Shift Happens. How the Infield Shift Can Affect Defensive Metrics
While shifting from typical defensive positions is often considered “new school” in baseball, the general concept is far from new. Defensive shifting was first documented in the 1920s, when it was employed against Cy Williams, but gained popularity against a more familiar Williams, Ted Williams, in the 1940s. Although defensive shifting has strong roots in baseball history, its use has been revolutionized by sabermetrics in the past decade and, accordingly, its use has seen explosive growth in that time. With increased shift usage in MLB, statisticians have spent the past few years attempting to determine the value of the shift, and the effect that it has on the outcome of at-bats. Although the data isn’t definitive, the general consensus seems to be that despite how difficult it is to quantify the numeric value of the shift, there is something there, and the steady increase in shift frequency from 2012-2017 in particular is clear evidence that teams see value in the shift.
One aspect of the defensive shift that has received very little attention, however, is the impact the shift has on individual defensive ratings. Statistics like Ultimate Zone Rating (UZR) and Defensive Runs Saved (DRS) attempt to quantify the quality of a player’s defensive prowess by using a system that rewards plays that are “less likely” to result in an out when fielded by a player in his position. By comparing a particular play to past data on similarly hit balls, it enables us to look at the likelihood of an outcome (hit/error/out), and see if the player performed above or below average for that play of interest. The Fielding Bible has a much more in-depth look at this here. As a simple example of how the shift can affect a player’s outcomes in a statistic like DRS, I’ve pulled a clip of Jose Altuve fielding a grounder up the middle in 2016. The video can be seen below:
In this first video, Altuve ranges to his right after his starting position (a relatively deep shift) put him in good position to make the play. This makes the play easier for him, although obviously not automatic. Since many of the average 2nd basemen on a ball similar to this one are playing a more “normal” location for 2nd basemen (much farther toward 1st, and away from the ball up the middle) the probability of an average 2nd basement making this play is close to 15-20% since they are much farther from the ball than Altuve in this video. Statistically, Altuve is rewarded for the improved starting position, and receives a boost in the DRS statistic for a successful putout on a low probability play. It is important to note that the actual difficulty of the play for the given player is not necessarily what is reflected in DRS. Many plays that are routine for a shifted fielder are some of the most valuable in terms of DRS. This disparity could result in inflated DRS numbers if the shift is always beneficial.
So, this anecdote provides insight into how the shift could manipulate statistics like DRS and UZR to be falsely inflated, but is it reflected in the data sets when we look at teams that shift a lot versus those that barely shift? As I mentioned before, the use of the shift in MLB has been on the rise over the past few seasons, and this increase in usage can easily be seen when I select a few teams, and examine defensive shift frequency over the last 6 seasons.
So, If we look at the infield DRS for each team, do we see any differences in runs saved depending on how frequently they shift? Unfortunately, we can’t separate plays made with versus those without the shift with publicly available data, but using the total DRS data, this is what we come up with for each infield position (excluding 1B) if we plot Shift Frequency vs. DRS from
THOSE ARE ALL FLAT LINES. If we saw a positive slope for our line of best fit in our data, we could reasonably conclude that there was an advantage for a position if that team shifted a lot. Unfortunately, it doesn’t look like there is a dramatic change in the DRS of any position whether a team shifts frequently or not. So what if we combine all of the infield positions together, and examine the 2012-2017 seasons?
Another FLAT LINE. Both of these graphs show that there is not a dramatic difference in infield DRS between teams that shift frequently and those that don’t. Basically, this data is trying to tell us that everything I fed you in that opening Altuve GIF is a giant hunk of baloney. So what is ACTUALLY going on here? I have a theory.
Allow me to present Altuve GIF #2:
In this example, Altuve starts the play on the 3rd base side of 2nd base, where Carlos Correa, the shortstop, normally sets up shop. Altuve makes a full sprint over to the right side of the infield to make the play. He probably gets very little statistical credit for this play since it is hit right at where a 2nd basemen would be standing in a standard defensive configuration. Basically, this play is marked as “routine” in the DRS formula, and any instance in which Altuve wouldn’t get a ball like this he would be penalized. The act of “beating the shift” like in the video below would be such a case where the fielder would take such a loss.
Ultimately, this investigation boils down to two things. First, there is likely a statistical benefit on some (if not a majority) of the plays in which an infielder shifts. Conversely, it is likely that many players take losses on their DRS numbers when the shift has them grossly out of position for what would normally be a routine ground ball. These actions seem to cancel out and result in very little difference when looking at DRS data across time. When measuring the shift, it seems every action has an equal and opposite reaction. Newton’s laws strike again.