Making MatchMaker

I’m Phil, I’m in my 60’s and still playing football every week. I play 4 to 7-side football every Wednesday night with my team the Strollers. The Strollers were formed in 1985 and I joined them in 1995. We don’t play in a league and are a group of friends who pick teams from within our squad with players ranging from teens to 60’s. 

Some of the original players have retired or moved out of London, where we play, but others still play with their sons and daughters joining, along with new friends arriving. We have a large and diverse squad so picking evenly balanced teams is difficult. 

For several years, from 2019 to 2022, our players took turns to pick teams from our squad of players and record the results of the match. A close game, like 5-4 was considered good and a one sided match like 7-1 was bad. For each match we recorded the team picker and the score and made a league table of team pickers based on each team picker's ability to pick well balanced teams (ie 5-4 not 7-1). 

After 3 years it became obvious that some players were much better at picking teams than others. We then started inviting players, who picked good, evenly balanced teams, to pick teams more regularly. This helped reduce the number of one sided matches. 

By 2022 I started to develop a software algorithm based on machine learning, which became an iPhone and Android App, called Football MatchMaker. We found that MatchMaker reduced the number of landslide victories and increased the number of close games even more than our best human team pickers. MatchMaker records things like match results, who played, who scored and much more. This information is then reused when picking future teams. 

Have fun and kind regards, Phil.