Introduction

A little matchmaking algorithm to create balanced teams based on the players' skillratings. This simple algorithm was written for a fun tournament of Quake Champions. We wanted to play team-based modes and rather than have the players form the teams themselves we wanted to balance out the skills between the teams to get exciting matches.

Streamlit frontend is available here.

1 Initial Seed

As a first step, players are subdivided into skill tiers. The number of skill tiers is the same as the team size. The best player from the highest skill tier is matched with the worst player from the lowest skill tier. From the middle skill tiers random players are chosen. For the next team the second best and the second worst are chosen plus random players from the middle skill tiers then the third best/worst and so on.

2 Balancing Algorithm

The optimization tries to minimize the score. As a scoring function first the average skill ratings of all teams are calculated. The score is calculated from the maximum deviation a team has from the total average skill rating. In each iteration of the balancing algorithm two teams are chosen at random or the minimum and the maximum scoring teams based on the min_max_pairing flag. Single players are swapped between the two teams. After swapping the new score is calculated. If the swapping improves the overall score the changes are kept. As stopping criteria either the number of iterations is used or the number of iterations without a change.

The Code

• The optimization algorithim is implemented in matchmaking.py.

• A streamlit frontend is implemented in matchmaking_fe.py.

• Command line implementation to read csv files and run the optimization: cmd_create_teams.py.

• Example input file: test.csv

Running from the command line.

You can run the algorithm from the command line.

Team size of 4 (4 even teams).

python cmd_create_teams.py -i test.csv -s 4

Team size of 3 (one player will stay without a team).

python cmd_create_teams.py -i test.csv -s 3

Run Streamlit Frontend Locally

streamlit run matchmaking_fe.py