Speedrunning—completing video games as quickly as possible—offers a unique window into how human performance records evolve. Unlike traditional sports, speedrunning blends quantifiable metrics (time) with unpredictable elements like luck, iterative strategy improvements, and even scientific discoveries (such as uncovering game glitches). Yet, despite its popularity, there’s little systematic research on how these records progress over time or what drives breakthroughs. Analyzing this could reveal insights into human optimization, competition dynamics, and the science of record-setting.

How the Idea Works

One way to study speedrunning records is by modeling their evolution using statistical and machine learning techniques. The approach could start with a baseline assumption: records are random fluctuations with no inherent improvement. From there, it could incorporate:

• Logarithmic time scaling to account for exponential growth in the number of speedrunners attempting records.
• Performance extrapolation to separate noise (like lucky runs) from genuine skill or strategy improvements.
• Game-specific factors such as genre, patches, or community size that might influence progress.

Existing leaderboard data (e.g., from speedrun.com) could be used to train and validate these models, helping predict future records and uncover hidden patterns in speedrunning progress.

Who Benefits and Why

This analysis could serve multiple groups:

• Researchers in game studies, psychology, or statistics, who could use the findings to understand human performance and competition.
• Speedrunners and fans, who might gain insights into record trends, helping them strategize or appreciate the history of their hobby.
• Game developers, who could learn how players optimize or break games, informing future design choices.

For example, if the data shows that certain genres (like platformers) see faster record improvements than others, developers might design challenges differently to balance speedrunning potential.

Execution and Challenges

A possible execution strategy could involve:

1. Collecting records for a few high-activity games to test the baseline model.
2. Expanding to include trend analysis (e.g., logarithmic scaling) and game-specific factors.
3. Scaling up to compare trends across genres or communities.

Key challenges include filtering out noise (like outlier runs) and handling sparse data for less popular games. One workaround could be focusing on games with robust leaderboard histories or grouping similar games to identify broader trends.

By combining empirical data with theoretical modeling, this project could fill a gap in understanding how speedrunning records evolve—and what that reveals about human performance under constraints.