Defensive shifts are on the rise in baseball, and data driven teams are at the forefront of the increasing use of the shift. Since the Astros are oriented toward sabermetrics, we shouldn't be surprised that the Astros exhibited the most significant increase in the use of shifts and were among the league leaders in using defensive shifts.
Jeff Zimmerman's recent Hardball Times article provides data on BABIP (batting average on balls in play) results for defensive shifts in 2013. I have analyzed some of the summary data he presents. (Note that some of the data I use is slightly different from the tables in his article, because he subsequently corrected errors in the underlying data spreadsheets.)
The team data below is based on shifts for balls in play (i.e., it doesn't include all instances of putting on the shift---only when the ball was put in play). The table shows the top ten ML teams in putting on the shift when a ball was put in play. BABIP-shift and BABIP-no shift reflect the BABIP when the shift was on and when it wasn't. Difference reflects the BABIP points difference between the team's overall BABIP and BABIP-shift.
The Astros had the fifth highest number of major defensive shifts. The Astros may have an above average overall pitching BABIP, but the shift appears to be producing a beneficial effect, slicing five points of BABIP compared to situations when the shift is not used. Only two of the teams, above, (Yankees and Cubs) produced worse results with the shift than without it. I would note that Zimmerman's data base shows a higher pitching BABIP for the Astros than either fangraphs or baseball-reference; it's unclear whether this is due to a different calculation of BABIP or the sample of batters used by Zimmerman. This discrepancy is not limited to the Astros.
In the image, below, Jose Altuve, from his shift position in short RF, catches a "fliner" hit by Raul Ibanez.
Because the data is not very detailed, my analysis is somewhat limited. But an examination of the relationships in the data for all 30 teams appears to confirm that the shift practice suppresses BABIP.
- On average, shifts produced a BABIP 1.5 points below the overall team BABIP. This decrease in BABIP has a 0.29 correlation with overall BABIP.
- The number of shifts used by a team has an inverse correlation (r= -.49) with team BABIP, suggesting that an increased used of shifts is associated with lower overall BABIP. The R-squared indicates that the number of shifts explains 22% of the variation in team BABIP. This is a surprisingly high percentage, considering that, on average, shifts are applied to only 5% of balls in play. Perhaps shifts are associated with other defensive or pitching relationships.
- BABIP-shift has a 0.55 correlation with overall BABIP. This isn't all that surprising since BABIP-shift is a component of overall BABIP.
- I wondered whether the lower BABIP-shift could be due to a selection bias caused because players subjected to shifts are likely to be low BABIP hitters. However, this doesn't seem to be the case. For all hitters who put a ball in play at least 90 times against the shift, the average BABIP--including the shift--is above average (.298 vs..294 MLB average). Without the shift, these hitters have a .320 BABIP. To the extent that a selection bias exists, it suggests that the reductions in team BABIP caused by the shift is understated.
"One thing I've learned in Tampa that they've reiterated here is that you want to catch line drives. We don't want guys to be kind of where they hit it," he explained. "We want to be exactly where they hit it the hardest, because then if they deviate from there, it's a little weaker contact. Weaker contact does what? It gives the infielder more time to get to those balls."It's all probability. It's all probability. If this guy hits the ball on the screws, what's the probability that you'll get it? You've got to be right there pretty much, or within arm's length. If this guy chops the ball then the probability of you getting the ball there is great."