Written By: Forrest Allen

Shifting defenders certainly isn’t a new concept, but before now there was a lack in precision, making it difficult for coaches to make informed decisions about its potential risks and benefits. Armed with mountains of data, decades of coaching and playing experience, and innovative data science techniques, 6-4-3 Charts set out to create a tool for coaches to experiment with how defensive configurations can impact outcomes. We wanted to quantify the impact of multiple defensive configurations, each uniquely tailored to an individual hitter’s batted ball profile. The result is an interactive tab within the 6-4-3 Charts’ TrackMan SYNC tile that allows users to see a hitter’s expected batting average (xBA), slugging percentage (xSLG), and weighted on base average on contact (xwOBAcon) versus multiple defensive configurations. In addition to preset alignments, users also can manually move infielders and outfielders while the system dynamically recomputes the expected outcomes. This functionality allows users to assess which configurations most meet their defensive objectives.

*The Defense Tab, located in the TM SYNC Tile on the 643 Charts Interface*

**A Tour of the Tool**

This tool lives within the TrackMan SYNC Tile under the Defense Tab. The Defense Tab enables users to interact with the defensive shift model. A picture of the infield appears on the top half of the page. There a user sees measurements of depth in feet from home plate along the foul lines, as well as measurements of horizontal distance. The distance across the field is shown as angles relative to a straight line extending from home plate to the centerfield wall. These precise measurements help bring the model from the screen to the field. Coaches can use the precise measurements to translate the configurations to their own communication style.

*The infield viewer in the Defense Tab*

The smaller red, white, and blue circles represent all the recorded batted ball events (BBE) from the selected hitter. Their position on the field represents their first contact with the field and the color shade shows the likely outcome. Blue represents likely outs, while red indicates likely hits. Finally, the larger orange circles represent where the defenders are positioned in the selected defensive configuration.

The bottom half of the screen shows a table with 13 commonly deployed defensive configurations. Each configuration has an associated xBA, xSLG, and xwOBAcon for the selected hitter. In addition, it shows exactly where the fielder is positioned in that configuration. The table defaults to show the most effective alignment (lowest xBA) at the top, but the table can be sorted by any of the columns dynamically. For portability into the dugout, the table can also be exported to PDF, CSV, or copied to clipboard.

*Sample table of defensive configurations for infield*

*Manual Adjustments*

While the present configurations give users a starting point, users can further refine positioning by using the “left side adjustments” and “right side adjustments” buttons on either side of the visualization. The left side adjustments allow users to move the third baseman and shortstop in and back (Depth Adjustment) and left & right (Bearing) a foot and a degree at a time respectively. The right side adjustments provide the same functionality for the second and first basemen. The system responds dynamically by updating the expected statistics underneath the field diagram for the customized alignment set by the user.

*Infield viewer with left & right side adjustments*

The outfield viewer uses the same format as the infield viewer, displaying the field, adjustments, and table of configurations.

*Outfield viewer & table of configurations*

**Methodology**

The first step was to assign a primary defender to every batted ball event (BBE). Popups, bunts, and homeruns were excluded since popups can be easily defended irrespective of positioning and no configuration can defend homeruns. Defending bunts is outside the scope of this model. All groundballs, defined as a BBE with a launch angle less than 12 degrees, were assigned to the closest infielder. Flyballs traveling at least 200 feet were assigned to the closest outfielder.

**Infielders**

To calculate our expected statistics for the infield, we examined two features of the groundball BBEs:

- Exit velocity
- Primary fielder’s distance from the BBE

Exit velocity is simply the speed of the ball off the bat. To calculate distance, we start with the position of each infielder at the time of contact and the spray angle (also known as bearing) of the BBE. By subtracting the fielder’s bearing and that of the BBE bearing, we determine how far the distance between the primary fielder was from the ball in terms of degrees.

Just considering exit velocity OR distance from the defender isn’t enough to accurately assess whether a given BBE will result in a hit. However, combining the attributes *togethe*r results in much more comprehensive results. For example, a 110-mph shot right at a defender and a weak ground ball near a defender are both quite likely to results in outs, despite being very different BBE profiles. Conversely, a 110-mph shot in the gap is quite likely to result in a hit.

*Expected stats for exit velocity/distance combinations*

Once assigned an exit velocity/distance combination, we count the number of hits this combination produced and divide them by the number of total occurrences. This allows us to derive an xBA for each EV/Distance combination. Despite pitchers and catchers being able to defend a subset of groundballs, we have excluded them from this analysis as they cannot be repositioned to reduce the probability of a hit. xSLG multiplies the hit by total bases of the hit while xwOBA multiplies the hits by wOBA coefficients before dividing by occurrences for those two statistics. Below is a heat map showing the results for 3B; the xBAs have been formatted for blue values to represent the lowest and the red values show the highest.

*Sample heatmap of xBAs for 3B*

The results align with our intuition. The highest exit velocities result in the highest frequency of hits, as shown as you move from the top to the bottom. Batted balls farther from fielders produce high expected batting averages and decrease as the distance shrinks between BBEs and the fielder. We repeat this same concept for all other infield positions.

