# Exploring the Shift Dynamic

Shifting is all the rage, I don’t think that’s news to the common baseball fan. That there are now more shifts than ever employed—against righties and lefties—in every-day baseball only gives credence to it’s legitimacy as a staple of baseball strategy. But I think many common baseball fans, those who are told that it’s ruining the game or horribly ineffective, would argue the legitimacy of the shift—something that looks, on the surface, a bit ridiculous. Well, like anything in baseball, it’s a science, it’s a gamble, (and for many teams) a highly calculated risk.

When shifting, a team must consider the probability of success and failure, and the subsequent run value (risk or reward) that come with it. Like stolen bases, a shift is a gamble between a few outcomes, in order to increase the chances of turning a batted ball into an out and neutralizing the bread and butter of the game’s biggest hitters. In most shifts, teams are betting that the extra value added by converting more ground balls and line drives into outs, outweighs the (sometimes slight) chance the player beats the shift by going the opposite way or splitting the shifted fielders.

Still, this begs the question: what determines a successful shift against a player and what factors encourage a team to employ a shift against said player? Running a regression of the 2013 data from Jeff Zimmerman, I ran a model of “times hit into shift”—proxy for which players were shifted the most—on a number of important factors. A few important takeaways, the *ratio of opposite hit outs to pulled outs* was highly predictive—explaining nearly 33% of the variation in the amount of times a player hit into a shift. ISO or wRVfb — which I introduced here yesterday — along with the aforementioned ratio, when added to the regression made it so that nearly 45% of the variation was explained by the regression.

However, in a shift setting there is still the possibility of everything but a pulled ground ball resulting in an out—a line drive to the opposite field, a pulled fly ball, an up-the-middle knock, a Matt Adams inspired bunt down the third base line. But it is debatable if these factors have to be accounted for in addition to the probability the shift works out and adds value. A shift is made to do one thing most importantly—diminish the rate of ground balls to the pull side that go through for hits. So given the recent onslaught of shifts in the Major League Baseball, we should except some kind of pattern—that suggest a decreased amount of batted ball production—to show across the league. Below, is a plot of Batted Ball runs (simply wRVbattedballtype * number of batted ball type) from 2008-2013:

Surprisingly we see a constant measure of GBruns league wide, despite the onset of an unprecedented amount of shifts. Meanwhile, LDruns and FBruns have fluctuated, with a decreased amount in recent years—likely due to the fact of more strikeouts (fewer balls in play) and also classification errors that make no two seasons of FB’s/LD’s similar. But what happens when we isolate for player’s who faced the most shifts (15+) in 2013, what does their trend look like against the increasing amount of shifts in baseball?:

So both line drive production and ground ball production have decreased over time from 2008, while FB runs created has seen the biggest impact in recent years. Why fly ball production? Does the shift really effect production of a power hitters fly balls? As previously mentioned, a player’s ISO or wRVfb correlates very well with amount of times hit into shift. For this reason, can the shift be taking away from a power hitter’s slugging abilities by forcing opposite field hitting?

If a player hits a three run bomb the opposite way, is it because of the shift? Did the shift influence the hitter to hit a fly ball home run? To test this theory, in the most rudimenteary way, I took all players who had been shifted 15 plus times in 2013 and graphed their opposite hit fly ball percent as a trend from 2008-2014:

Has shift effected frequency and value of opposite hit outcomes? |
---|

Season |
OppFB% |
run value |
OppGB% |
run value |
OppLD% |
Run Value |

2008 | 8% | -0.06 | 3% | 0.09 | 3% | 0.32 |

2009 | 8% | -0.02 | 3% | 0.07 | 3% | 0.36 |

2010 | 7% | -0.07 | 3% | 0.09 | 3% | 0.32 |

2011 | 7% | -0.07 | 3% | 0.02 | 3% | 0.32 |

2012 | 7% | -0.07 | 3% | 0.04 | 3% | 0.31 |

2013 | 6% | -0.11 | 3% | 0.03 | 4% | 0.26 |

2014 | 6% | -0.14 | 3% | 0.08 | 4% | 0.23 |

As it seems, the onset of the shift has not really changed the probability of an opposite hit ball. But, for those most shifted in 2013, there has been a large decrease across the board in the value of their batted balls, with the average ground ball losing 66% of it’s value until this year—where there seems to be a spike in ground ball run values. But the most surprising factor is the apparent effect of the shift on the fly balls and line drive value, which have both dropped significantly in value without encouraging the hitter to take a different approach to take it the other way. What we really need is to detect when the shift is on for all players, and compare the spray and outcomes of their batted balls in shifted versus non-shifted situations.

