Variance Analysis - MLB gametimes in 2023

Introduction

For most of MLB history, average game time has crept upward, for a variety of reasons: commercial breaks, more pitching changes, hitters becoming more selective and seeing more pitches per plate appearance, pitchers working more deliberately, lights for nighttime baseball reducing the urgency of finishing games by sunset, etc. But thanks to a series of rule changes introduced for the 2023 season, the average elapsed time of an MLB game has collapsed, reversing about 40 years’ worth of trend in a single year. The most notable of these changes are the “pitch clock” that limits how long the pitcher can hold the ball before delivering to home plate, and limitations on the number of pickoff throws a pitcher can attempt.

Data Exploration

Major League Miscellaneous Year-by-Year Averages and Totals

Our data will be scraped from the table at the link above. We specifically need the columns Time, PA/G and Pitches/PA. Baseball pitch count data is virtually non-existent before 1988, and incomplete up to 1998, so we will only present the years 1999-2023.

Variances and trends

Our objective is to take this change in time per game and identify its sources. We can use this identity to produce a variance analysis:

\[\begin{equation} \frac{time}{gm} = \frac{time}{pitch} * \frac{pitch}{PA} * \frac{PA}{gm} \end{equation}\]

This gives us three sources of variance: pure pace of play (elapsed time per pitch), selection effect (batters seeing more pitches per plate appearance) and offense effect (number of plate appearances per game.)

1. Pace of Play: time per pitch

The pace of the game as measured by average time between pitches had gradually slowed in the 21st century. MLB made some halfhearted rule changes regarding how often teams can change pitchers, the no-pitch intentional walk, etc. but the pace continued to slow. Finally in 2023, the adoption of the “pitch clock” resulted in a dramatic acceleration in the pace of play.

2. Pitch Selection: pitches per plate appearance

One thing that has contributed to the growing length of games is that hitters have become increasingly selective, preferring to see more pitches per plate appearance, or “working the count.” This not only results in more pitches being thrown, but more pitching changes as well.

3. Offensive context: plate appearances per game

Some of the other new rules (e.g. banning certain types of defensive shifts) may have led to increased league-wide on-base percentage. Scoring runs is mostly a function of how many men you send to the plate (i.e. team on-base percentage) and more batters make for a longer game.

Waterfalls

After identifying the sources of variance for the change in average game time between two seasons, we can use a waterfall chart to display their relative magnitude. The average MLB game in 2023 is 162 minutes long, or 24 shorter than in 2022, with the pure pace of play effect actually improving by 26 minutes per game.

Comparing 2023 to 1999, the pace of play per pitch has sped up, batters are seeing more pitches per plate appearance, and the number of batters per game is down.