Long-Term Contracts and Predicting Performance in MLB

Thumbnail Image




Thomas, Duncan
Arcidiacono, Peter S.

Journal Title

Journal ISSN

Volume Title

Repository Usage Stats



In this paper, I examine whether MLB teams are capable of using players’ past performance data to sufficiently estimate future production. The study is motivated by the recent trend by which teams have increasingly signed long-term contracts that lock in players for up to ten seasons into the future. To test this question, I define the “initial years” of a player’s career to represent a team’s available information at the time of determining whether or not to sign him. By analyzing the predictive ability these initial years have on subsequent performance statistics, I am looking to answer whether—and if so for how long—teams can justify signing players to long-term contracts with guaranteed salaries. I also compare the results of the predictive tests with actual contract data to determine the per-dollar returns on these deals for different types of contracts. I conclude from my analysis that a player’s past performance does in fact provide sufficient insight into his future value for teams to make informed decisions at the time of signing a contract. Teams are able to better predict the future production of potential signees by examining their consistency and relative value in the initial seasons of their careers. Furthermore, the results from examining the contract data coincide with my findings on performance; teams and players arrive at salaries for long-term contracts that divide the future risk between the two parties. The returns on long-term contracts are thus demonstrated to be higher than for short-term contracts, as the overall value of longer deals compensates teams for the associated higher annual salaries.






Goldstein, Drew (2017). Long-Term Contracts and Predicting Performance in MLB. Honors thesis, Duke University. Retrieved from https://hdl.handle.net/10161/14794.

Dukes student scholarship is made available to the public using a Creative Commons Attribution / Non-commercial / No derivative (CC-BY-NC-ND) license.