So it doesn’t *seem *like hitters have gotten any better going the other way, despite heavy shift age against them. For this reason, it is probably best to focus on opportunities where we believe the shift could improve a team’s defensive efficiency. So while, theoretically a shift should encourage hitters to poke the ball the other way, you wouldn’t necessarily consider a shift with a player who already sprays the ball all over the infield. Likely because standard positioning gives the team the best shot, and moving to a pull shift would mean more opposite hit balls finding the hole. Like in any gamble, there is a break even point—a point where the returns overmatch the risk.

### Introducing Shift Score

Given what we showed in earlier articles—that each player has distinct run values based on spray and batted ball type—what we want to know is the individual indicators where shifting that batter and getting the out, outweighs the probability where the shift does not work. Every player has different batted ball values, based on type and spray. For this reason, individual assessments of shifts should be based on individual run values rather than a general one.

In my first attempts I learned that for the most part (from what we know about 2013), those who rely on their fly ball for run production and tend to pull more ground balls and line drives for outs. Those who belong to said group tend to be shifted and hit into more shifts (Ryan Howard rings a bell). But in my estimation, a shift’s possible value is determined by more than just pull%, it is based on a myriad of factors:

- The value and frequency of a hitter’s ground balls relative to the value and frequency of his pulled balls to infielders.
- Standard deviation of ground ball angles, pulled and opposite (clustered means easier to position— think of Nick Punto‘s trademark ground out to the shortstop— while more variability means harder to predict spray).
- Ratio between standard deviation of distance between pulled outs versus ground balls (proxy for hard hit ground balls, those who hit ball harder will have larger ratio. EX: Miguel Cabrera versus Ben Revere)
- Overall power, ISO or wRVfb (due to high correlation with times hit into shift — some teams will just shift a power hitter for the heck of it).

While I imagine its pretty easy to look at a player’s spray chart and determine “shift-proof” or “shift-worthy,” it definitely is a subjective way of doing things. Each player has different batted ball run values. If I shift Yasiel Puig, I am more afraid of him hitting a rocket on the ground the opposite way (and stretching that for a double) than I am of a Shelley Duncan actually hitting the ball the other way, for once.

Unfortunately, right now we don’t have much of an objective way to define a shift pull hitter without looking at a spray chart, while I’m sure teams use a combination of scouting data and proprietary methods to determine their positioning. So today, I will introduce shift score, a way to measure what we see on the spray chart and express it in mathematical terms.

Today I will try to incorporate the first two points, but the third will have to hold up until we define “usual positioning” for each player. Play-by-play shift data is not readily available, otherwise this model may look different. However, using Gameday data, this is the best proxy I could think up as a “SHIFT SCORE” metric, to identify player’s who are suitable to be shifted:

SHIFT SCORE = (Pull_GB%/RVpullgb) / (Opp_GB%/RVgb) * (sd_angleopp/sd_anglepull)

This part is “mathy”, so skip it if you please.

Two disclaimers: one, this is based on the assumption that the batted ball distribution of a player wont change pre-shift to after-shift — when it very well could would in real circumstances. Plus, there might be a slight selection bias for players who are already shifted a bunch, so I included a section of players who did not see many shifts in 2013 so that we can identify players who should be shifted and are not already.

