Pebble Hunting: Clayto... (05/03)
10 comments have been left for this article. BP Comment Quick Links
| |
|
|
May 3, 2016 Baseball TherapyThe Pretty Good Case That the Shift Doesn't WorkHere’s a cheeky question that I ask in complete sincerity: How many home runs were hit against The Shift last year? I’m sure someone out there knows the answer to the question, but there are probably more people wondering why I even bothered to ask it. If the ball was hit over the wall, what does it matter whether The Shift was on or not? Either way, the fielders weren’t going to be able to get to it. It’s an important question. The infield shift was once the new, hot “it” baseball idea. Sending three infielders to the right side of the infield was seen as a sign of enlightenment, rather than as a sign of a shortstop with a poor internal GPS. Teams were finally able to break free of the invisible force field that said “two over here, two over there.” And for some hitters, that made a lot of sense. If a batter hit 80 percent of his ground balls to the right side, why not put three fielders over there? And so it was done. Now, shifting is so commonplace that Joe Girardi seems a little out of place when he suggests that The Shift be banned. Usage of The Shift has increased in the past few years. Even leaving aside the broader move toward more precise tailoring of fielder position to the tendencies of the particular batter in the box, the practice of putting three fielders to the right (or left) of second base has increased. Here are the number of three-on-one-side shifts by season for the past few years, according to data from Baseball Info Solutions.
Now that The Shift is just a regular part of the game, it bears asking whether it actually does what it says on the label. Might seem a strange question to ask. There have been several studies that have tried to answer the question, and generally come to the conclusion that yeah it works, even if it’s only a small effect. For the most part, these studies have relied on looking at the aggregate number of hits on balls in play when The Shift was on and when The Shift wasn’t. It’s not bad methodology, but I think it leaves out a few important questions. So today, we re-visit the question of whether The Shift actually works. Warning! Gory Mathematical Details Ahead! Our friends at FanGraphs now provide splits for individual players based on whether there was a shift on or not. So, I gathered all player-seasons from 2012 to 2015 and did a quick check of what happened when The Shift was on in those years. In theory, The Shift exists to reduce the chances that a hitter will get a hit if he puts the ball into play. It’s really only going to be effective on groundballs, and sadly, I can’t filter by both groundballs and shifts. Still if there’s an effect on groundballs, it should make itself fairly apparent. I started with players who logged at least 100 PA where there was no shift and used their BABIP against a non-shifted infield as my baseline. In theory, when a team shifts, it might not prevent all basehits getting through, but The Shift should be a better shield against basehits than the no-shift condition. At least that’s what it says on the label. Here’s the thing to remember. While BABIP is (kinda) random for pitchers, we have evidence that it is much more a skill that hitters have. So, it’s important to control for who actually gets shifted and how much and how good they are in general at getting hits on balls in play. So, I took the batter’s non-shift BABIP and the number of plate appearances which included a shift (and a ball in play) and multiplied them. This tells us, given the number of plate appearances that the batter had with The Shift, how many hits would we expect him to have against The Shift, if The Shift made no difference. Of course, we might be talking tiny sample sizes on the individual player level, so I summed across the entire league, both the expected hits and the actual hits that took place against The Shift. I also had SLG against The Shift (and not against The Shift). Since these are all balls in play, it allowed me to get an expected (and actual) count of total bases against The Shift.
Wait a minute. That means that… The Shift doesn’t actually work. Actually, it only works in odd-numbered years, according to these data. Well, now I guess we gotta figure out why. Time to squirm a little bit. As I mentioned above, the hits that we’re talking about could have come off of grounders, but also flyballs and line drives which will be minimally affected by The Shift. It’s entirely possible that The Shift really is doing its job on grounders in that when a batter hits a grounder, he hits it into the waiting glove of a fielder who is positioned in a weird place more often than he hits it into a hole where a “normally” positioned fielder would have otherwise been standing. There’s a hidden assumption buried within how we usually think about The Shift that we need to talk about. It’s the idea that when the batter walks up to the plate and sees The Shift on (it’s not easy to hide) he just shrugs and… doesn’t make any sort of adjustment for that. Turns out that this isn’t the case. Similar to my analyses above, I looked for batters who had 100 non-shifted, ball in play PA and found their groundball rate and pull rate in these non-shift situations and used that as a basis for an estimate of how many GBs and pulled balls we would expect if they made no adjustments at all when faced with The Shift. Expectations are again weighted by how many shifted PA they faced. Again, totals are summed across the league. It’s worth noting that I don’t have their pull rates on groundballs, just their pull rate in general and their grounder rate in general, but the data are interesting nonetheless.
