The Thing With Shifts

Sunday was the first day on the job for new baseball commissioner Rob Manfred. He wasted no time in wading into controversy. In an interview with ESPN's Karl Ravech that day, he said,

For example, things like eliminating shifts, I would be open to those sorts of ideas.

He later backtracked a bit, but the idea of eliminating shifts is now on the table.

This isn't new. Sports Illustrated writer Tom Verducci proposed eliminating shifts last summer, and I responded to his idea then. Then, as now, I don't think it's a good idea, but let me unpack the argument a bit more here. 

What kind of shifts are we talking about?

This isn't the usual shifts, like infielders guarding the line to protect against doubles, or outfielders playing shallow when a low-power hitter like the Phillies' Ben Revere is at the plate. These are what some call "extreme shifts," generally defined as three infielders on one side of second base. Take the example at right. That's a game between the Astros and the Mariners in April 2013. The Mariners' Franklin Gutierrez is at the plate. The Astros' second baseman, shortstop, and third baseman are all playing on the left side of second base. 

Why are they doing this? Because Gutierrez has a pronounced tendency to hit ground balls there. Here's a chart showing the location of his batted balls in 2012 and 2013. (Source: FanGraphs.) The green dots are grounders. As you can see, Gutierrez hits the vast majority of his ground balls to the left side. Put another way, if Gutierrez hits a grounder, he's likely to hit it within range of the Astros fielders above. That wouldn't necessarily be the case in a traditional defensive alignment, with the second baseman on the first base side of the bag and the shortstop further toward second.

How prevalent is this?

Well, shifts like this have been around for a long time. The first instance I could find was used in the 1920s against Phillies outfielder Cy Williams. The most famous one was probably the 'Williams Shift" that teams employed against a different Williams, Red Sox slugger Ted. A left-handed batter, he tended to pull balls to the right side of the field, so teams would line up fielders there. It began in a 1946, in a doubleheader in July. In the first game, Williams went 4-for-5 with three homers and eight RBI in a game against the Indians. In the second game, Indians manager Lou Boudreau moved his team into a crazy alignment when Williams came to the plate. (I figure Boudreau effectively said, "Shit, I don't know, let's try this" as a last resort.) The outfielders shaded to the right with the rightfielder playing deep. The shortstop (Boudreau) moved between first and second, the third baseman played to the right of third, and the second baseman was in the short outfield, near the right field line. There were effectively four outfielders (two in right), three infielders to the right of second base, and none to the left of second. Williams got two walks and a double in four trips to the plate, scoring two runs in a 6-4 Red Sox victory, so in the context of Ted Williams, maybe it worked. One of the baseball card companies, Fleer, commemorated it with a card, shown above.

Anyway, the Williams Shift had limited success, given that the Splendid Splinter finished his career with arguably the best hitting statistics of all time. Shifts like this didn't really catch on widely until recent years, as baseball analytics have enabled teams to chart every batted ball, resulting in charts like the one I showed above for Franklin Gutierrez.

The 2015 Bill James Handbook has a section on shifts, full of bad puns (e.g., the title is "Who Gives a Shift?"). I'm not going to reproduce everything here--buy the book, it's really good, if you want details--but one of the interesting figures is that there were 2,357 shifts employed in 2011 but 13,296 in 2014. That's a 464% increase in three years. In terms of shifts per game, that works out to 0.97 shifts per game in 2011 (that's the total for both teams) and 5.47 shifts per game in 2014. In 2014, the Houston Astros shifted the most, 1,341 times, or 8.3 times per game. That's extreme, and there were only six teams (Astros, Rays, Yankees, Orioles, Blue Jays, and Pirates) that shifted more than four times per game.

So to answer how prevalent shifts are: Growing rapidly, but other than a few teams, not deployed all that often.

How much have shifts suppressed offense?

The Bill James Handbook answers that pretty clearly:

If you look at the league as a whole, the average decrease in batting average when players are shifted is 30 points.

That's a lot. When Boston's David Ortiz, who faced a shift in 95% of his plate appearances, hit a ground ball or a soft liner--the types of batted balls the shift is designed to turn into outs--he batted .201 when the shift was on but .250 when it was off. 

The Bill James Handbook estimates the impact of shifts in 2014 was 195 runs. In other words, shifts prevented 195 runs from being scored. Thirty points in batting average, 195 runs--no wonder Manfred's thinking about stamping out shifts, right?

