Roma President James Pallotta is aiming to use machine learning to help “identify the next Francesco Totti”.
The American owner spoke at the MIT Sloan Sports Analytics Conference last week, and he has some innovative ideas for securing the best talent for the future.
“One of the toughest things is actually getting the coaches to listen to the stuff we’ve been doing,” Pallotta explained.
Roma President James Pallotta is aiming to use machine learning to help “identify the next Francesco Totti”.
The American owner spoke at the MIT Sloan Sports Analytics Conference last week, and he has some innovative ideas for securing the best talent for the future.
“One of the toughest things is actually getting the coaches to listen to the stuff we’ve been doing,” Pallotta explained.
“About nine months ago we decided that we were going to try to go to the next step. You know, if you want to try to identify the next [Lionel] Messi – if there is ever another Messi – or you want to identify the next Ronaldo or Totti at Roma, it’s really difficult to do that.
“You could look at 6,000 players. It came about because we are looking at 6,000 players, but if you want to look at certain characteristics of a player you’re looking at hours of film on a particular player.
“So we wanted to get to the point where simple things like ‘ok these are the characteristics that we’re looking for in a player’, and I don’t mean number of tackles or things like that, but other characteristics of play – acceleration, things like that – the only way we thought we could end up doing it was through machine learning.
“We actually got lucky and we met a number of people, and a couple of women who played football in Italy and one was an astrophysicist and the other had a PHD in statistics.
“That started the group we’ve been building out on the machine learning.
“So if someone like Monchi, who is our football opps guy [sporting director] says ‘I want to see this, this, this and that’ we actually can start building a system that identifies those players, just to filter out the 6,000 players.
“That’s what we’re trying to do is really just get a filtering system as much as anything else.
“You might think that’s easy but it’s not, because the factor is really how much computing power you need for all the video that’s out there.
“The data is somewhat out there, the video is out there, but it’s a lot of computing power because of the vision aspect of it.
“We think we’ve found somebody who has solved that at another college but that’s kind of the direction that we’d like to be heading.”