1 Background

Major League Baseball (MLB) is the premier professional baseball competition in North America, featuring 30 teams across the United States and Canada. The season ends with the MLB Postseason, where the top 12 teams compete across four rounds: the Wild Card Series, Division Series, League Championship Series, and World Series. The 2023 Postseason saw the Texas Rangers defeat the Arizona Diamondbacks in five games to claim the World Series title.

The data used in this report was sourced from Baseball Savant, the official Statcast data platform operated by the MLB. Statcast is a state-of-the-art tracking system installed across all 30 MLB stadiums, using a combination of high-speed cameras and radar to capture detailed data on every pitch, hit, and player movement during games (1). The system records a wide range of measurements including pitch velocity, spin rate, launch angle, and exit velocity (1). This enables teams, analysts, and fans to quantify player performance. This report focuses specifically on pitch-by-pitch tracking data, capturing detailed information on every pitch thrown throughout the 2023 MLB Postseason.

2 Data Description

The dataset contains 11,828 pitch-by-pitch observations across 41 games, spanning four rounds of the 2023 MLB Postseason. A total of 136 unique pitchers and 148 batters are represented, with 12 teams competing in the postseason. The pitching cohort was predominantly right-handed, with 98 right-handed pitchers (RHP) accounting for 75.6% of all pitches thrown. The remaining 24.4% were thrown by 38 left-handed pitchers (LHP). A similar trend was observed among batters, with 98 right-handed batters (RHB) facing 56.9% of all pitches, and 63 left-handed batters (LHB) facing the remaining 43.1%. Table 1 provides a high level summary of the dataset.

2.1 Dataset Overview

Table 1: Dataset Overview
Metric Value
Total Pitches 11,828
Total Games 41
Total Rounds 4
Total Teams 12
Unique Pitchers 136 (98 RHP, 38 LHP)
Unique Batters 148 (98 RHB, 63 LHB)
Pitches by RHP 8,946 (75.6%)
Pitches by LHP 2,882 (24.4%)
Pitches to RHB 6,734 (56.9%)
Pitches to LHB 5,094 (43.1%)
Table 2: Pitches per Round
Round Games Pitches
Wild Card 6 2,281
Division Series 15 4,116
League Championship 14 3,942
World Series 6 1,490
Table 3: Pitch Type Distribution
Pitch Count Proportion
4-Seam Fastball 4343 36.7%
Sinker 1645 13.9%
Slider 1553 13.1%
Changeup 1114 9.4%
Curveball 1088 9.2%
Cutter 754 6.4%
Sweeper 600 5.1%
Split-Finger 373 3.2%
Knuckle Curve 337 2.8%
Slurve 21 0.2%

2.2 Data Processing

Prior to analysis, several data processing steps were undertaken to clean and prepare the dataset. The raw dataset contained 92 variables, which was reduced to 75 following the removal of eight fully deprecated columns containing no data, two duplicate columns, seven fielder identification columns, and additional redundant variables including game year and fielding alignment. Additionally, a single “Pitch Out” observation was also removed as it did not represent a genuine attempt to retire the batter, resulting in a final dataset of 11,828 pitches.

Several variables were recoded and renamed to improve clarity. Base runner columns (on_1b, on_2b, on_3b) were recoded from NA values to a binary format, where 0 indicates an unoccupied base and 1 indicates an occupied base. Game type codes were recoded to their full round names, and several column names were updated to better reflect their content, including renaming stand to batter_stance, p_throws to pitcher_hand, and type to pitch_outcome. Additionally, the outcomes of pitches were recoded from single letter codes to descriptive labels (Ball, Strike, In Play). The pitch_type column was removed as it was redundant with the more descriptive pitch_name variable.

New variables were created to extend the dataset for analysis. A pitch count variable was created by combining the balls and strikes columns into a single string (e.g. 3-1), providing a clearer representation of the count situation on each pitch. A runners on base variable was derived by summing the three binary base runner columns, producing a value between 0 and 3 representing the total number of runners on base. Finally, a score differential variable was calculated by subtracting the fielding team’s score from the batting team’s score, providing a measure of game state at the time of each pitch. Additionally, pitches were grouped into three categories: Fastball, Breaking Ball, and Offspeed, based on their movement and velocity characteristics as per Lebovitch (2).