So both Pull_GB% and RVgb are directly proportional to shift score, while RVpullgb and OPP_GB% are indirectly proportional. Basically that means the higher a player’s Pull_GB%, the higher his shift score—same deal for his run value on all ground balls. In other words, players who pull the ball a lot, and have high amount of ground ball production are shift candidates. Now factor in their production to the pull side—where a higher run value on pulled ground balls means a smaller shift score, and vice versa. Meanwhile, a player who hits a lot of ground balls the opposite way will see a lower shift score, while player’s who don’t will receive just the opposite. Lastly, batted ball angles are accounted for, where the standard deviation of angle of opposite hit ground balls is directly proportional, at the same time, the standard deviation of a player’s pulled ground balls is indirectly proportional to shift score. In simpler terms, a player who tends to hit the ball in the same spot, with the least variation, on the pull will receive a higher shift score—while someone who sprays the ball the opposite way will receive a lower shift score.

All players with at least 200 BIP since 2008 were included, as long as 50 came in 2013. Pull was defined as anything left or right of 2nd base depending on handedness, determined by hit angle. Angles that were accounted for were the ones that were fielded by infielders. In GB%, I also included LD’s under 200 ft, perhaps playable or effected by pull shift, etc.

Top Ten “Unshiftable” Righties |
---|

Name |
Stand |
Shift |

Ed Lucas | R | 0.74 |

Nick Punto | R | 0.82 |

DJ LeMahieu | R | 0.97 |

Emilio Bonifacio | R | 1.06 |

Willie Bloomquist | R | 1.06 |

Ryan Hanigan | R | 1.13 |

Jamey Carroll | R | 1.14 |

Eric Young | R | 1.15 |

Everth Cabrera | R | 1.17 |

Shane Robinson | R | 1.22 |

Top Ten “Unshiftable” Lefties |
---|

Name |
Stand |
Shift |

Ichiro Suzuki | L | 0.87 |

Norichika Aoki | L | 1.06 |

Endy Chavez | L | 1.11 |

Melky Cabrera | L | 1.11 |

Emilio Bonifacio | L | 1.12 |

Dee Gordon | L | 1.13 |

Brett Gardner | L | 1.16 |

Gregor Blanco | L | 1.21 |

Omar Quintanilla | L | 1.22 |

Everth Cabrera | L | 1.22 |

Top Ten “Shiftable” Righties |
---|

Name |
Stand |
Shift |

Carlos Beltran | R | 6.37 |

Josh Willingham | R | 5.93 |

Casper Wells | R | 5.18 |

Kevin Youkilis | R | 5.06 |

Martin Maldonado | R | 5.05 |

Jose Iglesias | R | 4.88 |

Jason Bay | R | 4.43 |

Kelly Shoppach | R | 4.34 |

Devin Mesoraco | R | 4.31 |

Mike Napoli | R | 4.28 |

Top Ten “Shiftable” Lefties |
---|

Name |
Stand |
Shift |

Jimmy Rollins | L | 4.59 |

Ryan Howard | L | 4.44 |

Justin Smoak | L | 4.34 |

Carlos Pena | L | 4.31 |

Seth Smith | L | 4.27 |

Curtis Granderson | L | 4.08 |

Chase Headley | L | 4.08 |

Jarrod Saltalamacchia | L | 4.07 |

Nate McLouth | L | 4.04 |

Jason Giambi | L | 3.81 |

Here, I disregarded all who were shifted 15 times or more in 2013, so that this metric could be more informational.

Top Ten “Should Be Shifted More” Lefties |
---|

Name |
Stand |
Shift |

Jimmy Rollins | L | 4.59 |

Nate McLouth | L | 4.04 |

Alex Avila | L | 2.69 |

Dexter Fowler | L | 2.65 |

Nate Schierholtz | L | 2.63 |

Roger Bernadina | L | 2.58 |

Asdrubal Cabrera | L | 2.54 |

Will Venable | L | 2.41 |

Alejandro De Aza | L | 2.35 |

Daniel Descalso | L | 2.31 |

Top Ten “Should Be Shifted More” Righties |
---|

Name |
Stand |
Shift |

Martin Maldonado | R | 5.05 |

Jose Iglesias | R | 4.88 |

Kelly Shoppach | R | 4.34 |

Devin Mesoraco | R | 4.31 |

Franklin Gutierrez | R | 4.00 |

Wilin Rosario | R | 3.93 |

Steve Pearce | R | 3.78 |

Scott Hairston | R | 3.74 |

Danny Valencia | R | 3.72 |

Matt Kemp | R | 3.60 |

If you’d like to try the eye test, I recommend trying Bill Petti’s Interactive Spray Chart Tool, try isolating for LD’s and GB’s under 200 ft to see if the shift score passes the eye test to you. I tried Aoki, one of the most unshiftable lefties, versus Carlos Santana, one of the most shiftable active lefties (when he isn’t hitting righty). From Bill’s tool, Aoki versus Santana:

