Soccer AI getting ‘smarter’; David Beckham becomes even less relevant

If you like sports, or physics, watch the video. Just, dear God, mute it first.

I’ve been playing soccer — both real and virtual —  for probably more than a decade now, and I’ve never scored from a free kick. For one, there’s always three or four dudes, generally standing about 10 yards away, whose only job is to bone you over. Trying to make a soccer ball curl around an obstacle is hard enough — putting it past a goalkeeper (again, his only job is to sh*t things up) is damned near impossible.

Thankfully, in my FIFA career, I don’t think I’ve ever conceded a free kick, either — probably owing to the fact that the kick-taker and goalie are controlled by the exact same AI, resulting, basically, in a tie. Unfortunately for me, that’s about to change, if the winners of the 2008 Machine Intelligence competition have anything to say about it.

The MI competition is pretty self-explanatory — it’s an annual conference about artificial intelligence. The DeMontfort Leicester team won the competition by using Darwinian evolutionary theory to make virtual free kick-takers “smarter.” Unfortunately, I don’t know enough science to tell you how it works, but the implications are interesting. Says Simon Coupland, half of the winning two-man team:

In video games, scenarios tend to be scripted, rule-based, deterministic systems. This means if you get the same situation twice, the exact same things will happen in the exact same sequence.

Any hardcore video gamer will tell you this is a significant limitation of current games. This new approach gives a diversity of good computer game play and means the computer agent players are unpredictable.

The team believes that not only would it greatly improve the way soccer sims are made, but that the technology would be easy to implement.

The bottom line? My keeper’s ability to ass everything up for other people is slowly, albeit surely, slipping. 

[Via Computer Weekly

Joseph Leray