2.3 Missing Values and Outliers

The dataset was assessed for missing, erroneous, and incomplete values, each requiring a different treatment approach. Eight columns were found to be entirely missing across all observations, suggesting Statcast variables that were no longer in use, all were removed from the dataset.

A second category of missing values was identified in batted ball metrics such as launch speed, launch angle, and hit distance. These missing values were expected by design as these variables are only populated when the ball is hit into play. Similarly, the events and bb_type columns contained empty strings for pitches that did not end a plate appearance or result in contact.

Three isolated missing values were identified: effective_speed for a Sinker thrown by Brandon Pfaadt, release_extension for a Sinker thrown by Orion Kerkering where effective_speed was also recorded as 0, and delta_run_exp for a single pitch. These were determined to be isolated Statcast tracking anomalies with no impact on the analysis and remained in the dataset.

One unusual value was identified in the release_spin_rate variable, where a Slider thrown by Corey Seager was recorded with a spin rate of 790 rpm, substantially below the expected range for this pitch type. This value is likely a Statcast tracking error and while it was retained in the dataset, it is noted as an outlier that may influence spin rate analysis for the Slider pitch type.

2.4 Variable Definitions

The dataset contains 75 variables capturing a wide range of information on each pitch, including pitcher and batter traits, pitch mechanics, location, and game context. Key variables relevant to this analysis are defined in Table 4 below.

Table 4: Key Variable Definitions
Variable Description Type
pitch_name The name of the pitch type thrown (e.g. 4-Seam Fastball, Slider) Categorical
pitch_category The broader category of the pitch (Fastball, Breaking Ball, Offspeed) Categorical
release_speed The velocity of the pitch at the point of release, measured in miles per hour Continuous
release_spin_rate The rate at which the ball spins after release, measured in revolutions per minute Continuous
plate_x The horizontal location of the pitch as it crosses home plate, measured in feet Continuous
plate_z The vertical location of the pitch as it crosses home plate, measured in feet Continuous
pitch_outcome The outcome of the pitch (Ball, Strike, In Play) Categorical
events The outcome of the plate appearance (e.g. strikeout, home run, field out) Categorical
batter_stance The batting stance of the batter (L = Left, R = Right) Categorical
pitcher_hand The throwing hand of the pitcher (L = Left, R = Right) Categorical
game_type The round of the postseason (Wild Card, Division Series, League Championship, World Series) Categorical
count The ball-strike count at the time of the pitch (e.g. 3-2) Categorical
runners_on_base The total number of runners on base at the time of the pitch (0-3) Numeric
score_differential The score difference between the batting and fielding team at the time of the pitch Numeric

2.5 Data Visualisations

Figure 1 displays the frequency of each pitch type thrown across the 2023 MLB Postseason.

The 4-Seam Fastball was the most frequently thrown pitch in the 2023 MLB Postseason, accounting for 4,343 pitches (36.7% of all pitches thrown). This is consistent with baseball norms, as the 4-Seam Fastball is the main pitch for most pitchers and is typically used to establish velocity and set up secondary pitches.

The Sinker (1,645) and Slider (1,553) were the next most common pitch types, reflecting the emphasis on generating weak contact and swing-and-misses.

The Curveball (1,088) was also well represented, as one of the most recognisable and widely used breaking balls in baseball, known for its sharp downward movement. At the other end of the spectrum, the Slurve (21) was rarely used, likely reflecting that it is a speciality pitch limited to a small number of pitchers in the dataset.

Figure 2 displays the distribution of release speed for each pitch type, coloured by pitch category.

Figure 2 illustrates a clear velocity hierarchy across pitch categories, with Fastballs sitting at the top of the speed range and Breaking Balls at the bottom.

The 4-Seam Fastball and Sinker lead in median release speed, both averaging around 93-95 mph, consistent with their role as power pitches designed to overpower batters.

Offspeed pitches, including the Changeup and Split-Finger, sit in the mid-80s. Their effectiveness relies not on raw velocity but on slight speed differentials from fastballs, disrupting a batter’s timing.

Breaking balls are the slowest category, with the Curveball and Knuckle Curve dropping into the low-80s and below, relying instead on spin and movement to deceive batters.

