Best of BP 2018: Walking Away From The Shift

Last week in this space, I discussed the infield shift and how it has a major flaw that may actually make it a net negative for the defense it’s supposed to be helping. The reason is that even though The Shift helps to prevent singles from getting through the infield, among other things, there are a whole bunch of walks that seem to happen in front of The Shift, more than we would otherwise expect. Last week, all we saw was that the walks happened. Today, we’ll ask the question of why they happen.

One of two things is going on. Teams are either aware of all the additional walks that happen in front of The Shift but believe there is some other countervailing force that makes the additional walks worth it, or they … just never asked the question. Today, we’ll be looking at potential reasons why the walks might be happening and whether it’s plausible that the extra walks might just be a silly mistake.

There are a few theories that we can fairly easily discount. First, the effect is not because teams shift more commonly in front of hitters who are more likely to walk. That is a true statement, but the methodology that I used specifically controlled for that. (I used a baseline of each individual player’s stat line when not shifted and looked to see how he performed when shifted.) The problem is that we had to sum across the whole league to get any usable findings, and in doing so we can only make conclusions about The Shift in general, rather than individual players. But still, something’s going wrong.

The other idea is that teams are aware The Shift increases walks, they are OK with that, because it has some other effect that they like more. This would be perfectly reasonable if it were true. But it’s not. Using data from 2015-2017 (as I will be throughout this article), with players who had at least 100 non-shifted plate appearances within a season, here’s what we would have expected from them if that same group of players (weighted for how often they saw The Shift) had performed exactly as well in front of The Shift (and here, I’m talking about a full three-infielders-on-one-side version) as they did against a standard defensive (two on the left/two on the right) formation, how they actually performed, and the differences between those numbers.

Outcome (n = 49,052)	Expected Number	Actual Number	Difference	Percentage Point Difference
Strikeouts	10,478	10,279	-199	-0.4%
Walks	4,310	4,884	+574	+1.2%
HBP	484	510	+26	+0.0%
Single	7,103	6,610	-493	-1.0%
Double/Triple	2,463	2,556	+93	+0.2%
Homerun	2,019	2,086	+67	+0.1%
Out in Play	22,644	22,577	-67	-0.1%

We see that while The Shift is doing a good job of taking away singles from hitters, it is “giving back” an even larger number of walks. Hitters are also coming away with a few more extra-base hits, including home runs, and are being hit by more pitches. They aren’t striking out as much and not hitting into more outs in the field (per PA). Those aren’t huge effects, but even if you want to hand wave them away, they are literally all pointing in the wrong direction other than the singles. If pitchers are trying to strategically pitch more carefully in front of The Shift, allowing more walks but picking up value elsewhere, it isn’t working.

So if it isn’t a selective sampling problem and it’s not part of a grand strategy, where are the walks coming from?

Warning! Gory Mathematical Details Ahead!

Hypothesis 1: It’s an accounting error.

The numbers that I’m quoting above are based on what happens at the end of a plate appearance, and all plate appearances are classified as “shift” or “no shift” based on the defensive configuration on the final pitch. Of course, teams don’t have to maintain the same formation throughout a plate appearance. Fielders may move where they want to when they want to. Is it possible that the spike in walks is because teams are more likely to switch into a full shift at points in a plate appearance where a walk is more likely?

Among plate appearances whose last pitch included a full shift (three infielders to one side of second base), 13 percent of them had at least one pitch in which the defense was identified as being in a different formation (and some of those were situations where one of the infielders were shaded up the middle, but hadn’t crossed over second base, something which the MLBAM data set would consider a “strategic” rather than a “full” shift). Perhaps teams were lining up in a standard formation until the count has two or three balls, and then for the fourth ball, they switched over to a full shift? (In that case, The Shift deserves only a small portion of the blame for that walk.)

The nice thing about the data recently released by MLBAM on The Shift is that we can look at the strategy on a pitch-by-pitch level. We can see when teams switched formations. Looking at pitches that happened in front of a full shift, 23.5 percent were with no balls in the count (i.e., 0-0, 0-1, or 0-2). However, for the pitches immediately following when teams moved into a full shift, 32.4 percent were with no balls (and 35.9 percent were with one ball). Teams tend to switch into a full shift when it’s actually least likely that a walk is going to happen. (Teams also switched into a standard defense earlier in the count as well.)

