Computes Optimal Pickleball Pairings

Given a list of players and their ratings, runs thousands of simulations to find the teams/pairings that are the most fair.

Used for the Portland "Dinkies for Drinkies" mini pickleball tournaments.

How does it work?

The algorithm works in two phases: generating teams and then generating pairings.

Team Generation:

• Each team consists of 4 players, and the "strength" of the team is the cumulative rating of that team.
• Each team will have 1 player from the bottom 25%, 1 player from the 25-50% percentile, and 2 players from the top 50%.
• We randomly generate teams num_attempts times and pick the one with the lowest variance in team strength.

Pairing Generation

Dinkies for Drinkies has six rounds of gameplay. Each player keeps the same partner for 2 rounds, and they play with all three partners. No partners can play anyone on their team, and no partners will play the same exact opponents twice.

We run a graph algorithm to find a pairing schedule respecting the above constraints. We do this num_attempts times and choose the pairing with the lowest average rating difference between matches.