New model clears up messy baseball rankings

Statisticians from Rice University and Cornell University have developed a new method to better understand complex sports rankings. Traditional rankings can be inconsistent, with different sources providing varying results and opinions shifting over time. This collaboration aims to create clearer insights from messy ranking data.

The method, called Bayesian Multivariate Rank Regression (BMRR), merges rankings from multiple sources and time periods while accounting for disagreements and incomplete lists. BMRR also quantifies the level of certainty associated with the final consensus. The researchers tested their methodology using Major League Baseball's annual top player rankings from ESPN, CBS, Bleacher Report, Yahoo Sports, and MLB from 2021 to 2024.

By focusing on 55 players who consistently appeared in the rankings each year, they analyzed how rankings were influenced by factors such as player performance, age, salary, and wins above replacement (WAR). The study identified three key variables associated with improved rankings: higher WAR, lower age, and higher salaries. By employing BMRR, the researchers were able to reveal that rankings are not only a reflection of recent performance but also expectations about a player's future potential.

For example, Shohei Ohtani, Aaron Judge, Mike Trout, and Mookie Betts were consistently ranked among the top four players during the 2023 season, with Ohtani winning the American League MVP Award. Additionally, the analysis showed that Ohtani's rankings fluctuated significantly over the four-year period, highlighting the importance of considering a player's trajectory when assessing their overall value.

The method also allowed for the comparison of individual rankers' behavior, demonstrating which sources tend to align with the broader consensus and which bring unique perspectives.