Don’t Blame the Shift

Addressing the Issue

 

Speculation aside, let’s deal with some definitives. At the time when Rob Manfred became commissioner of Major League Baseball and first suggested the idea of limiting or banning the shift, offense in baseball was down, and quite dramatically. Offensive production in any sport can be simply defined by how much you score; in baseball, the main capital is runs per game, and the trend over the decade from 2005 to 2014 is undeniable:

 

 

For most of baseball history, the main offensive statistic not named runs per game has been batting average (abbreviated BA), which is a simple rate statistic that calculates the ratio of the number of hits a player records against the number of at bats that player has. Over this same time frame, the league batting average followed this same trend:

 

 

Offense was down in baseball, at an alarming rate. When Rob Manfred earned the seat in the commissioners chair, runs per game in Major League Baseball was the lowest it had been since the 1970’s, before the American League removed weak hitting pitchers from the lineup and implemented the designated hitter. It was seemingly straightforward to connect this decline with the expanded use of the shift, and the idea that a dramatic change in the game, like the one brought about in the 70’s, was not unrealistic.

 

Except, that was 2014. This is 2020, and five whole MLB seasons have run their course since this low point of offense in baseball. So, what happened in these five seasons? The trends may surprise you:

 

 

So, what happened in those five years? Baseball happened, that’s what. The league rebounded immediately, with runs per game increasing from 4.07 to 4.25 in the first year, and then up again to 4.48 runs per game in 2016, continuing to increase to 4.83 runs per game in 2019, the highest runs per game total since 2007, and the second highest runs per game total this millennium.

 

These rebound trends are happening in spite of the fact that the shift, which began being officially tracked by MLB Statcast, the artificial intelligence baseball tracking system, in 2016, was happening across baseball at an all-time high:

 

 

So, that’s the end of the argument, right? Offense over the last 5 seasons rebounded tremendously, even though the shift reached all-time highs across baseball. Period.

 

Except, that isn’t the end of the argument. In spite of these positive offensive trends across baseball, there are those around the league that still believe offense is down overall, and that the shift is largely to blame.

---

Using the Proper Statistics

 

There are still those that cite the drastic decline in batting averages as one of the most significant consequences of the shift. Part of that argument is correct: despite a slight rebound after the 2014 season, batting averages have reached historic lows in this past decade:

 

 

However, there’s a flaw in this argument. The shift is designed to turn balls in play, pitches that are hit in some capacity, into outs. Turn a ground ball that would have been a single against a standard defensive alignment into an out. Batting average is not limited to balls in play, and therefore there use can be used to mask the actual on field trends. If you’re going to argue that the shift is bad for offense, look at offensive outcomes that the shift can actually influence: balls in play.

 

Batting Average on Balls in Play, or BABIP, measures exactly this. By isolating situations that can actually be influenced by the shift, you can more easily quantify the impact the shift can have. And, although batting averages have dropped dramatically, BABIP has actually held fairly steady over the past 8 seasons:

 

 

Looking at this chart, we see that BABIP is almost exactly the same as it was ten years ago, which doesn’t help our friends who believe that the shift is having a significant detrimental effect on offensive output, in spite of the fact that the use of shifts continues to rise.

 

Yet this is only the first step in the right direction. While BABIP is a considerably better way of observing on field trends across baseball than a simple batting average calculation, it’s still imperfect. When you take a closer look at the formula for BABIP, a significant flaw stands out:

 

\[ BABIP = \frac{H-HR}{AB-HR -SO+SF} \]

\[ (H = hits, \ HR = home \ runs, \ AB = at-bats, \ SO = strikeouts, \ SF = sacrifice \ fly) \]

 

On an extreme technicality, home runs are hit out of the field of play, and therefore are not considered a ball in play. Thus, they are not counted in BABIP. The home run, one of the most significant plays in baseball, is actually excluded from the main statistic used in arguments about the use of the shift in baseball.

 

To combat this, BaseballSavant, the MLB Statcast Database, introduced a new stat called BAcon. And who doesn’t love bacon? BAcon, or ‘Batting Average on Contact,’ uses essentially the same formula as BABIP, but includes home run totals:

 

\[ BAcon = \frac{H}{AB - SO + SF} \]

 

Whereas BABIP has held fairly steady over the past several seasons, BAcon, which was first used at the beginning of the Statcast Era in 2015, has steadily increased over that time:

 

 

Using the formula for BAcon and the data from BaseballReference, we can extend this plot to encompass the same time frame we’ve examined throughout this analysis:

 

 

Since 2011, the same time frame over which BABIP remained relatively unchanged, BAcon not only steadily increased, but quickly surpassed the levels from 2006, which was the peak year for runs per game and league batting average.