Yeah, I would try not to shift on Aoki. Luckily, no one really dared to in 2013. But he still only had a .295 BABIP and a 3 GBruns in 2013. While, it’s not the end all be all of the argument, sometimes unshiftable does not mean unbeatable. It does seem that for the most part the shift score passes the eye test, but I’d like to run a few tests on it. Mostly, I’d like to know how measures of spray, batted ball rates, and then wRVbb effect this measure. If the the metric did what we wanted it to do, we should see a relationship between wRVgb on the metric, and less of a relationship for batted ball rates. In the future I will look at season-to-season shift scores and see how much they fluctuate, etc. The rest of the Shift score data is available upon request, for all players since 2008 with at least 200 balls in play.

### Next Steps

So today, we incorporated spray and batted ball run values into our understanding of batted ball performance. In my next few articles I will look at the stabilization of batted ball spray, which will give us insight into how the shift score relates to time in the big leagues. Meanwhile, I will also look into estimating fielding positions for each individual player and regressing their batted ball run values based on the distance hit from the fielder. It would also be awesome to have shift play-by-play data so we could calculate actual shift success and shift failure values, instead of my proxy one. That way we could see how GB distance, standard deviation of pulled GB angles, pull to opposite LD/GB ratio would affect the actual shift effectiveness against a player. If I ever do get my hands on the right data, I would love to tackle a Shift BEP.

For now, I’ll sit and wait for that data. Feedback is welcomed and encouraged, so let’s discuss. And as always, if you want the code, please do reach out.

### References and Resources

- Thanks to Gameday and Major League Baseball Advanced Media.
- Zimmerman, Jeff. “Early Hitter Shift Data,” RotoGraphs.
- Zimmerman, Jeff. “Expanded 2013 Infield Shift Data.” The Hardball Times.
- Petti, Bill. “Introducing the Interactive Spray Chart Tool,” FanGraphs.

Print This Post

Ok, gotta ask this! If you remove Matt Adams from the “Has shift effected frequency and value of opposite hit outcomes?” chart, how much do the numbers change? His production has shifted so dramatically in 2014 (at least for the first few months) that I wonder if he skewed both the OppFB and OppGB average run values all by himself.

Max, excellent, excellent stuff! I’m gonna have to plow through this in more detail over the weekend (gotta get back to my day job). Thanks!

Adam Dunn has also changed since mid-2013, not as dramatically but still significantly.

Google spreadsheet with all shift scores, please?

I’m out of town but will post it here as soon as I get back.

I don’t understand why the standard deviation of opposite side groundballs is proportionalto shift score. Wouldnt that mean a batter that sprays the ball oppo would have a high shift score not a low one like you say?

That should be “line drives and home runs” not “fly balls and home runs.”

Thank you for superb article, so what was the shift score fot Yasiel Puig, if possible, splitting two seasons so far ?

I will post it once I get back to town, thanks.

at the first line, by “the percentage of ground ball hits”, I meant “ground ball hits/ground ball”.

Looking at the aggregate, I don’t see any changes to league wide BABIP. The shift is giving as much as it is taking from my view.

As for individual hitters, I think the assumption is very nebulous that a hitter won’t alter his approach to the shift. Obviously, for hitters that don’t and who fit the necessary profile, the shift will be effective. For those who do, the shift could actually be counter productive.

See MGL’s comment above. It is very hard for hitter’s to try and hit ground balls in the first place, so going the opposite way would actually be harder. I think the assumption is fine, given that those who are shifted in the first place are those who cannot change their approach. The hard part is identifying those who cannot change which seems to be 90-95% of MLB.