Sports Analytics Challenge

Participate in a unique contest: revolutionise football using your science skills!

ページを翻訳

The issues of the challenge

To use a data set we provide to create an algorithm that will help optimise all aspects of sports performance before, during and after matches!

For the prediction

In the resources of your participation area, you will find a file with an explanation of the databases. The following procedure will be applied several thousand times to your algorithm:

  • We randomly choose a match from the test database (F24 Opta files of all matches in the second half of the 2016-2017 Ligue 1 season).
  • We randomly choose the first or second half for this match.
  • We randomly pick 15 minutes of the selected halftime (all events between t and t+15 minutes with t randomly selected).
  • We replace all the names of the teams by "1" (home) or "0" (away). In the files, it means we replace "team_id" by "1" or "0".
  • We delete all the players IDs and write "0" instead, except for one randomly chosen player (who played more than 800 minutes on the learning dataset and did not change team in January). We write "1" for the ID of this specific player. In the files, it means we replace "player_id" by "1" or "0". In addition, when: Type ID=140, Type ID=141, qualifier_id=140 or qualifier_id=141 appears, we replace the values by " ".
  • We delete all position information (y, x) for all events except the last 10 Opta events within 15 minutes. In the files, this means that we replace "y" with "0" or "x" with "0".
  • We will also remove everything that is written in the F24 files before the first OPTA event. We also replace the values of Event timestamp, Event id, Q id and version with " ".
  • We reduce some of the information of the last 10 events. In the files it means:
    - For the last 10 events we replace "outcome" by "
    "
    - For the last 10 events we get rid of information about qualifier_ID 
    So : we replace the "value" by " " and we replace all the "qualifier_id" by " ".
Ready to start?

Click on the "Participate" button to access your participation space and discover all available resources.