 

Another step in the right direction, but this still isn’t perfect. A major, fundamental flaw with BABIP and BAcon is that they treat every hit with an equal value. Obviously, this isn’t the case; home runs are considerably more valuable than singles, and should therefore be treated as such. And while the main objective of the shift is to turn potential singles into outs, there’s more to the shift than that. The shift also hopes to turn doubles into singles, or even outs, and possibly potential home runs into weak singles as players try to alter their swing to avoid placing a ball in the teeth of the adjusted defense.

 

Moreover, if the shift is in place to reduce hits, then hits become more valuable. The fewer hits we have, the more important the ones that remain become. And more valuable hits (i.e. extra base hits) become more valuable. The increased value of these remaining hits can account for the loss of hits, or loss of value of hits, as a result of increased shifting.

 

Thus, we need a stat that examines exclusively balls in play, as reducing the effect of those is the purpose of the shift, while also attributing the proper value to each different type of hit that can occur in fair play. In short, we need some adjusted, or ‘weighted’ version of our BAcon stat used above.

 

The formula for this exists, in the form of a stat referred to as wOBAcon, or Weighted On-Base Average on Contacted Balls. The formula for wOBAcon is:

 

\[ wOBAcon = \frac{((w1B*1B)+(w2B*2B)+(w3B*3B)+(wHR*HR))}{AB - SO + SF} \]

 

In this equation, \(w1B\), \(w2B\), \(w3B\), and \(wHR\) refer to the particular weights that are attributed to each type of hit, single, double, triple, or home run. These weights change slightly each year, as determined by an advanced linear weighting system from FanGraphs, the leading website for advanced baseball analytics. For 2019, the weights were: \(w1B = .870\), \(w2B = 1.217\), \(w3B = 1.529\), and \(wHR = 1.94\).

 

The ‘standard’ version of wOBAcon, wOBA (weighted On-Base Average) has become widely accepted as the new best statistic for measuring offensive output. As an all encompassing statistic that properly values each individual type of offensive outcome, wOBA has earned that distinction, and thus wOBAcon has become the best measure of offensive production on balls that are put into the field of play. In other words, wOBAcon is the best statistic to measure the very thing defensive shifts are out to minimize: offensive output when a batter hits the ball.

 

Now that we have the proper statistic, we can plot how wOBAcon has changed over the past 15 years:

 

 

Aside from a slight dip during the 2010-2011 stretch, production on balls in play, as measured by wOBAcon, has held fairly steady over the 10 year stretch where runs per game was dramatically down, and has actually shown to increase in the most recent stretch of time.

 

This graph almost seems to acquit the shift of this wrongful charge of it being the cause of decreasing offense. In the same time frame that the amount and degree of shifting has been being tracked by Statcast (2016-Present), wOBAcon increased from .369 to .378. Offensive output on balls in play isn’t decreasing; in fact, it is actually higher than it’s been in over 20 years!

---

Killing the Shift

 

While Major League Baseball has pondered the idea of getting rid of the shift in some capacity, Major League Baseball players went and did their dirty work for them. As the percentage of plate appearances where some kind of defensive shift occurs increases, production on balls in play has also jumped, as shown by our previous graph of league wOBAcon numbers.

 

But let’s take it a step further, for those who still believe that the shift has a significant impact on the game. The plays where a defensive shift has the largest affect is a ground ball; ideally, the shift puts the infielders in the best position to turn difficult ground balls routine, turn ground ball hits into outs.

 

Using the batted ball data from Statcast, as well as the league-wide shift data referenced earlier, we can see how these statistics relate. Ideally, the more often a team shifts, the more often a ground ball should be turned into an out:

 

 

Except, across the league, the exact opposite trend has followed over the past 4 seasons where Statcast has tracked shifting around the league. The implementation of a defensive shift occurred on a record number of plate appearances in 2019, yet that year saw the lowest proportion of ground balls get converted into outs, a trend that increasingly devalues the use of some type of shift.

 

Combining the league wide wOBAcon trends with the league wide ground ball out percentage versus shift percentage trends, believers and blamers of the shift are shown an image exactly opposing what they believe. The league may be considering some type of limitation on the shift, but not only should they not consider this, they may not have to; if these trends continue, the value of shifting will continue to decrease, and the shift will slowly kill itself.

