In 2012, data scientists from the University of Southampton, UK, developed an algorithm that could beat 2M human players at the Fantasy Premier League. Two years on, they are now launching the RateMyTeam tool that allows all football fans around the world to use the algorithm to improve their team. This effort was led by Paul Morgan, who runs the FantasyFootballFirst blogging site, in collaboration with Southampton researchers Dr. Sarvapali Ramchurn and Tim Matthews. “We thought it would be really cool if we could let the general public interact with the algorithm, and Paul really helped us take it to the next level. We realised every user needs personalised advice on their team and we adapted the algorithm to provide advice rather than only find the best team.” said Dr. Ramchurn, who co-developed the algorithm with Tim Matthews, currently doing his PhD in Machine Learning at the University. “I’m really excited to partner with the team at Southampton University. The Fantasy Football algorithm has an excellent track record, proving their statistical model works. By joining forces we have created a very powerful Fantasy Football tool.”, said Paul. “Most important to the success of the Rate My Team tool is the combination of the simulation with the judgement of experienced Fantasy Premier League managers. We feel this is a winning formula and we look forward to Fantasy Football players trying it at fantasyfootballfirst.co.uk.”
So, does this mean the users of the tool will all get the same advice? “The algorithm cannot predict the future but it definitely does better than most human players. Our hope is that the sum of human expertise and machine intelligence will be greater than the parts” said Ramchurn.
The RateMyTeam tool can be accessed at: http://fantasyfootballfirst.co.uk/rate-my-team-player-picker/ The work on the fantasy football algorithm was done as part of the ORCHID project (www.orchid.ac.uk), a research project looking at the development of advanced information systems called Human-Agent Collectives.