This is how I would try to program it if it were me and that announcement was my motivation:
- First get a pool of players with as little gp diff. as posssible, say 50k ranges.
- Assign some weight to 12+ pieces
- Assign less weight to total amount of g12 in a players roster
- Assign less and less weight to each gear tier after that and higher thresholds
- Assign some weight to amount of total zetas
- Assign some weight to total amount of mods and their tier distributions
I would try to then try to match the players in the pool where this full picture looks as similar as possible. I already got to the accuracy of previous -probably GP only- matchmaking we had before these considerations.
So each of these weightings would have threshold values. i.e. the amount of g12s are +-%20 different players in comparison, they won't get matched.
Any weights can be picked for these various parameters. i.e. maybe g12 has a 0.20 multiplier while mods have a 0.10 one. Depends on what matchmaking wants to prioritise.
-------------------------------------------
*Note that I would completely ignore metas, which metas the players would have or which toons they chose to develop. That's player advantage and strategy how they operated within all the avaliable toon options.
*I would also fully ignore the amount of speeds on mods, only consider how developed the mods are and compare the amounts players has. Naturally stat distributions would be different between players with similar amounts of 6E, golds, purples etc. That's both luck and player strategy of mod developmental paths.
-------------------------------------------
I think whatever weighting CG is going with, they will completely ignore mods there and that's also what will make the difference in the outcome of higher leagues like Kyber.
