The Psychology of the Infield Shift

Over the past 10 seasons, the shift has gone from a quirky defence against some pull-heavy left-handed hitters to being employed against over 50% of lefties and over 20% of righties. In that time multiple researchers have shown that the batting average on balls in play (BABIP) is lower when players are shifted, and therefore shifting works.

But BABIP alone does not paint the whole picture, a shifted at-bat does not take part in a blind test where the pitcher and hitter do not know the defensive alignment. They both do and this will affect the decisions they are making during an at-bat.

Recent research (Tom Tango in 2017, myself in 2018 and 2019 & Russell Carleton in 2020) has shown there is more than just BABIP to consider when evaluating the shift. We have all shown, with different methodology, that there is an inherent walk penalty when shifting which in some situations suggest that shifting is inherently bad for the defensive team. 

I wanted to investigate the decision-making that pitchers and hitters are making when there is a shift on and when there is not, focusing on pitch location from pitchers and location-based swing rate from hitters. It analyses these to see if there is anything different occurring.

Pitch level info for shifting exists in the Statcast dataset from just 2015 onward. I used that data to investigate the following three things for each season since 2015.

  • Outcome of all at-bats, not just balls in play
  • Swing rate of hitters – split by attack zones
  • Pitch location for pitchers – split by attack zones

To investigate this, I have split the data into four separate sets based on the handedness of the pitcher and the hitter. And for each of these sets look at the impact of a standard (full) shift and a strategic (partial) shift.

Firstly, there are other reasons that determine where the infield players are beyond who is the best; if there is anyone on base this influences the positioning. So, to truly interpret if the changes in outcomes and behaviours are the impact of the shift alone, we should look at cases where the bases are empty. So, all data in this piece is looking at bases-empty situations only.

Secondly, I need to account for the hitters and pitchers for when the shifting is happening.

To account for hitters, I have weighted each player on the number of times they were shifted and not shifted. If a player were shifted 100 times and not shifted 500 times, then their weight would be 100. If they were shifted 300 times and not shifted 300 times, their weight would be 300. I repeated this for all hitters and came up with the weighted average. 

To account for pitchers, I calculated the same rates for all pitchers and pro-rated them to give expected rates from these at-bats. Using those figures, I then calculated how much above or below this expected outcome the results were. I scaled this to 100, with 100 being inline, less than 100 being below average, and greater than 100 being above average.

If we take the data from 2015-2020 seasons and combine them together, we can see the overall performance of the shift.

For the highlights, green is good for the defensive teams and red is bad.

The graph above shows the differences of the shifted plate appearances to the non-shifted plate appearance, and is shown in percentage points.

For example, the 11.4% increase for L v L strikeout rates equates to roughly a 2.2% increase in actual strikeout rate, as the expected strikeout rate is around 22%.

As expected, the BABIP is lower in all scenarios, bar the RHH v RHP strategic shifts (which has the lowest volumes by far). And there is the walk penalty which has been previously researched.

If you look at the wOBA and xwOBA columns the thing that clearly jumps out though is, shifting works against lefties but not against righties and that seems to be driven largely by the differences in strikeout rate.

Do we see any differences with the approach of pitchers and hitters?

Looking at the pitch locations split by Statcast zones, with green highlighting showing more pitches thrown in that location when shifted compared to no, and red showing fewer.

This shows that pitchers have been pitching a little more often out of the zone against lefties but a little more often in the zone against righties. That is interesting but does not to me explain the differences we have been seeing, so let us look at the hitters swing rates for these zones.

This shows a clear difference in the behaviour of right-handed hitters compared to left-handed hitters.  Lefties are swinging more at all types of pitches when they are shifted, righties are not.

If we look at players who faced more than 100 shifts and non-shifts, and plot their swing rates in the shadow zone, we get the following.

We can see that these changes are not just the impact of a handful of players, there are trends that are happening across multiple players.

This difference looks to me to account for not only the changes in outcomes but probably the behaviour of the pitchers. But that might be a chicken and egg scenario so let us look at past seasons individually to see when this behaviour came about.

How did we get here?

We will start with 2019, I am excluding 2020 here as the volumes for some of the data points are quite low but do note it follows a similar pattern in general.

In 2019, the behaviour here is in line with the overall behaviour so these trends must have started earlier. Interestingly though, the change in BABIP in six of the eight scenarios was to the detriment of the defensive team.

In 2018, the behaviour here is still in line with the overall behaviour so these trends must have started earlier.

In 2017, the behaviour here is still in line with the overall behaviour so these trends must have started earlier. The big changes in Waste Zone swing rate can be ignored as the volume behind these is small.

In 2016, we see that behaviours of the hitters are still the same but that of the pitchers is less present than the years afterwards.

In 2015, we finally see no discernible pattern the behaviour of pitchers. But we still see the behaviour of the hitters which suggests that the hitters changed their behaviour, and pitchers changed their approach to respond to that.

As stated earlier, sadly I do not have access to pitch level shift data before 2014 so we cannot see when this trend for hitters started. This does mean we can only speculate the reasons behind this behavioural difference.

I believe that right-handed hitters saw what left-handers had to deal with and then adjusted because of it. They became more patient in their approach when facing the shift. But that sadly is conjecture.

But it does seem to me that it would behove some left-handed hitters to speak to the right-handed counterparts and learn from their approach.

Limitations

  • The shift may be working well on hitters who are now shifted nearly 100% of the time; these individuals would not appear in this methodology.
  • It does only look at bases-empty situations which account for 60-65% of all shifted plate appearances in a season. So, there is the possibility that the shift is working well enough in those other situations to make it worthwhile.
  • Date only goes back to 2015, so it is conjecture for before then.

Russell is Bat Flips and Nerds’ resident analytical genius, and arguably Europe’s finest sabermetrician. If you’re not following Russell on Twitter @REassom then you’re doing baseball wrong.

 

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.