Outliers are visible on the slower end of the Cutter and the faster end of the Split-Finger, which may reflect individual pitcher styles.

Figure 3 displays the distribution of spin rate for each pitch type, coloured by pitch category.

Figure 3 reveals an interesting contrast to the velocity distribution observed in Figure 2, with Breaking Balls now leading in spin rate despite being the slowest pitch category.

The Curveball and Knuckle Curve sit at the top of the spin rate distribution, with median spin rates exceeding 2,500 rpm. This high spin is what generates their sharp movements, making it difficult for batters to make solid contact.

Offspeed pitches have the lowest spin rates across all categories. Reduced spin causes the ball to tumble and drop naturally due to gravity, creating a distinct movement profile that differs from both fastballs and breaking balls.

One notable outlier is visible in the Slider distribution, consistent with the Statcast tracking anomaly identified in the data quality assessment.

Figure 4 displays the relationship between release speed and spin rate for each pitch, coloured by pitch category.

Figure 4 reveals three broadly distinct clusters corresponding to each pitch category, reinforcing the velocity and spin rate patterns observed individually in Figures 2 and 3. Fastballs form a tight cluster in the high velocity, moderate spin rate region, generally sitting above 90 mph with spin rates between 2,000 and 2,500 rpm.

Breaking balls occupy the lower velocity range but spread across a wide range of spin rates, reflecting the variety of movement profiles within this category. Offspeed pitches form the most compact cluster, sitting in the mid-80s for velocity and the lower end of the spin rate range, consistent with their design to mimic fastball trajectory while dropping out of the zone.

The scatter plot also highlights the overlap between categories, particularly between Offspeed and Breaking Ball pitches in the lower velocity range, suggesting that velocity and spin rate alone do not perfectly distinguish between all pitch types and that movement profile and grip are equally important factors.

Figure 5 displays the density of all pitch locations across the 2023 MLB Postseason, with a strike zone overlaid.

Figure 5 reveals that pitches were most densely concentrated around the lower half of the strike zone, particularly just below and within the bottom portion of the zone. This is because pitches in the lower zone are more likely to induce ground balls and weak contact, while also being harder for batters to elevate the ball. The spread of pitches extends well outside the strike zone on both sides horizontally, which reflects the deliberate use of balls to set up certain throws.

Figure 6 displays the density of pitch locations broken down by pitch type.

Figure 6 reveals distinct location tendencies across pitch types, reflecting the different roles each pitch plays within a pitching arsenal. The 4-Seam Fastball shows concentration in the upper half of the strike zone, consistent with its use as a power pitch designed to overpower batters.

The Sinker sits notably lower, reflecting its design to generate ground ball contact.

The Cutter and Slider tend to cluster toward the outer edges of the zone, as these pitches are commonly used to attack the outside and generate weak contact.

Curveballs, Knuckle Curves, Slurves, and Split-Fingers show the lowest vertical location of all pitch types, often landing well below the strike zone, reflecting their use as chase pitches designed to get batters to swing out of the zone.

Figure 7 displays the density of pitch locations to left and right-handed batters.

Figure 7 reveals that pitch location patterns differ subtly between left and right-handed batters. Against right-handed batters, pitches are concentrated toward the lower and outer portion of the zone, a common strategy to generate weak contact or induce chases.

Against left-handed batters, a similar low zone concentration is observed, however the horizontal density shifts slightly toward the inner half. This may partly reflect that left-handed batters are less common, meaning some pitchers may default toward inner half locations more typical of their approach against right-handed batters, rather than fully adjusting their targeting.

Figure 8 displays the density of pitch location by pitcher handedness.

Figure 8 reveals subtle but notable differences in pitch location between left and right-handed pitchers. Both groups show the greatest density in the lower/middle portion of the strike zone, consistent with the overall pitching tendencies observed in Figure 5.

Right-handed pitchers display a much more concise and concentrated location profile, with density tightly clustered in the strike zone.

Left-handed pitchers show a more spread out distribution, with pitches extending further low and to the left of the zone, which is the arm side for a left-handed pitcher. This wider spread may reflect the natural movement tendencies of left-handed pitchers, whose pitches naturally tail away in that direction due to their arm angle and release point.