It’s worth saying over and over again that these data represent only balls that end up in play, so this increase in groundballs is stealing directly from flyballs and line drives. But far from the usual narrative about The Shift being a way to more effectively defend certain hitters, we’re not seeing a decrease in BABIP (in fact, we see a slight increase). Groundballs, in general, have a greater chance of ending up as hits compared to flyballs. While The Shift may be a somewhat more efficient way of dealing with the groundballs that do get hit, The Shift has the side effect of putting more groundballs into the system, and groundballs are not as efficient as flyballs as a way to generate outs on balls in play. Sometimes the best way to defend against a groundball is to have the batter fly out to the right fielder. Now, we return to my original question. How many home runs were hit against The Shift last year? My original inspiration for this piece was a column written by Fox Sports reporter Ken Rosenthal on the Red Sox and their shifting tendencies this year. From the article: Red Sox pitching coach Carl Willis, bullpen coach Dana LeVangie and infield coach Brian Butterfield plan to begin asking certain pitchers to eliminate a certain pitch or location to take better advantage of shifts, Butterfield said. The way Butterfield explained it, some right-handers occasionally like to elevate their fastballs against left-handed hitters; both Jake Peavy and John Lackey did that when they were with the Red Sox, and Rick Porcello does it now. Early in the count, the pitcher might want the left-handed hitter to chase; late in the count, the elevated fastball might be a way for the pitcher to finish the hitter. The problem, Butterfield said, is that when the ball is up, the chances increase that the left-handed hitter will go the other way. Eliminate the location and the hitter becomes more likely to pull, giving The Shift a better chance of success. We don’t know what the other 29 teams are telling their pitchers about how to pitch in front of The Shift, but I’m guessing that others have asked their pitchers to pitch in ways that “take advantage of The Shift.” Perhaps a hitting coach or two has picked up on the idea as well, and in the pre-game briefing, told their hitters to not think much about pitch X or location Y when up at bat. They’re not going to throw there. If nothing else, even if a batter is up at the plate guessing, it’s one option that he can eliminate from his range of guesses. That’s bound to make him a better hitter. The fact that he can plainly see that The Shift is on means that he might be up there laying off of pitches that he might have swung at in the past because he feels he’d just hit it into that $%&@ing shift. We also know that The Shift tends to annoy pitchers, probably because they feel that they have to change the way that they pitch and don’t want to. It’s hard to know what the effect of that annoyance is, but I doubt it would be positive, and perhaps a little extra frustration occasionally leads to a careless meatball every now and then. But we don’t know the answer to that. Is the reason that the proportion of groundballs that remain in play goes up with The Shift a result of the fact that while the same number of groundballs and flyballs are being hit overall, a greater proportion of the flyballs that are hit are leaving the sample… by leaving the yard? I suppose that there is room for the idea that The Shift somehow leads to more strikeouts and fewer walks than we might expect when The Shift is not on. Maybe home run rates don’t go actually go up (or down). But the facts on the ground tell us that at least the way The Shift is being implemented now, it is—at best—a net neutral strategy for the balls in play. If the other outcomes are being affected in a negative way, then teams are actually shooting themselves in the foot by employing The Shift, even as a cursory look shows that The Shift is working just dandy. Is Joe Girardi Right? This is a lesson in not looking for the unintended consequences of a strategy. A few months ago, I suggested that everyone’s favorite new toy, catcher framing, might be shiny to look at in isolation but might be more muted when you really take a closer look. I think the infield shift is similar. There’s information that we just don’t know that would help to answer this question, but there is a decent chance that despite the fact that the infield shift looks like it’s working, it might actually be doing more harm than good… And now a phrase that I want to give its own paragraph: in the aggregate and in the way in which it is currently practiced. My argument here is that if teams had not shifted at all between 2012 and 2015, it’s likely that they would have been not much better or worse off. That’s not to say that it all shifts are meaningless. There’s space in between those extremes, and it might be time to explore them more thoroughly. I specifically want to emphasize that I don’t think that The Shift should completely disappear. There are going to be cases in which it makes complete sense. In fact, Matt Jackson found that players who were shifted the most often found the biggest shift-versus-no-shift differences in BABIP. In those cases, teams really should line up three guys on the right (or left) side of the infield. There are a couple of reasons that I can think of for why we’re seeing these numbers. Maybe all of them are true, but it’s worth asking whether any of these are in play. 1. Teams have fallen overly in love with The Shift and are shifting on players for whom The Shift doesn’t end up making sense. It seems sensible that some players really are going to hit the ball to the right (or left) no matter what you do. But teams