Since we have the fielding profiles for all four infielders, we know the probability of each defender recording an out based on the exit velocity and distance away from each of them. As such, we take a weighted average of xBAs from all four infielders by using the number of BBEs to each position as weights.

*Visualizing an Example BBE to Calculate xBA*

To demonstrate the process, below is a calculation of xBA for a BBE with an exit velocity of 84 mph and a bearing of -17 degrees hit against a straight up defensive alignment.

*Sample BBE within the ‘straight up’ alignment*

In this formation, the third baseman’s location is -32, shortstop is at -15, second baseman is at 15 and the first baseman is at 32 as denoted by the orange circles. The distance is the difference between the BBE bearing and each fielder’s bearing. Using these data points, we know how far the ball is from each fielder and how hard the BBE was hit and can look up the xBA in each infielder’s profile.

Position | Exit velocity classification | Fielder location | Batted ball bearing | Distance from ball (in degrees) | xBA | BBEs |

3B | Routine | -32 | -17 | 15 | .7744 | 0 |

SS | Routine | -15 | -17 | 2 | .116 | 55 |

2B | Routine | 15 | -17 | -32 | 1.000 | 1 |

1B | Routine | 32 | -17 | 49 | 1.000 | 0 |

The weighted average of an 84 mph BBE with a bearing of -17 degrees against a straight up configuration will result in an xBA of .131. However, this result changes drastically in a different configuration. Should the infield defense line up with the third baseman and shortstop shifted toward the left field line and the second baseman playing behind second base, the xBA increases .505 points to .636. As the table below shows, these distances place defenders further from the BBE.

Position | Exit velocity classification | Fielder location | Batted ball bearing | Distance from ball (in degrees) | xBA | BBEs |

3b | Routine | -34 | -17 | 19 | .666 | 0 |

SS | Routine | -23 | -17 | 6 | .625 | 32 |

2b | Routine | 3 | -17 | -20 | 1.000 | 1 |

1b | Routine | 22 | -17 | -39 | 1.000 | 0 |

Knowing what the average fielder is capable of effectively defending, the next step is to bring in a specific hitter’s profile to determine the optimal defensive configuration. Specifically, we bring in all BBE for ground balls and, using the methodology described above, the system provides xBA, xSLG, and xwOBA for the 13 pre-configured shifts. Should a user want to explore other options, they can use the columns on sides to move each defender as the system dynamically updates the expected statistics.

**Outfielders **

Finding xBA for BBE to the outfield is like infield in the sense that we need to consider how long the fielder has to react and the distance required cover to make the play. To calculate our expected statistics for the outfield, we examined two features of the flyball BBEs:

- Reaction Time Required
- Distance to Cover (horizontal and vertical)

*Reaction time required*

To analyze how long the player has to react on the infield, we used exit velocity and bearing, but they are suboptimal for outfield. For example, the probability of a hit for a BBE that was 110-mph off the bat is quite different if it stays in the air for 5 seconds or 2 seconds. Accordingly, hangtime, the number of seconds the ball is in the air, is the better attribute to consider when assessing hit probability of flyballs because it literally measures how long a defender has to cover the required distance to make the play.

*Distance to cover*

We calculate both horizontal distance (between the foul lines) and vertical distance (between the outfield wall and home plate) a fielder must cover to catch the ball since we know both where the fielder starts and where the ball lands. But not all distances are created equal. We distinguish between distance types since catching a BBE 20 feet in front of a defender presents a much different level of difficulty than one 20 feet over a defender’s head.

*Creating the model*

To assess hit probability for the outfield, the model considers the following factors: hangtime, vertical distance to cover, and horizontal distance to cover. Even with a sample approaching 6000, there are too many different combinations of these factors to simply divide hits by occurrences the way we did for infield. Rather, we use a machine learning predictive model for each position as the defending talents are not uniform across the positions.

The model learns which combinations are most likely to result in hits and outs, thus able to generate a probability of a hit for any combination of hangtime, vertical and horizontal distance. We exposed the model to almost 4500 flyballs showing the result (out or hit), hangtime, vertical and horizontal distance. We then evaluated its performance against almost 1500 BBE the model had not seen before, but we knew the outcome of.

*xSLG and xwOBAcon*

While the outfield model type is convenient to predict xBA, it poses a challenge in creating xSLG and xwOBAcon. The model doesn’t distinguish between hit types which is required to predict SLG and wOBAcon. To remedy this, we consider two additional features of the BBE, launch angle and exit velocity. Using the data we have on launch angle, exit velocity, and distribution of hit types, we can calculate the weighted average number of bases for a given type of BBE. To calculate xwOBAcon, we simply substitute the wOBA weights for the total bases gained.

**Conclusion**

When the more analytically-inclined MLB teams started shifting in the 2010s, the initial debate centered around whether shifting would work. The argument went that good hitters could just push the ball to a virtually vacant side of the infield and guarantee themselves a hit. But by the end of the decade, the debate…no, please don’t say it…*shifted…* to how we could reign in shifting due to its frequency and success suppressing offense. While MLB outlawed extreme shifts in 2023, there is no current restrictions at the college level. Having this tool allows coaches to explore how shifting defenders fairs against each hitter their team may face.