It's about entire models of game play style and running simulations with a prospective player to see how they fit in. The data science there is not querieable science. It's more about simulations with numbers.
"This summer had a familiar feel in many respects as Arsenal faltered in the transfer window (...) yet it concluded with Granit Xhaka, Rob Holding, Skhodran Mustafi and Lucas Perez arriving for a combined fee approaching £90million.
“Four years ago, they bought a company called StatDNA in Chicago,(..)They will take a player and collate data based on every pass, sprint, shot, assist, goal – you name it – he has ever made. These stats are fed into a computer that assesses strengths and weaknesses against every professional in the game before producing an overall value of that player.
Arsenal tend to use statistics to raise questions or confirm observations on possible targets and sometimes unearth players for scouting. "
https://www.standard.co.uk/sport/football/the-deal-arsene-we...
Manchester United for example have spent VERY heavily on new players since 2013, and the ROI has been awful almost across the board. They acquired Alexis Sanchez in 2017, who was joint top scorer in the 2016/17 season with 24 goals. At United, he has scored only 5 goals in 18 months.
Chelsea FC are scoring one duck after another by signing Morata, Giroud and Higuain, all of whom have flopped as strikers despite doing well in their previous teams.
Arsenal are an example of a team that uses analytics to suss out bargains in the transfer market. And yet, two of their most expensive analytics-driven signings, defender Shokrodan Mustafi and Granit Xhaka, have fared poorly while cheaper buys like Sokratis Papasthopulous have thrived.