both sides actually fall, perhaps without knowing it. And that’s why it persists.