https://www.sciencedaily.com/releases/2018/07/180723142932.htm

Estimating working life expectancies, or "life tables," particularly in sports, is seen as such a labor-intensive, detail-grinding task that demographers and statisticians have tended to avoid doing it. The result is a dearth of information about MLB career length, despite an abundance of historical data. Swanson, Jack Baker, Jeff Tayman, and Lucky Tedrow set to remedy this by showing that calculating life tables is actually quite easy if you use cohort change ratios. And using cohort change ratios allowed them to make forecasts with the same data they used for looking at working life expectancy.

"We used pitchers in the 1980-1981 season because that gave almost everyone enough time to finish their careers," says Swanson. "But we excluded those who had interrupted careers, such as those who played for a bit but might have been sent back down to the minor leagues for more training or because of injury. We wanted to look only at uninterrupted careers in the MLB, and it turns out that, on average, they're really short."