Adding Value: Shifting More May Help Nationals in 2014

Jeff Zimmermann of the Hardball Times wrote in February about which Major League teams shifted the most last season. The Nationals ranked last in Major League baseball on "major infield shifts," or where three players were on one side of the infield, with just 41 major shifts (the Orioles led baseball with 470 shifts).

Is it any surprise that, with the rise of advanced analysis in the sport, teams would look to defensive shifts to prevent hits and, therefore, runs? Davey Johnson might have eschewed such an approach; sure, he wasn't old school enough to "put 'em in a locker," but if he called the defensive shots last year, he certainly didn't favor adjusting the defense.

No matter, according to Chris Teeter of Beyond the Box Score: The Nats still saved a run or two when they shifted in 2013. But as the author of The Book, Mitchel Lichtman, pointed out in the comments, there were some underlying issues with Teeter's conclusions. And what about Nats' hitters? Stuart Wallace of District Sports Page did an excellent job of highlighting who was shifted, what happened, and how it affected the D.C. squad.

via nbchardballtalk.files.wordpress.com

Presently, then, there's some really good information about when the Nats shifted, how it might have affected their defense, and how it likely affected their offense. I was interested in drilling down more on the defensive aspect of the Nats performance, and Lichtman's comments had me wondering whether we could look at little deeper here. There are plenty of caveats as we move forward, but I think Williams' use of more shifts this year -- as he has said will be the case -- has real potential to improve the team.

The Data

First caveat: I've limited the data set examined to only left-handed hitters who either A). were primarily N.L. hitters who were shifted against more than 58 times last season or who B). were N.L. East opponents who were shifted at least 17 times. Here's the group, with "times hit into major shift" and BABIP with and without shift on (all data from Zimmerman's Hardball Times updated spreadsheet):

So, this is really just a fraction of the overall sample of hitters against whom the Nats could have shifted last year.  That's fine, because we'll see that even using this set, there's room for converting more batted balls into outs.

Here's what this information doesn't tell us:

• How many times did the Nats shift against each player?
• Were there runners on base which would affect when the Nats could shift?
• How many ground ball outs did this group make to the right (pull) side of the infield?
• How many ground ball hits did this group make to the right (pull) side of the infield?
• How many ground balls did this group hit to the left (opposite) side of the infield?
• Did the Nats shift these guys when they should have, and if not, what advantage was lost?

Right, almost more questions than Grady Little faced when he didn't pull Pedro in the 2003 ALCS.

I can't conclusively answer the first two points with publicly available data, although I tried (and if anyone knows, please drop some knowledge in the comments). For the second point, I think we can make some reasonable -- and even aggressive -- estimates to illustrate the overall point.

We can get pretty close on the remaining bullets. To investigate, I looked at every at bat of the above players against the main Nationals' starters last season -- Stephen Strasburg, Gio Gonzalez, Jordan Zimmermann, Dan Haren, Ross Detwiler, Tanner Roark, and Taylor Jordan (for short, and with apologies to Ross Ohlendorf, I'll refer to the group as the "Nats' starters") -- and whether the play resulted in a ground ball out to the right, ground ball hit to the right, or ground ball to the left (regardless of result).

With this more limited sample, and the awesome database of Bill Petti's Ball In Play spray chart,* we can figure out just how many grounders were pulled by these lefties, and how many weren't. In turn, with some estimates regarding baserunners, we can figure how many opportunities the Nats had to shift, and what the run value difference between when they did (and didn't) shift might be. Again, however, there are some limitations, and you're justified in crediting or debiting how worthwhile extra shifts might be based on your tolerance for our guesses.

Spray Charts

Let's check out the spray charts to visualize things. By viewing a park comparison of, say, Matt Adams' at bats last season in D.C. and St. Louis against Nats starters, we can see pull grounders that were hits (the dots in the outfield represent where the ball was fielded, and both were labeled grounders)...

and pull grounders that were outs (red dots):

Several batters also hit a ground ball or two to the left side, although you probably know by now what those would look like, so I won't image that up here.

