Over the past couple weeks, Russell Carleton has mounted a mini-attack on the widespread use of the shift in baseball, claiming that while some amount of shifting is surely a good idea, there is evidence that teams have taken it too far. It's a topic that matters to Rays fans, as the Rays are among the teams that shift the most.
But some of his conclusions (like, the ones in the headlines) are suspect. So now read on.
Here are the good and worthwhile questions Carleton proposes:
- Does the infield shift save runs on ground balls?
- Does the infield shift affect the outcome of at bats in other less obvious ways?
It's entirely possible that there are unintended consequences to moving defenders around the infield. As pitchers and catchers select pitches that take advantage of the shift they might make choices that create disadvantages in other ways, showing up as changes in home run, walk or strike out rates. Hitters changing their approaches to overcome the shift might find new sources of advantage, as described in this analysis of Mark Teixeira and Mike Moustakas' approaches.
In the first article, Carleton finds that when the shift is on, hitters hit more groundballs, as well as more groundballs the other way, than expected. The result is actually a slightly raised batting average on balls in play (BABIP) and a slightly lower slugging (SLG).
As he discusses in part two, his methodology was excluding some of the most-shifted players, and when they are included, the shift did in fact result in slightly lower BABIPs.
He doesn't yet have answers to most parts of his second question, so let's acknowledge that those are good points to think about, and move on without drawing conclusions.
In the second part, he looks specifically at the difference between individual hitters' production when they are shifted and when they are not, and finds that the shift has a very weak positive effect for the defensive team.
How Do You Know When It's Too Much?
So some of that data are interesting, but they fall way short of providing clear evidence to say: "the shift is overused."
The first issue is that teams don't actually care about about BABIP or about SLG. They care about runs. So to determine whether the shift is helping or hurting teams, Carleton needs to express the effect in run values. Anything else is potentially misleading.
The second issue is a question of game theory.
Every situation is different, and smart teams shift accordingly (or at least change their shifts constantly). Rays fans will see their fielders go to different spots on the field based on hitter, on pitcher, on who's on base, and on count.
One can assume that batters are changing their approach to each pitch based on many of the same variables. There is a give-and-take between the hitter's desire and ability to hit in his most-normal manner and his desire and ability to beat the shift.
So with full acknowledgement that "the shift" is not a binary decision and that categorizing plays as "shifted" or "not shifted" is very reductive, we can still approach the question the way we approach other game theory decisions such as whether to throw a fastball or a changeup.
In this type of situation, optimal play should show both decisions as contributing equal value.
If teams were shifting too much, then an analysis of the results of the shift should show that players hit better when shifted against. If teams were shifting too little, that analysis should show that they hit worse when shifted against.
But if the analysis finds, as Russell claims his does, that there is almost no perceptible difference between shifting and not shifting, that means that teams are shifting more or less at exactly the right amount.
Trying to answer some of the questions Carleton proposed in his first article, reader Brad McKay looked at the correlation between the frequency of each teams' shifts their pitchers' HR/FB, BB%, and K%. Carleton then integrated this work in the second article.
With apologies to McKay, I don't buy it.
The issue here is that there's something else that almost certainly has a much stronger affect on all of those pitching metrics than does a teams' shift percentage: quality of pitchers and quality of opponents.
For instance, sluggers who are great at hitting home runs also tend to pull the ball. A team that faces more of these types of players is likely to apply more shifts, and their pitchers are also more likely to give up more home runs.
What we really need to see is whether individual players hit more home runs when they're shifted against than when they aren't. Anything less than that? We should shrug our shoulders and move on.
The other thing to keep in mind is that not all high-shift teams are the same. They represent different managers and different pitching coaches applying different research by different researchers. To simply categorize teams as "high shift" and "low shift" is nearly as reductive as categorizing all shifts the same way.
My favorite article about the effectiveness of the shift was written last year by MGL.
In it he takes advantage of the fact that UZR throws out all shifted plays to compare success on shifts to success generally. His goal is to separate fielding ability from smart positioning, and what he finds is that some teams are better at shifting than others (the Rays are good); that there's not a strong correlation between how often teams shift and how efficient they are at their shifting; and that there's also not a strong year-to-year correlation in how good teams are at shifting.
There are a lot of questions that neither MGL nor Carleton have answered about whether the shift works. It's something worth thinking critically about and pursuing further. But there's almost no evidence to say that it doesn't work.
Defensive positioning is complicated, and its impact on hitters even harder to understand fully. No doubt smart teams will continue to evaluate when and how to shift and how to maximize offense in the face of shifts, but let's not grab onto "shift-less" arguments without fuller analysis.