Figure 9 displays the pitch type usage by batter stance.

Figure 9 reveals that pitch type usage was largely consistent between left and right-handed batters, with Fastballs dominating in both cases at around 57.9% and 56.3% respectively. However, a notable difference emerges in Breaking Ball and Offspeed usage.

Pitchers threw a considerably higher proportion of Breaking Balls to right-handed batters (32.9%) compared to left-handed batters (27.1%), while Offspeed pitches were used more frequently against left-handed batters (15.0%) than right-handed batters (10.8%).

This suggests that pitchers adjusted their secondary pitch selection based on batter stance, likely reflecting the different movement profiles of breaking balls and offspeed pitches relative to each side of the plate and the handedness matchup between pitcher and batter.

Figure 10 displays pitch type usage for left-handed pitchers.

Figure 10 reveals that left-handed pitchers leaned heavily on the 4-Seam Fastball as their primary pitch at 28.1%, followed closely by the Sinker at 24.4%, indicating a strong preference for fastball variants among left-handers.

The Curveball was the most used breaking ball at 15.1%, with the Changeup and Slider both sitting at 12.2%.

The Cutter was used more sparingly at 5.9%, while the Sweeper (1.8%) and Split-Finger (0.3%) were rarely used, reflecting that these are specialty pitches limited to only a small number of left-handed pitchers in the dataset.

Figure 11 displays pitch usage for right-handed pitchers.

Figure 11 reveals that right-handed pitchers relied most heavily on the 4-Seam Fastball as their primary pitch at 39.5%, a notably higher proportion than left-handed pitchers at 28.1%.

The Slider was the most common secondary pitch at 13.4%, followed by the Sinker at 10.5% and Changeup at 8.5%.

Compared to left-handers, right-handed pitchers showed a more diverse breaking ball arsenal, with the Slider, Curveball (7.3%), Sweeper (6.1%), and Knuckle Curve (3.8%) all featuring meaningfully. Notably, right-handed pitchers also made use of the Knuckle Curve and Split-Finger, which were either absent or negligible in the left-handed pitcher arsenal. The Slurve remained a rare offering at just 0.2% for both groups.

Figure 12 displays pitch outcomes by pitch type across the 2023 MLB Postseason.

Figure 12 reveals notable differences in outcome profiles across pitch types. Among the more frequently thrown pitches, the 4-Seam Fastball generated the highest proportion of strikes, consistent with its role as the primary pitch used to attack batters and get ahead in the count.

The Sinker produced a notably higher proportion of in play outcomes compared to other pitch types, which aligns with its design to generate ground ball contact rather than swings and misses.

The Changeup recorded the highest ball proportion of all pitches, reflecting its use as a pitch thrown deliberately off the plate to disrupt batter timing and induce chases.

The Slurve shows the highest strike proportion of all pitch types, however given it was thrown only 21 times this should be interpreted with caution as it is not a reliable sample.

Figure 13 displays pitch type usage by round across the 2023 MLB Postseason.

Figure 13 reveals that pitch type usage was remarkably consistent across all four rounds of the postseason, with Fastballs dominating in every round ranging narrowly from 56.6% in the League Championship to 57.6% in the World Series.

Breaking Balls were the second most used category in every round, peaking in the League Championship at 31.9% and dropping to their lowest in the World Series at 27.5%.

Offspeed pitches were the least used category throughout, however they saw their highest usage in the World Series at 14.9%, coinciding with the lowest Breaking Ball usage of any round.

It is worth noting that as teams are eliminated across rounds the pool of pitchers changes, meaning the pitch arsenal available in later rounds differs from earlier rounds, which may influence the proportions observed rather than reflecting a deliberate strategic shift.

Figure 14 displays home runs hit by pitch type.

Figure 14 reveals that the 4-Seam Fastball gave up the most home runs in raw count terms at 48, well ahead of the Slider (17) and Changeup (16). However, raw counts alone are misleading given each pitch type was thrown a different number of times.

Table 5 below presents the home run rate per pitch type to provide a more meaningful comparison.