---

An Alternate Solution: If Not the Shift, Then What?

 

There are still some unanswered questions. If the shift wasn’t responsible for the rapid decline in batting average and runs per game, then what was? Instead of the shift, I offer a considerably more practical solution: strikeouts.

 

It’s no secret that batting averages have dropped dramatically, even when runs per game, BABIP, BAcon, and wOBAcon have been on the upswing as of late. Because of this, there are some around baseball that will continue to insist that offense is down.

 

It’s also no secret that strikeouts are up in baseball:

 

 

Strikeouts have grown steadily over the last 15 seasons, quite similarly to how batting averages have dropped. The correlation between the two is difficult to deny:

 

 

If this seems pretty intuitive, that’s because it should be. The more often a player strikes out, the less often they make contact with the ball. The less often they make contact, the fewer opportunities they have to get hits. The fewer opportunities to get hits, the fewer total number of hits occur.

 

The problem with examining batting average is that batting average, to the pain of baseball traditionalists and lifetime fans, is no longer the ideal statistic for calculation offensive output. It’s easy, it’s familiar, but it’s far from the best way to measure a batter’s prowess. Quite frankly, batters don’t care as much about high batting averages. The best statistic, as mentioned above, is wOBA, or weighted On Base Average, which can be calculated similarly as to wOBAcon:

 

\[ wOBA = \frac{((wBB*BB)+(wHPB*HBP)+(w1B*1B)+(w2B*2B)+(w3B*3B)+(wHR*HR))}{PA} \]

\[ (BB = Walks, \ HBP = Hit \ By \ Pitch, \ PA = Plate \ Appearances) \]

 

The weighted values for wOBA are the same as the ones used for wOBAcon. However, this statistic measures total offensive output, and therefore includes all possible ways of reaching base safely. So, if this is our best statistic for analyzing offensive production, we should be examining this statistic:

 

 

While one might be quick to say that the league wOBA numbers follow the same trends as batting average, and as a reciprocal, strikeouts, there’s some issues with this. First off, wOBA has begun to rebound in recent years, much like many other offensive statistics, which, if strikeouts were a high contributing factor to wOBA, wouldn’t make sense. The relationship between the two statistics is actually very unclear:

 

 

The reality is, there is quite literally no correlation between wOBA and strikeout percentage. Players have always wanted to maximize their offensive output, and in the modern MLB, that means maximizing your wOBA, regardless of what that does to your strikeout totals, and, in turn, your batting average.

 

If batting average is the stat you insist on using for some reason, 1) stop being so stubborn, 2) consider this next point. If you use the BAcon statistic we used earlier and subtract regular batting average from BAcon, the difference between them represents the difference in offensive output when you remove strikeouts from the equation:

 

 

Plotting this difference between BAcon and batting average against strikeout percentage tells an interesting story. Basically, what this is saying is that players are making contact less often, as shown by the increase of strikeouts, but players are getting hits more often when they do make contact. Quite pleasingly, this further disproves the idea that defensive shifts are bad for baseball. Players are quite literally getting more hits when they hit the ball. However, the increase in strikeouts is so much that even though batters are having more success when they hit the ball, they’re hitting the ball at such lower rates that batting averages are dropping.

 

We can perform this same analysis with wOBA and wOBAcon, subtracting the two values to gauge what kind of difference strikeouts make in the equation:

 

 

We see, for all intents and purposes, the exact same trend as above. Batters are making contact less often, but their production when they do make contact has increased greatly. By using wOBA/wOBAcon as opposed to BA/BAcon, thus implementing the weights on each particular outcome, this further illustrates that the offensive output on balls that are being hit is increasing. Players are creating more offense when they hit the ball (sorry shift) than they have in over two decades. And, if weighted production is increasing, in spite of the fact that strikeouts are also increasing, then why should it matter if batting average is decreasing?

---

2014: The Dark Age that Led to a Revival of Offense

 

Almost every chart and graph we’ve produced has displayed one particular trend: 2014 sucked for hitters in baseball. In fact, 2014 was such a bad year for offense around baseball, that a pitcher, a guy that only plays in 20% of his team’s games, won the league Most Valuable Player Award (no offense Clayton Kershaw). Runs per game was the lowest it had been since 1976, which was part of a rebound from the era of baseball history known as the “Dead Ball Era,” a period of time that brought about a myriad of changes within baseball, including lowering the pitching mound from 14" to 12" and implementing the designated hitter and removing the pitcher from the batting order.

