Zero-Sum Series Splits of One Run

The San Francisco Giants and San Diego Padres split their four-game series last week, with each team winning two games by one run. In our #social-baseball slack channel, my coworker asked:

any data whizzes wanna figure out how many MLB 4 game series have ended 2-2 with all one-run games?

I put myself on the case.

UPDATE JULY 6, 2025

I finally took the time to do this properly. I exported game-by-game data from Retrosheet since 1900 and (with the help of Claude) used pandas to analyze the data for these split series decided by one run each game.¹The weirdest part about debugging code made by an LLM is interrogating their odd assumptions. Claude was exceedingly cautious in how it identified a four-game series. It originally allowed for a gap of up to 7 days between games of a pair of teams, which makes no sense. In baseball, a series is nearly always played on consecutive days. Once I corrected this error, my final count went from 32 to 49.

The list of 49 such series can be found here. The most recent occurrence before 2025 was a 2019 series between the Reds and Mets.

ORIGINAL APPROACH

I had no clean way to run a single query that pulled these because Stathead does not allow “OR” qualifiers in their search. So, this took slightly more manual work. This is what we do in the quest for knowledge.

This Stathead query contains the longest streaks of at least four regular season games such that every game’s run differential was between -1 and 1. I was only interested in streaks of exactly four games, so I went straight there.

I then sorted each page by total streak run differential, which is necessarily zero when both teams win two games in the series. Sorting by streak end date, I was able to peruse each page for consecutive rows with matching start and end dates—if the home team had one of these streaks, so did the visiting team.²There is a slight issue with this approach: what if one team had a streak of 4 games, but the other team had a streak of 5 or more because they had the necessary run differential before or after the four-game series? I only realized this problem while writing this post, but I’m too far into it to fix it this moment. I then manually checked that there wasn’t a coincidence in the matching dates, and also confirmed for older dates that it wasn’t part of a five-game series.

All that led to a list of the following four-game series³Again, I’m probably missing some because of the issue I pointed out above. I’ll try to fix that at a later point. where the teams each won two games, and all four games were won by exactly one run.

• 1
  The weirdest part about debugging code made by an LLM is interrogating their odd assumptions. Claude was exceedingly cautious in how it identified a four-game series. It originally allowed for a gap of up to 7 days between games of a pair of teams, which makes no sense. In baseball, a series is nearly always played on consecutive days. Once I corrected this error, my final count went from 32 to 49.
• 2
  There is a slight issue with this approach: what if one team had a streak of 4 games, but the other team had a streak of 5 or more because they had the necessary run differential before or after the four-game series? I only realized this problem while writing this post, but I’m too far into it to fix it this moment.
• 3
  Again, I’m probably missing some because of the issue I pointed out above. I’ll try to fix that at a later point.