Grounders: Pull Versus Opposite Field Results

Pull Ground Ball Outs

Overall, the sample had 254 at bats against Nats' starters, which isn't all that huge. Graphically -- and it's probably no surprise -- many of the hitters pulled ground balls to the right side that resulted in outs.  Here's that chart first:

Twenty-seven percent of the sample's at bats ended in a pulled ground ball out. You can see how guys like Ryan Howard, Ike Davis, and Jason Heyward had a tendency to turn on starters' offerings and burn worms. For context, the set averaged a pull percentage of 24% against all MLB competition, and this excludes hits up the middle which would presumably be covered by a shifted infielder in the middle of the diamond.

Pull Ground Ball Hits

This is something we're interested in, to the extent that these were hits that a shift might have prevented. We'll do some estimates about that shortly, but for now, here's how the sample fared:

Of the 254 at bats against Nats' starters identified above last year, only 15 were pull grounders that went for base hits; that's just about 6%, and it doesn't sound like a lot. But before we go rushing a shift, let's see how often these batters went the opposite way on grounders.

Opposite Field Ground Balls

If a hitter can consistently go to the opposite field, he's probably not a good candidate for a shift, regardless of the contact result (for example, Joey Votto pulled the ball just 14.8% of the time, yet he was among the league leaders in batters shifted). How did the sample do here?

The average opposite field ground ball rate for the sample against the Nats' starters here is around 4%. These guys aren't going oppo, as Bryce Harper would say, on the ground all that much. Of course, it is impossible to determine what percentage of this was situational hitting, although I was able to exclude obvious bunts from this set.

Petti's spray charts also provide individual run values for these events. The overall run value of the sample, accounting for ground ball hits and outs to either side of the infield, was -9.91, or almost a win to the good. However, keep in mind here that these values are in a vacuum (in that they do not account for base/out state).

To be sure, none of the numbers perfectly represent the performance of all Nats' pitchers against all shift candidates.  But we've reduced our look to assess the pitchers who most often face shifted batters, and we have a total pulled ground ball value of -- again, just from these sixteen lefties -- 84 occurrences.  Keep in mind that the Nats shifted just 41 times, according to Zimmerman's data.

To better figure out when a shift could have helped last season, we'll need to make some assumptions about factors affecting when a defense can use a shift, namely, whether a runner was on base.

Guestimating Runs Saved

Quick recap:

• The Nats didn't shift a lot in '13 -- just 41 times, in fact;
  
• When we look at how Nats' starters performed against only sixteen frequently shifted lefties, we see that at least 84 times a ground ball was pulled, and 11 times where it was not;
• We don't know when the Nats were shifted within this sample, and we don't know when runners were on base, which would affect when the team might be able to implement a shift.

While we don't know when the team shifted, or when runners were on base, we can use some estimates to ballpark some run value numbers (keeping in mind that these too are used generally).  Buckle up; it's about to get sketchy.

1). Let's first say that of the 254 total at bats against the sample, runners were on base half the time. That's a pretty aggressive estimate, considering the MLB OPB last year was .310.** That leaves us with 127 instances where no runners were on, and the often-shifted lefty is at the plate.

2). We'll also assume that, at least with respect to the pulled ground balls, the Nats were playing in a shift for half of that total -- so, of the 84 times a ground ball was pulled by the sample, we'll call the Nats having shifted in 42 of those occasions. BIt's not perfect, unfortunately, but we're just illustrating here.

3). We also have to account for when the Nats actually did shift in 2013, which is somewhat tricky. Indeed, perhaps the team shifted on 41 of the 42 instances assumed above. If that were the case (and it could be), then we'd be left with just one "shift-able" case. We've been halving everything thus far, so for consistency's sake, we'll say that the 2013 Nats were in the shift half the time during the 42 pulled ground balls events. In other words, we're guessing they burned 21 of their 41 shifts on these guys. This leaves us with 21 non-shift instances against the sample with no one on where the ball is pulled on the ground, but where perhaps the team could have used the shift.

4). We'll keep the overall pull grounder rate constant as we reduce out shift-able cases, as well as the percentage of pull ground ball hits to overall pulled ground balls (~18%), and figure pulled grounders that are hits are all singles.

5). I'm ignoring the ~2 opposite field ground balls that we'd get if we halved the original number of opposite field grounders twice (again, once for runners on, and another time for them just not using the shift for whatever reason).

6). Using these numbers, we can estimate how many runs the team could have saved by employing a defensive strategy which (ostensibly at least) puts them in a more favorable position to convert pulled ground balls to outs (the shift). I've used Petti's run values from his spray charts for the sake of consistency, where a single is worth +.5 runs and a ground ball out is -.28.  Limitations to keep in mind.