Table 5: Home Run Rate by Pitch Type
Pitch Name Pitch Category Total Pitches Home Runs Home Run Rate (%)
Changeup Offspeed 1114 16 1.44
Cutter Fastball 754 9 1.19
4-Seam Fastball Fastball 4343 48 1.11
Slider Breaking Ball 1553 17 1.09
Sweeper Breaking Ball 600 6 1.00
Knuckle Curve Breaking Ball 337 2 0.59
Split-Finger Offspeed 373 2 0.54
Sinker Fastball 1645 8 0.49
Curveball Breaking Ball 1088 2 0.18
Slurve Breaking Ball 21 0 0.00

When accounting for usage, the Changeup had the highest home run rate at 1.44%, followed by the Cutter (1.19%), with the 4-Seam Fastball dropping to third at 1.11%. This suggests that while the 4-Seam Fastball gave up the most home runs in total, it was not the most home run prone pitch relative to how often it was thrown.

The Sinker and Curveball were the most effective at suppressing home runs, with the Sinker giving up just 8 home runs from 1,645 pitches and the Curveball only 2 from 1,088, reflecting how their in air movement limits a batter’s ability to get under the ball.

Figure 15 displays strikeouts by pitch type.

Figure 15 reveals that the 4-Seam Fastball recorded the most strikeouts in raw count terms at 234, followed by the Slider (101) and Curveball (100). However, as with home runs, raw counts alone are misleading given each pitch type was thrown a different number of times.

Table 6 below presents the strikeout rate per pitch type to provide a more meaningful comparison.

Table 6: Strikeout Rate by Pitch Type
Pitch Name Pitch Category Total Pitches Strikeouts Strikeout Rate (%)
Sweeper Breaking Ball 600 61 10.17
Knuckle Curve Breaking Ball 337 34 10.09
Slurve Breaking Ball 21 2 9.52
Curveball Breaking Ball 1088 100 9.19
Split-Finger Offspeed 373 33 8.85
Slider Breaking Ball 1553 101 6.50
4-Seam Fastball Fastball 4343 234 5.39
Changeup Offspeed 1114 60 5.39
Cutter Fastball 754 35 4.64
Sinker Fastball 1645 72 4.38

When accounting for usage, breaking balls were the most effective pitch category for generating strikeouts, with the Sweeper (10.17%), Knuckle Curve (10.09%), and Curveball (9.19%) all ranking among the top strikeout pitches. The Split-Finger (8.85%) was the standout Offspeed pitch for strikeouts.

Fastballs were the least effective category, with the Sinker posting the lowest strikeout rate of all pitch types at 4.38%, the Cutter at 4.64%, and the 4-Seam Fastball and Changeup both sitting at 5.39%.

Interestingly, the Sinker presents a contrasting profile when considered alongside the home run rate data from Table 5. The Sinker gives up very few home runs (0.49%), but also takes the least strikeouts. This reflects its role as a set-up pitch or a pitch designed to generate ground ball contact rather than swing and miss outcomes.

3 Data Use Cases

The following section proposes two data use cases that leverage the 2023 MLB Postseason pitch-by-pitch dataset. Each use case identifies a practical application of the data for teams, players, or coaching staff, supported by preliminary visualisations and statistics. For each use case, a written description, hypothesis, and discussion of limitations are provided.

3.1 Use Case 1: Pitch Selection and Effectiveness on a Full Count

3.1.1 Description

A full count, where the batter has three balls and two strikes, represents one of the most high-pressure situations in baseball for both the pitcher and batter.

For the pitcher, throwing a ball results in a walk, awarding the batter first base, while throwing a strike results in a strikeout. This pressure creates an interesting strategic dilemma. Does the pitcher play it safe with a fastball, relying on its speed and control to guarantee a strike, or do they gamble with a breaking ball or offspeed pitch in the hope of inducing a swing and miss?

From the batter’s perspective, a full count also presents a dilemma. Do they sit back and take a pitch hoping to draw a walk, or do they swing aggressively knowing the pitcher is under equal pressure to throw a strike?

This use case explores pitch selection and effectiveness on a full count from both the pitcher and batter perspective, with the aim of informing pitching staff on the most effective pitch to throw in this situation, and providing batting teams with insight into what they might expect to face.

3.1.2 Visualisations

Figure 16 examines how pitch type usage on a full count compares to all other counts, to determine whether pitchers adjust their selection under full count pressure.

