Can A.I. rebuild Man United?

For all of our efforts to better understand professional soccer, there's really only one golden rule: The more you spend, the more you win.

As the economist Stefan Szymanski has found, multiple times over the years, the single number that has the largest connection with the number of points a team wins over the long run is the amount of money a team pays its players. You can't, of course, just automatically double your wage bill and expect a 100% increase in performance. But the best players and the best coaches will eventually find their way to the clubs that can pay them the highest wages.

That is, on average. For the past decade, one club has consistently thwarted the wage-to-wins gravity -- in the wrong direction: Manchester United. While revenue doesn't exactly equal wages, it's close. And in the 10 reported seasons since 2012-13, when they won their last league title and Sir Alex Ferguson then left the club, United's average annual revenue ranked third in the world and first in England. Over that stretch, their average finishing position in the Premier League table is fifth.

They've underperformed their resources by such a degree, so consistently, that it almost seems like you'd have to actively be trying to sabotage yourself. The relative performance is so bad that I've often wondered if it's even possible to be worse. Like, if you let a teenager pick players based on ratings from the FIFA video game or you exclusively trawled Manchester United's Reddit page for tips, I'm not sure you'd do worse than an average of fifth-place over 10 years. In fact, I think you'd very likely do better.

Stream on ESPN+: Man United vs. Man City (FA Cup final)

Or, you know, what if you just let a computer run the team? And I'm not talking about a bespoke analytically precise algorithm like the smartest clubs in the world are using. I'm talking about one of the publicly available large-language models. What would artificial intelligence say about rebuilding Manchester United? And would it do better than the humans who built the team in the past or the ones now tasked with building it for the future?

Let's find out.