Except that's not really big, as Ben Lindbergh points out in this article and Dave Cameron does in this one. (For comic relief, read this post by Grant Brisbee.) That 30 point suppression of batting averages, as I pointed out, occurs in only 5.5 plays per game, on average. And 195 runs scored...well, there were 19,761 runs scored in 2014. 195 equals a shade under 1%. Teams scored 4.07 runs per game last season; adding back 195 boosts the figure only to 4.11.

And I want to reiterate a point I made in July. Let's look at the league-wide batting average on balls in play and the batting average on grounders. (The difference between overall batting averages and batting averages on balls in play is that the latter excludes strikeouts and home runs, neither of which can be defensed.) Here's the table over the past ten years, from FanGraphs:
           Batting Avg   Batting Avg
   Year   Balls in Play   Grounders
   2005      .295           .233
   2006      .301           .236
   2007      .303           .239
   2008      .300           .237
   2009      .299           .232
   2010      .297           .234
   2011      .295           .231
   2012      .297           .234
   2013      .297           .232
   2014      .299           .239
Do you see a pattern there? I don't. Batting average on balls in play went up in 2014 despite the explosion of shifting. And the batting average on grounders was the second-highest in the 13 seasons for which FanGraphs has data (.2386 compared to .2394 in 2007). What gives? If shifts are hurting batting averages, why aren't these batting averages going down? And if the averages aren't going down, why are runs being suppressed?

Why aren't batting averages and grounders and balls in play lower?

Because batters are really good. Seriously. I mean, we hear all the time about how pitching's dominating the game. The average fastball velocity last season was 91.6 mph, the highest in history. Pitcher struck out 19.4% of batters, the most in history, and walked only 7.1%, the fewest since 1921, which is just crazy. But at the same time, starting pitchers are typically done after six innings. (2011 was the last time starters averaged more than six innings per start.) The reason? Yes, there are pitch counts and eight-man bullpens and all that, but also, there aren't easy outs. Despite all the challenges for hitters, they're still batting about .300 on balls in play and in the .230s on grounders. That hasn't changed, and it's a reflection of the skill of major league batters.

So why is scoring down?

Well, let's look at that batting average on balls in play in two ways. First, let's consider the type of hits that batters get. I'm going look at the number of singles, doubles, and triples batters get when they hit balls in play. Add them together, and you get the total batting average on balls in play. Here's the table: 

Season	1B	2B	3B	Total
2005	0.221	0.067	0.007	0.295
2006	0.225	0.069	0.007	0.301
2007	0.226	0.070	0.007	0.303
2008	0.224	0.069	0.007	0.300
2009	0.224	0.068	0.007	0.299
2010	0.224	0.066	0.007	0.297
2011	0.222	0.066	0.007	0.295
2012	0.223	0.066	0.007	0.297
2013	0.226	0.065	0.006	0.297
2014	0.227	0.065	0.007	0.299

Singles and Doubles per Batted Ball, 2005 = 100

That's not very helpful, is it? But given that triples are pretty rare events, occurring in about 0.7% of plays, let's focus just on singles and doubles. In 2005, when balls were put in play, batters got a single 22.1% of the time and a double 6.7% of the time. Let's use that as our baseline, and set those years equal to 100. Look at the chart at right. Over the past few years, singles are up and doubles are down. Overall batting averages on balls in play, as we've seen, are holding steady. But there are more singles and fewer doubles. That's not because of shifting--shifts suppress singles, not doubles--but it affects scoring overall.

Home Runs and Strikeouts per Plate Appearance, 2005=100

Now, let's look at batting averages on balls in play another way. What's excluded? Home runs and strikeouts. Here's another chart like the one above. But this time, let's look at home runs and strikeouts per plate appearance. This one's more pronounced. Home runs are in decline, while strikeouts have moved unremittingly upward. When batters are hitting balls in play, they're getting more singles but fewer doubles, as the chart above demonstrates. When they're not hitting balls in play, they're striking out more and hitting fewer homers.

So are shifts the problem?

There is no question that offense is down. Teams scored 4.07 runs per game in 2014, the lowest in a non-strike year since 1976. The reasons, as I've shown above, are clear: More strikeouts, fewer homers, fewer doubles. Shifts have played a minor role and are pretty insignificant compared to the other factors. If the new Commissioner wants to address scoring, he should look at ways to promote more contact, and more hard contact, by batters, not worry about where the fielders are standing.

(Thanks to JBM for inspiring me to write this.)