have undersold the fact that some hitters are able to make adjustments and have gotten burned by shifting against those guys. 2. The Shift has become an idol unto itself and there’s social pressure—whether explicit or implicit—on pitchers to generate a groundball “to take advantage of The Shift” rather than “get the guy out.” This introduces enough problems into the system that it negates whatever positive effects The Shift would otherwise bring. Sure, The Shift helps when the hitter hits a groundball, but the point isn’t to generate outcomes that make The Shift look good, but to generate outcomes that make the batter walk back to the dugout with a sad face on. 3. My own analysis could be flawed. The way that the data are available, we have a simple “yes/no” answer as to whether a shift was on. There’s a bit more nuance to The Shift. What to call it if the shortstop is playing very much up the middle, but not to the right of second? 4. The fact that we’re talking about the league aggregate could be masking the fact that some teams are just better at shifting than others. (If I may for a moment break the rarely-even-acknowledged fourth wall around here, team people, there’s a Lake Wobegon trap hiding in there. Don’t all 30 of you go assuming that you are above average.) Using the data that are publicly available, I can’t panel the data by opposing team, so I can’t rule that out. The solutions to each of these potential problems differ, but they all have a similar bedrock: the need for a much deeper understanding of everything that happens when The Shift is on. What’s fascinating about these data is that if there is a benefit to The Shift, it’s not the one that we thought it would be. The Shift seems to actually produce more hits than we would have otherwise expected, though fewer total bases. The actual effect of The Shift might not be in more groundballs being turned into outs, but that the presence of The Shift gets everyone so freaked out about needing to generate a groundball to justify the fact that The Shift and they start trying to get groundballs rather than outs. More groundballs mean more singles, although likely a reduced chance of a double or triple. And that might have some value. You trade off a few singles for a reduction in extra-base hits. Some. The net effect is pretty minimal. But if that’s what’s really going on, then it’s possible that the freak-out factor is affecting the rest of the at-bat. If pitchers are pitching differently, is “differently” also “less efficiently?” Is it possible that The Shift could have defended against that home run by not being put on in the first place? What we’ve learned suggests that The Shift might not necessarily be the sign of a “smart team” but the sign of one that hasn’t fully thought this all the way through. If more home runs (or fewer strikeouts or more walks) are being hit against The Shift, then that’s a major problem. But we don’t know the answer to that question yet, and even though it seems like a non-sense question, it might make the entire difference on whether The Shift is actually an innovation or just a shiny gadget.
Russell A. Carleton is an author of Baseball Prospectus. Follow @pizzacutter4
|
Closer Report: Week Five
Good stuff, Russell. Thought-provoking, as always.
One thing that I think is important that isn't given much play in the article is that we should have fairly strong priors that the shift is a net-positive strategy as it is currently implemented, even if we can't detect it in publicly available data.
Early adopters of the shift seem to increase usage over time. This suggests that whatever evaluation methods teams have in place, they must be giving positive feedback. I don't think the Pirates would implement their shifting strategy, lack empirical evidence of it's efficacy, and ratchet it up anyway. This is especially true due to cultural challenges of implementing widespread shifting; I would think the burden of proof is higher when a strategy is against the grain and very public. They must at least be convincing their players and themselves. Rival teams observed the strategy, presumably evaluated it somehow, and decided to follow suit. Is it possible that this is all group-think and confirmation bias? Perhaps, but that seems like an odd way for such a strategy to gain hold so fast.
One thing that's mentioned in the article as if it's an anti-shift point which could in fact be a pro-shift point is that hitters appear to change their approach in response to shifts. They hit more grounders and they pull fewer balls. You talk about this as if its a strategic response by hitters that mitigates the shift's advantages; I think the shift may be employed specifically to cause hitters to change their approach in this way. If David Ortiz (or whoever) decides buck the shift by smacking a few extra grounders to the left side, he is almost certainly going to hit for less power. I would be shocked if the higher grounder rates and lower pull rates are a result of extra home runs to the pull-side, as you suggest. It seems far more likely that hitters are pulling the ball less and hitting it in the air less, leading to similar (or even higher) BABIPs and less power on contact, and that this is a trade-off teams are happy to make. The public data to evaluate this isn't great, but I'd be willing to bet the private data is much better.
It's a frustrating walk, because I don't have the super-secret data to look at. (If I did, I'd probably be forbidden from talking about it publicly anyway... I don't.) I wouldn't say that just because the "smart" teams are shifting more means much. Having a lot of data or analysts doesn't make you smart. If you're asking the question wrong, it makes you a wrong person with a lot of data.