Data analysis simply has not gone far in football. Things like the flight arc or height of a floated ball that can make a massive difference in a given situation are yet to be consistently captured by data much less understood. Even player ability is not well captured at all.
https://twitter.com/ddoublepivot/status/1219402983093522433
Data science won't go very far until there is an algorithm that understands football. Some AlphaZero of sorts that learns football by itself from scratch by running billions of match simulations (which presents the problems of dimensionality and realistic ball control in making).
If I were a corporation like Google or something interested in 'hacking' football, I'd probably hire the developers at Pro Evolution Soccer and give them resources to create the most realistic soccer simulation engine that they could and have an algorithm run a billion simulation
Until it comes up with pretty much similar patterns to what we already have in the game. Each pattern the algorithm develops will be given a weighted score to encourage it (for example, having an overload in buildup) and copied into a new algorithm without the weighted score to
eliminate optional bias, as we all know that football is not played in a rigid way.
So much can still be done and little will until clubs or corporations lean in on their massive resources to gain competitive advantage. Or some genius with a genius algorithm or metric comes.
So much can still be done and little will until clubs or corporations lean in on their massive resources to gain competitive advantage. Or some genius with a genius algorithm or metric comes.
Liverpool's current pattern of football cannot be achieved without massive amounts of athleticism. How will an algorithm from the present know that without human input?
• • •
Missing some Tweet in this thread? You can try to
force a refresh