On top of that, when teams moved into a full shift, the next pitch was actually less likely to be a ball, and more likely to be a ball that the batter swung at, made contact with, and hit into play. So when teams switch into a full shift, the next things that happen all make it less likely that they would produce a walk. In general, most formation switches happen early in the count, so while perhaps some of the credit for a few walks shouldn’t be laid at the feet of The Shift, that can’t nearly explain all of the effect.

***

Hypothesis 2: Shifts happen in situations where teams would prefer a walk to a single

Assume for a moment that your team must endure the other team getting either a walk or a single. Which would you choose? While neither is a great outcome, the walk is probably the better option. With the bases empty, both are functionally the same, with the batter ending up on first base (although a single might occasionally turn into a single and an error, which gives it a slight edge.) But with runners on base, the amount by which a team prefers the walk to a single changes a lot. A single definitely scores a runner on third base, and might score a runner from second base. A walk only scores the runner from third if the bases are loaded and will not score a runner from second. When runners are on base, a team would have even more incentive to take away singles, even if it meant giving back some walks.

This is, unfortunately for the hypothesis, the opposite of what we actually see. Most (71.4 percent) full shifts happen with no one on, compared to 57.4 percent of standard defensive formations. (Note: This is not a surprise. It’s hard to pull off a shift when you have a runner at third—with no one on the line, you’re basically conceding a huge lead to him—or at second—where if you shift too far, he might just outrun your one guy on the left side of the infield and take third.)

To put it another way, in situations where there was no one on, teams lined up in a full shift 15.0 percent of the time, compared to 8.2 percent when there were runners on. Teams might very well pitch to the walk when runners are on, but they are more likely to be in a standard formation when they do so. That’s going to put more walks into the non-shift column.

***

Hypothesis 3: The boringly obvious explanation.

Let’s instead turn to the pitchers. My initial hypothesis was that the walks were simply a matter of a behavioral difference caused by The Shift. In other words, The Shift freaks the pitchers out. We can look into that. To start with, pitchers actually start to shy away from their fastball a bit.

Similar to last week, we can look at how often a pitcher throws his fastball without The Shift, and use that as a baseline for expectations to look at how he behaves in front of The Shift. For example, if he throws fastballs 50 percent of the time without The Shift, we would expect 50 fastballs per 100 pitches in front of The Shift. They might throw more or less on an individual basis, but we can sum across the league to find out what “pitchers in general” do. And it turns out that they use their fastball a little less than expected. We can even confirm that by looking at the batters and seeing how often they are usually fed fastballs with no shifting compared to how often they see fastballs when The Shift is on.

Outcome	Expected Number	Actual Number	Difference	Percentage Point Difference
Fastballs (pitcher’s perspective; n = 212,810 pitches)	108,488	105,886	-2,562	-1.2%
Fastballs (batter’s perspective; n = 181,951 pitches)	90,637	90,365	-272	-0.2%

If pitchers are throwing (a little) more off-speed stuff, perhaps they are also changing their approach to pitching location. For example, leaving a pitch outside is an invitation for a hitter to slap the ball the other way. It may not be something he normally does (that’s why he’s being shifted to begin with), but he’s a professional hitter and he might be tempted to try it. And maybe it might work. That would be embarrassing!

I classified each pitch as being thrown on either the inner third of the plate (or inside), over the middle third of the plate, or on the outer third of the plate (or outside).

Location (pitcher’s perspective)	Expected Number	Actual Number	Difference	Percentage Point Difference
Inside Third/Inside (n = 139,978 pitches)	44,223	44,921	+698	+0.5%
Outer Third/Outside (n = 139,978 pitches)	70,310	71,177	+867	+0.6%

So, if there’s something that pitchers are doing, it’s not favoring one side of the zone or another, but staying off the middle of the plate. In fact, this teases what’s actually going on.

