Please accept this as an admission of overcommittment, rather than laziness, but I posted something on bluesky and realized immediately afterwards that this can probably be easily tested.
If someone wants to take it upon themselves to regress wins over player-games lost to injury, I’d be most gracious. If they further wanted to interact that variable with total payroll expenditures (player payroll only, please), that would go further towards really testing the hypothesis. While I don’t tend to think there is much to be intuited from correlation coefficients, I would be curious to know how much the R-squared increases when you run a regression strictly over payroll and the lagged wins and then subsequently add player-games lost to injury to the independently variables. The delta on R-squared could be charted over time. There are other metrics that could be applied to try to control for overall talent, but real question is how accurately could you predict the final standings in a sports league if all you knew was player expenditures and injury luck, and if this has changed over time.
I’ll happily sit on a masters or undergraduate thesis committee for anyone who pursues this!
(Not sure there is enough meat on the bones for a PhD thesis, but happy to be proven wrong)
Economist Writing Every Day 