Figure 16 reveals that pitch selection shifts notably on a full count compared to all other counts. Fastball usage increases from 56.8% to 61.8% on a full count, suggesting that pitchers do indeed favour the safety and control of a fastball when the risk of walking the batter is at its highest.

Offspeed usage remains relatively stable, increasing slightly from 12.5% to 13.1%. Most notably, Breaking Ball usage drops from 30.7% to 25.1% on a full count, suggesting that pitchers are less willing to gamble with a breaking ball that carries a higher risk of missing the zone and awarding the batter a walk.

To further explore this, Figure 17 examines the outcomes of each pitch category specifically on a full count, to assess which pitch type is most effective in this situation.

Figure 17 reveals that despite fastballs being the most frequently thrown pitch on a full count, they actually produced the lowest strike rate of the three categories at 42.9%, while also producing the highest in play rate at 33.2%.

Offspeed pitches produced the highest strike rate at 50.0%, closely followed by Breaking Balls at 49.3%.

Notably, Breaking Balls produced the lowest ball rate at 22.5%, slightly better than Fastballs at 23.8%, which challenges the assumption that pitchers avoid breaking balls on a full count due to control concerns.

Offspeed pitches produced the highest ball rate at 30.6%, suggesting they carry the most risk of walking the batter on a full count.

3.1.3 Hypothesis

It is hypothesised that pitchers would favour fastballs on a full count due to the pressure of avoiding a walk, and the data supports this.

Fastball usage increases from 56.8% to 61.8% on a full count while breaking ball usage dropped notably from 30.7% to 25.1%. However, the outcome data reveals an interesting finding that challenges conventional pitching.

Despite being thrown less frequently, breaking balls and offspeed pitches both generated higher strike rates (49.3% and 50.0% respectively) than fastballs (42.9%) on a full count. This suggests that while pitchers default to fastballs for safety, breaking balls and offspeed pitches may actually be more effective at retiring the batter in a full count situation.

However, offspeed pitches also produced the highest ball rate at 30.6%, meaning the risk of walking the batter is at its greatest with this pitch type.

Ultimately, the most effective pitch on a full count may simply be the pitch the pitcher executes best. A pitcher with an elite curveball may be better served throwing it on a full count than defaulting to a fastball, suggesting that individual pitcher strengths should be considered alongside these broader trends when making pitching decisions.

From the batter’s perspective, sitting on a fastball on a full count is a statistically sound decision given that 61.8% of pitches in this situation are fastballs. However, a batter who commits too heavily to expecting a fastball risks being completely fooled by a breaking ball or offspeed pitch, meaning the same uncertainty that faces the pitcher also applies to the batter.

3.1.4 Limitations

Several limitations should be considered when interpreting the findings of this use case.

First, the dataset covers only the 2023 MLB Postseason, meaning the sample of full count pitches is relatively small and may not be representative of broader pitching tendencies across a full regular season.

Postseason baseball also features the best pitchers and batters in the league, meaning these findings may not generalise to regular season play where the overall quality of pitching and hitting varies more widely.

Second, the analysis groups pitches into broad categories of Fastball, Breaking Ball, and Offspeed, which masks individual pitcher differences. As noted in the hypothesis, a pitcher who excels at a particular pitch may deviate significantly from the broader trends observed here, meaning these findings should be interpreted as general tendencies rather than prescriptive advice.

Finally, batter quality and matchup context are not accounted for in this analysis. A pitcher may adjust their full count pitch selection depending on the strength of the batter, the game situation, or the score differential, all of which are factors that would need to be considered in a more detailed analysis.

3.2 Use Case 2: Pitcher Fatigue and Workload Management

3.2.1 Description

Pitcher fatigue is one of the most critical factors in pitching performance and a key consideration for coaching staff when making in-game decisions.

As a pitcher throws more pitches throughout a game, their physical condition deteriorates. This may result in diminished velocity and control, or a shift in pitch selection.

Identifying the point at which a pitcher’s performance begins to decline is a valuable tool for coaching staff, as pulling a pitcher too early wastes a valuable resource, while leaving them in too long risks giving up runs.

This use case explores how pitcher performance changes as pitch count increases. Specifically it will examine whether release velocity declines, whether pitch selection shifts towards breaking balls as fatigue sets in, and whether spin rate on breaking balls changes alongside velocity drops.