7). Here are the run values that a shift might have saved. We start with the same ratio of pulled grounder outs to pulled grounder hits, and back the numbers up to assume an increased out rate due to the efficacy of a shift:

So, if we add 21 shifts, and keep the overall pull grounder rate out/no out constant, we see around a three run improvement as one accounts for the increase in the shift's effectiveness (indicated by the fewer pulled ground ball hits), and over a half win on the top end of the assumptions. Your comfort with how many runs this would save depends on several pieces of conjecture from yours truly, but it does illustrate the range of scenarios. And, naturally, the shift can sometimes work against you as well.

Anyway, on the surface of it, this seems about right: A defensive shift shouldn't drastically change the outcome of a team's fortune. But we were also pretty aggressively limiting the sample -- with the obvious exception of the total times the Nats shifted over the 162 game season, since we know that -- so perhaps we can loosen the lid a bit.

Runs Saved Using Average Base Runners Allowed by Nats' Starters

When the Nats' starters were pitching, an opposing player reached base either via hit, walk (including IBB), error, or hit batsman 29.6% of the time.  I've rounded up and used 30%. Using this number to drive how frequently runners were on base during the 254 sample at bats changes our numbers a bit.

We do the math and end up with just about 178 instances where no runners are on and our sample is at the plate. Keeping the numbers proportional, we have 59 total pulled ground balls (30% off the original pull total of 84) with no runners on.

We can account for the actual Nats shifts a couple of ways here. Let's just say the Nats burned every 2013 shift with these pitchers against these sixteen hitters with no one on base. Well, 59 less 41 is 18, so we still see 18 instances where a shift would seem to make sense. Here are the run values, with ground ball out/no out rates constant:

These are the gains the team might have realized on top of those from shifting the other 41 times. At the high end, we approach over half a win, assuming 100% out conversion, which is dumb.

Last example. Here, we assume the real Nats only shifted on half of the 59 occasions (in other words, they burned 30 of their 41 shifts), leaving us with, call it, 29 cases where a shift might have been used but wasn't.

Remember, we're isolating ground balls, so other batted ball types exist outside this set. By and large, those couldn't hurt when a shift was used, because they are either fly balls to the outfield or line drives that no infield defensive alignment would prevent.

As before, if you believe that roughly 18% of pulled ground balls won't go for outs when using the shift, you've still seen a return of close to half a win within this section of data.

Thoughts

To review the limitations, there are several things we don't know. We don't know the most important pieces of information, which are exactly when and against whom the Nats shifted last year. And we don't know when base runners were on for the sample above, either. But I've tried to be forgiving enough elsewhere to overcome those items, because after all, the team couldn't have used a major shift more than 41 times, and we know that there were 84 times where these hitters against these pitchers pulled a ground ball (never mind every other batted ball event against the team generally last year).

It's also fair to question whether a shift is helpful, period, as Bill James did here. I looked at some of his questions while collecting information for the sample, and some of his concerns weren't realized. For example, in no instance was there an error on the left side of the infield (although, again, we don't know when they shifted). This does not diminish his point, though.

Ultimately, I don't think there's not a prime takeaway in the sense of, "how much better the team could have been in 2013 if they shifted more" or "how much better they will be in 2014 when they do," but instead my goal was to illustrate how both shifting more and shifting successfully (so to speak) can help. Yes, there's a worry about opposite field ground balls, but that's really it -- and at least against this sample, that's not much of a concern.

James Wagner had a great piece in the Washington Post yesterday as a follow up to his article earlier in the week about shifting, and I believe both are great explanatory supplements to this numbers-heavy piece if you want additional information or something much more digestible. Or you could probably pop your favorite beverage, too.  That would work.

Last week, we looked at how the Nats could be a better team just by moving the roles of a few pitchers around. This week, we're seeing again how a little movement -- this time defensively -- could benefit the team.

Thanks to Baseball Prospectus, Baseball Reference, Fangraphs, and of course, Bill Petti's spray charts for essential information and statistics.

*Seriously, before finding this, there was a complete dead end of BIP data -- even Katron was down.

**I realize this leaves out errors and other methods of reaching base, and counts home runs.