Result (all pitcher’s perspective)	Expected Number	Actual Number	Difference	Percentage Point Difference
Ball Outside GameDay Strike Zone (n = 239,233 pitches)	122,412	127,154	+4,742	+2.0%
Called Ball (per all pitches; n = 213,059 pitches)	75,788	79,495	+3,707	+1.7%
Called Ball (per taken pitches; n = 115,998)	78,043	80,457	+2,414	+2.1%

If you don’t believe the GameDay strike zone, then believe the umpires (because the rule book says you have to). What’s really happening is that pitchers are staying out of the strike zone more than they usually do. They don’t throw everything out of the zone. It’s a small effect at the margins, but if there’s something that catcher framing taught us, it’s that changing a few balls into strikes (or strikes into balls) is worth much more than we could possibly have guessed.

How do the batters respond?

Result (all batter’s perspective)	Expected Number	Actual Number	Difference	Percentage Point Difference
Swing (n = 182,083 pitches)	83,802	84,242	+440	+0.2%
Contact (per swing; n = 80,899 pitches)	62,364	63,026	+662	+0.8%
Foul Balls (per contact; n = 53,945 pitches)	27,194	27,444	+250	+0.5%

They swing a tiny bit more, and spoil a few more pitches, but if there are about two percent more pitches out of the strike zone and swing rate only goes up by a bit, pitchers aren’t getting more chases. They’re just going to ring up more balls. Do I need to connect the dots on how that would lead to more walks?

Once we get down to this behavioral level, it’s pretty obvious why the extra walks happen. Pitchers don’t throw as often in the strike zone. It’s not that they’re working too far inside or too far outside. They’re just not filling up the zone, and if you do that enough, eventually someone gets to walk to first. Even if the extra ball doesn’t produce an extra walk, it does tilt the count a little more in favor of the hitter. And while The Shift does help in gobbling up would-be ground-ball singles, it seems to be spitting out a few more doubles, triples, and homers. Maybe someone felt a little more comfortable at 2-1 than he would have at 1-2, and squared one up.

What’s the Problem Here?

Ten years ago, the fight over The Shift was for buy-in from the players on the field. With no one guarding the third base line, there were going to be grounders that went through to the outfield when they could have been turned into easy outs. Even if the math said that there would be more ground balls turned from hits into outs, it still took a while before players felt emotionally OK enough with The Shift that they stopped objecting. Or that they said they felt emotionally OK about it. Eventually, when your boss tells you that you need to be OK with something, you find a way to make peace with it. Sorta.

Despite the fact that the use of The Shift has increased by a factor of 10 in less than a decade, it’s still the “other” in the room. It isn’t rational to think this way, but humans over-value something being taken away from them (the hit that could have been an out, if only they had just played the “normal” way) versus something being given to them (the out that would have been a single through the hole). And pitchers are humans. The incentives are there for them to do something to avoid that icky feeling that comes with a hit that should have been an out. Maybe that’s staying out of the zone a little more than normal.

When I saw this data, the clinical psychologist in me took over. I’m sure all of these pitchers would say that they are fully on board with The Shift. The problem is that people say all sorts of things they don’t mean and their behavior tells a different story. I’m sure teams were hoping for (and believed that they had gotten) an enthusiastic “YES!” from their pitchers, when in reality it seems that they’d gotten a grudging “Ugh … yeah, sure, fine.” That’s not the same thing, and that unstated discomfort seems to be coming out in a way that’s destructive to the ultimate goal. When people aren’t fully committed to something, sometimes they sabotage it in what ways they aren’t even fully aware they are doing.

There are going to be a lot of people who read the findings I’ve presented above and use it as a bludgeon to say, “See! I told you that The Shift should be vanquished!” I don’t think that’s the right takeaway. The Shift, if deployed properly, seems to be a way to improve defensive efficiency. And the stumbling block here is a psychological one. I’m not entirely sure how easy it would be to overcome in a way that would make the math work, but in theory it could be overcome. And at that point, teams should go crazy with shifting.

Instead, I take this as a lesson in how hard it is to really get buy-in into anything, not just in getting to “yes” but getting to “YES!” I think that’s the major project that still faces The Shift. The positioning of fielders is just a matter of looking at some spray charts. Pitchers overcoming that icky feeling? Now that’s a little tougher, but worth doing.