The practical aim of this use case is to provide coaching staff with a data driven framework for identifying fatigue thresholds in individual pitchers, informing more effective pitching change decisions during games.

3.2.2 Visualisations

Figure 18 examines how release speed changes across pitch count for the five highest workload pitchers in their longest individual postseason outing, providing an indication of how velocity is managed throughout a game.

Figure 18 displays the smoothed release speed across pitch count for each of the five highest workload pitchers in their longest individual postseason outing.

The patterns vary considerably across pitchers, suggesting that fatigue effects differently depending on the individual.

Eovaldi starts with the highest velocity but shows an early decline until around pitch 40 before recovering, never quite returning to his opening speed. Gallen shows the most consistent upward trend, rising steadily until pitch 80 before a slight decline in his final pitches, suggesting strong endurance through the bulk of his outing.

Kelly follows a similar pattern of early decline before recovering through the middle portion of his outing, dropping off again in his final pitches. Montgomery starts with the lowest velocity but rises sharply until pitch 40, before a notable dip around pitch 65 and then a strong recovery to finish around 95 mph.

Wheeler is the most stable of the five, maintaining consistent velocity before rising sharply in the final portion of his outing from pitch 80 onwards, suggesting he may have been conserving energy early and elevating late.

Overall, the varying patterns across pitchers highlight that there is no universal fatigue threshold that applies to all pitchers. This reinforces the core argument of this use case, that coaching staff need to monitor individual pitcher velocity trends in real time rather than applying a blanket pitch count limit, as each pitcher manages their effort and endurance differently throughout a game.

It is also worth noting that pitch counts of 88-100 per game are considerably higher than what is typically seen in regular season baseball, where teams more readily utilise bullpen swaps to manage pitcher workload.

This limits the generalisability of these findings to regular season play, however it makes this analysis particularly valuable for postseason pitching strategy where starters are often pushed deeper into games. Future analysis incorporating regular season data would allow for a more comprehensive understanding of pitcher fatigue across different contexts.

Figure 19 examines how pitch type usage shifts across pitch count bands for the same five pitchers in their longest individual postseason outing, to determine whether pitch selection changes as fatigue sets in.

Figure 19 reveals distinct and highly individual pitch selection patterns across the five pitchers, with no consistent trend of increasing breaking ball usage as pitch count increases.

Wheeler displays the most consistent and predictable arsenal throughout his outing, relying almost exclusively on fastballs with a steady proportion of breaking balls across all five bands and throwing virtually no offspeed pitches. While this consistency may reflect strong command and confidence in his primary pitches, it could also make him more predictable to opposing batters.

Montgomery remains relatively consistent throughout, maintaining a similar fastball and breaking ball split across most bands before shifting slightly toward more fastballs and offspeed pitches in his final band.

Kelly shows the most erratic pattern, with large swings in pitch category usage between bands, which may reflect deliberate in game adjustments to keep batters off balance rather than fatigue driven changes.

Gallen presents perhaps the most interesting profile, beginning with an unusually high proportion of breaking balls before gradually transitioning toward offspeed pitches mid game and finishing with almost exclusively fastballs in his final band, suggesting a deliberate sequencing strategy rather than a fatigue response.

Eovaldi stands out for his heavy early reliance on offspeed pitches, maintaining an almost equal split between fastballs and offspeed in his opening bands before introducing more breaking balls mid game and returning to fastballs late.

Overall, the highly individualised patterns observed across these five pitchers suggest that pitch selection changes throughout a game are driven more by individual strategy and in game adjustments than by fatigue alone, which has important implications for how coaching staff interpret and respond to pitcher workload.

Figure 20 examines whether spin rate on breaking balls changes across pitch count for the same five pitchers in their longest individual postseason outing, to determine whether pitchers compensate for velocity loss by generating higher spin on their breaking balls as fatigue sets in.

Figure 20 reveals highly individualised spin rate patterns across the five pitchers, with no consistent evidence of pitchers compensating for velocity loss by generating higher spin on their breaking balls as fatigue sets in.

Montgomery and Gallen display the most stable spin rates throughout their outings, with Gallen showing a slight rise toward the end and Montgomery experiencing a minor drop in his final pitches, suggesting strong mechanical consistency on their breaking balls deep into games.

