FFA start point randomizer
|Reported by:||beherith||Owned by:|
The way FFA random points are currently chosen is the following: If there are K start positions, ordered from 1..K, and there are N players, the first N start points are used and one is randomly assigned to each player. This is very difficult to balance, so I was thinking a k-means algorithm might work better: Cluster all the metal spots and geo spots (each geo is worth 1 metal spot) to N-means, and then greedily choose the start point closes to each cluster center. This wont take into account 'adjacency' of start points, where players with more neighbours are worse off, but it might still be better than default.