 

However, the question we’re asking isn’t “why was 2014 so horrendous,” but instead, “what changed after 2014 to fix that?” If the metrics we’ve used thus far have only illustrated a trend, but not helped explain that trend, what can we use to quantify this trend? Let’s begin by first examining the most basic building block of offense in baseball: the number of hits a player, or team, or league, has:

 

 

While the number of hits recorded by the league steadily declines from 2005 to 2014, they become fairly stable from that point on, with a slight uptick in total hits in 2015 before becoming stable. But the important thing this graph shows us is that the number of hits being recorded didn’t jump drastically to match the jump that the other offensive categories made.

 

What if, instead, we examine the quality of these hits. A metric like wOBA or wOBAcon, with the way the weights work, can be heavily influenced by an increase in extra base hits, doubles, triples, and home runs. Namely, we want to look at doubles and home runs, since triples are so rare:

 

 

 

After 2014, both doubles and home runs, home runs in particular, increased at a considerably quicker rate than overall hits, which barely increased at all. There were approximately 2500 more home runs in 2019 than there were in 2014, approximately 400 more doubles in 2019 than in 2014, but only approximately 400 more total hits.

 

Well, that is pretty revealing. Hits barely increased, but home runs increased by 62% in just 5 years. However, raw numbers can be unreliable, so it may be better to view these numbers as a rate statistic:

 

 

 

Although the total number of hits remained relatively unchanged, the percentage of hits that led to extra bases increased dramatically. Considering how much home runs and doubles can influence a weighted value metric like wOBA, it’s no wonder that wOBA and wOBAcon had such significant rebounds after the 2014 season, bringing baseball out of a season starved of offense, and in just 5 short years, producing one of the best offensive seasons in recent baseball memory.

 

This can also give us insight as to why offense had plummeted across baseball leading up to 2014. Not only were hit totals down from 2006-2014, but doubles were disproportionally down over that same time frame; the percentage of hits that resulted in doubles dropped a full percent. One percent might not seem like much, but a one percent decrease of 9,000 total doubles has a significant impact on a stat like wOBA, where the difference between .300 and .310 is fairly significant.

 

While these metrics help to answer why offense was heading in the wrong direction, and then rebounded after 2014, one question still remains: why 2014? What changed in the winter of 2014-2015 that helped save baseball?

 

In a word: Statcast. During the 2014-2015 off season, Major League Baseball installed the artificial intelligence systems into every Major League stadium and made every piece of data collected available to the public. This changed everything. Coaches, players, management, and player development staffs were privy to a new wealth of information. Launch angle, the angle the ball has off the bat, and exit velocity, the speed in which the ball is moving off the bat, became part of baseball vernacular. They could quantify what people already knew: hit the ball hard, you’ll get more hits, hit it in the air, you’ll hit more home runs.

 

These very basic aspects of baseball offensive success were now quantifiable, which meant you could now train those specific aspects of performance. Weighted On-Base Average became a new normal around baseball, as players began to understand the seemingly obvious parts of the game; extra base hits are worth more to your team than singles, getting on base more often means you can score more runs. Statcast made all of this quantifiable in a way never seen or understood before in baseball. Chicks always dug the long ball, but now, so do the metrics.

---

Where This Leaves Us

 

The debate around the shift continues to rage on. Regardless of any amount of data or statistical evidence there is contradicting the idea that the shift is killing offense, there are still those around baseball who believe this extreme defensive strategy is bad for the game. On the other hand, there are people who simply believe if a defense is going to play so extreme, hitters should just adjust; hit the ball where the defenders aren’t. Find a way to beat the shift

 

Hitters did adjust, albeit not the way baseball expected; instead of trying to hit more ground balls, just not at the defenders, hitters decided they wanted to hit more doubles and home runs. Hitters beat the shift. With balls in gaps, off the walls, and out of the park. They didn’t hit around the shift, they hit over it. Doubles and home runs are being hit at all time rates. Offensive production on batted balls is as high as it’s ever been. Runs per game have rebounded back to near their highest point in the last 15 years.

 

Baseball hasn’t been one of the most popular sports in America, and growing around the world, for the last 150 years for nothing. Baseball is resilient. One aspect of the game changes, and the rest finds a way to adjust.