Wheeler starts as the second highest and shows a relatively stable profile with a slight upward trend after pitch 40, maintaining his spin rate well throughout.

Kelly starts with the highest spin rate of the five but shows considerable variation throughout, declining early before recovering mid game and dropping off again after pitch 50, suggesting inconsistent breaking ball mechanics rather than a clear fatigue pattern.

Eovaldi presents the most concerning profile, starting with the lowest spin rate and declining steadily from around 2,200 rpm down to approximately 1,700 rpm by pitch 65, before partially recovering toward the end of his outing. This sustained drop in spin rate could indicate genuine fatigue affecting his breaking ball quality in the middle portion of his outing, and may be a useful signal for coaching staff monitoring his performance in real time.

3.2.3 Hypothesis

It was hypothesised that as pitch count increased, release velocity would decline and pitchers would compensate by increasing their use of breaking balls and generating higher spin rates.

The data partially supports this hypothesis, however the patterns were far more individualised than expected, with no consistent universal fatigue trend observed across the five pitchers.

Eovaldi was the only pitcher to show a clear decline in both velocity and spin rate, suggesting he may have experienced the most notable fatigue effects of the five. Wheeler’s highly consistent and fastball heavy approach throughout his outing is an interesting finding, as his predictable pitch selection with virtually no offspeed pitches could potentially be exploited by opposing batters late in games who can reasonably anticipate either a fastball or breaking ball.

Gallen presented perhaps the most strategically interesting profile, beginning with an unusually high proportion of breaking balls before transitioning to offspeed pitches mid game and finishing almost exclusively with fastballs, suggesting a deliberate sequencing strategy designed to keep batters off balance rather than a response to fatigue.

Overall, the highly individualised patterns observed across velocity, pitch selection, and spin rate reinforce the argument that coaching staff should monitor individual pitcher trends in real time rather than applying a universal pitch count threshold, as fatigue manifests differently across pitchers.

3.2.4 Limitations

Several limitations should be considered when interpreting the findings of this use case.

First, the analysis is based on only five pitchers and a single game appearance each, representing a very small sample from which it is difficult to draw broad conclusions. A larger sample of pitchers and games would be required to identify more reliable fatigue patterns.

Second, as noted in the visualisations, postseason baseball features considerably higher pitch counts per game than regular season play, meaning these findings may not generalise to regular season contexts where pitchers are managed more conservatively and bullpen usage is more frequent.

Third, the dataset does not contain a direct measure of fatigue, meaning fatigue is being inferred from changes in velocity and spin rate, which may be influenced by other factors such as in game strategy, opponent quality, or deliberate effort management rather than physical fatigue alone.

Fourth, individual game context is not accounted for, as a pitcher may adjust their approach based on the score differential, the opposing lineup, or weather conditions, all of which could influence velocity and pitch selection independently of fatigue.

Finally, pitch count alone may not be the most appropriate measure of fatigue, as other factors such as innings pitched, days of rest between appearances, or cumulative postseason workload may be equally or more relevant indicators of a pitcher’s physical condition.

4 Conclusion

This report has explored pitch-by-pitch tracking data from the 2023 MLB Postseason, providing a comprehensive overview of pitching tendencies, pitch characteristics, and game context across 11,828 pitches.

The data visualisations revealed clear patterns in pitch selection, velocity, spin rate, and location, highlighting the strategic complexity of pitching at the highest level of baseball.

The two proposed use cases demonstrate the practical value of this data for coaching staff and players, offering data driven insights into pitch selection under pressure and individual pitcher fatigue management. While both use cases are limited by the scope of the postseason dataset, they provide a compelling foundation for future research using larger regular season datasets.

The increasing availability of detailed tracking data through platforms such as Statcast represents a significant opportunity for teams to gain a competitive edge through data driven decision making, and this report highlights just a small portion of the analytical possibilities that this data affords.

5 References

MLB. About Statcast MLB; 2026 [Available from: https://www.mlb.com/glossary/statcast.

Lebovitch J. Performance RP, editor. RPP Baseball; 2023 [Available from: https://rocklandpeakperformance.com/baseball-pitches-a-comprehensive-guide/.