clock menu more-arrow no yes

Filed under:

You've gotta be shifting me!

New, comments

With the release of publicly available shift data comes another opportunity to see just how much it has (or has not) suppressed offense.

Jonathan Dyer-USA TODAY Sports

As baseball teams have continued to embrace unconventional opportunities for improvement, one of the supposed advantages that has swept through the game like the Great Awakening has been the infield shift. While its origins are up for debate, one thing we can all agree on is that the shift isn't new; rather, it was brought to the forefront of our collective consciousness by spectacled sabermetric guru Joe Maddon and his merry band of versatile defenders when he was helping the Rays squeeze the extra two percent out of their limited budget.

Back when they were still the Devil Rays, Maddon's pitchers allowed BABIP values of .312 and .331 in his first two seasons as manager in Tampa Bay, both above league average. The Rays shook the trend, along with the Devil, in 2008. They went on to put up below-league average BABIP values for the next seven season's of Maddon's reign, as he implemented the shift more and more frequently.

In 2010, the first year for which shifting data are now available — thanks, FanGraphs! — the Rays shifted most often, either by way of a traditional (i.e., Ted Williams) shift or some other unconventional alignment of infielders. While this could be attributed to playing in the same division as David Ortiz, the Rays shifted more than fifty times more often than the next closest team. They also had the second-lowest BABIP that season (.278). The Brewers shifted the least that season, and also put up the second-highest BABIP in the league. But there may be a confound present in these two cases. Were teams that cared more about defense more likely to subscribe to the ideology of the shift?

Perhaps because of Maddon's reputation, or perhaps due to some subjectivity in how the impact of the shift is calculated, other teams quickly adopted the strategy. Consider that in 2010, only 3,323 infield shifts were on when a ball was put into play. While that may sound like a lot, it's just a drop in the balls in play bucket, making up just three percent of all these events. Shifting declined slightly in 2011 before more than doubling the subsequent season. It continued to shoot up until last season, where it was employed on nearly one out of every four balls in play leaguewide. The craziest part is no team has been left behind in this movement. Even the Nationals, who shifted four times less often than the league leading Astros last season, put the shift on for 159 percent more often than the Rays did in 2010.

With this public release of shifting data, we don't need to correlate overall BABIP with the frequency of the shift, because we can actually look at the difference in BABIP when teams are putting on the shift and when they're not. While these data have come out for single seasons in the past, we can now look for trends under these different conditions as the shift has increased in use.

In the first four seasons of data, when shifts were less frequent, BABIP on balls hit into the shift alternates between being higher and lower than BABIP when the shift wasn't employed. Although it may be tempting to apply a narrative to this pattern, the variation is likely based on the low number of shifts being employed and the volatility of the statistic. What is interesting is that as the frequency of shifting increases, the two BABIP values converge around .300. Even with the record high number of shifts in 2015, this evidence suggests that it hasn't done anything to reduce the number of hits overall.

But leaguewide BABIP probably isn't the best way best way to measure the value of the shift, because the signal that we're looking for, the conversion of ground balls hit into the shift, is diluted by the noise of all the other batted balls. These aren't just the fly balls that don't leave the park, or the line drives, but also the grounders that don't obey the typical pattern of pull-happy hitters. Ideally, you would look at BABIP for ground balls only, but we, the great unwashed masses, don't currently have access to that level of detail, so for now, the broad strokes will have to do.

It's been suggested that a better way to look at the effect of shifting with the data that we do have is to compare differences in shift and no-shift BABIP for individual players. So let's look at players with at least 500 balls in play between 2010 and 2015 who had at least one at bat under each of the two conditions, and compare their BABIP when the shift is on and when it's off.

Not surprisingly, players who received very few shifts showed a wide range of differences. While there is a slight trend to lower BABIP values against the shift for players who were shifted against most often, as you might expect for those batted ball profile suggests the shift would be most effective, the relationship is about as weak as you can get.

Isolate players who received the shift more than a quarter of the time, and the strength of the relationship is three times as large. However, the effect is roughly half as pronounced.

While the relationship between the amount a player is shifted and the decrease in BABIP against the shift is small, it's important to remember that we're looking at all batted balls, not just ground balls. We're also not looking at a very large sample size for many of the players. For Ryan Howard, David Ortiz, and Adam Dunn — the three at the right hand of the spectrum — we're talking about players who weren't shifted against less than 800 times, between the three of them! Let's look a little deeper at these last two standing with shift rates above 80 percent to see how the shift has worked against them.

We'll begin with the player with the least difference between their shift and non-shift BABIP between 2010 and 2015, David Ortiz. Big Papi has found himself at the centre of this topic on several occasions, both a case for, and against, the value of the shift. This is partly because his no-shift BABIP alternates in a similar patter to the one seen in the first figure. But, multiply the net difference in BABIP over these six seasons (0.008) by the number of times he was shifted against (not including sacrifice hits), and it seems that all the shifting against Ortiz saved about 15 hits. Shifting against Adam Dunn, on the other hand, cost teams about 40 hits. Remember, though, that the error bars on BABIP, especially no-shift BABIP, are huge. Saving 15 hits over six seasons is certainly preferable to giving them up, but it's still not convincing evidence to support the effectiveness of the shift.

Year %Shift Shift BABIP No-Shift BABIP
2010 84% 0.329 0.232
2011 77% 0.325 0.309
2012 77% 0.316 0.316
2013 78% 0.306 0.375
2014 93% 0.256 0.259
2015 97% 0.268 0.154
Total 85% 0.296 0.304

But look to the far right of the second-to-last figure and you'll see a lone data point pounding on the table in clear support of the shift under ideal conditions. The player most shifted against was Ryan Howard, and boy did it work, saving teams an estimated 183 hits! The same sample size issues apply to Howard as they did to Ortiz, but it's hard not to look at the consistency of the pattern and the magnitude of the effect and not be convinced that shifting might be the only reason that Ryan Howard isn't living up to his contract. (Just kidding, I know about the thing with the lefties.)

Year %S Shift BABIP No-Shift BABIP
2010 84% 0.313 0.431
2011 85% 0.281 0.418
2012 73% 0.266 0.341
2013 89% 0.327 0.524
2014 92% 0.281 0.367
2015 96% 0.266 0.417
Total 87% 0.289 0.410

Finally, using this approach, let's look at the difference in BABIP for all batters who have been shifted against, and convert it to hits based on the number of eligible balls in play for each player. The effect of the shift, overall, seems to be a wash, resulting in just a net 33 hits being saved between 2010 and 2015. This suggests something that has been, well, suggested before: The infield shift may be used too often, and the positive effect seen against players like Howard, as well as players like Chris Davis (56 hits saved) and Brian McCann (45) may be sacrificed in situations where a player's batted-ball profile is less defined.

Since 2010, we've seen the use of the infield shift proliferate through the league to the point that there wasn't a team that didn't use it regularly last season. What isn't clear is whether the infield shift is actually suppressing offense. There are limitations to the data that's currently publicly available, but it could also be that teams have gone shift happy and are returning some of the hits they've saved against the most predictable players by using it against those whose batted ball tendencies aren't as prominent. Teams, it would appear, do not agree. The infield shift has been on for 30 percent of balls in play already this young season. There may be a good reason for the continued rise of the shift; we just don't have the evidence to support it, yet.

. . .

Matt Jackson is a featured writer for Beyond the Box Score and a staff writer for Royals Review. You can follow him on Twitter at @jacksontaigu.