… or “why anet will never be able to make Glicko-2 work with gw2”.
First off, the current matchmaking is atrociously random’ed together. There is a reason anet doesnt release mmr for each account, because in no way does it reflect actual skill level.
MMR reflects one thing for soloqueuers (majority of gw2 players): luck.
Do you get underrated teammates that will ensure your victories? Do you get stuck with an overrated teammate that will ensure your loss?
Why is that, you might ask yourself. After all, gw2 uses fancy glicko2 as a matchmaking algo.
Well, using glicko2 to begin with is like trying to fit the square box into the round slot.
The reason is quite easy. Glicko (and Glicko-2) were made for either full teams, or single players in a 1v1 environment. They were not made for a mix of both.
Now, anet, in their endless wisdom, thought they could just take glicko2’s teamrating algo, and use it for everything.
Meaning… you queue solo, you play solo… and at the end of the match, you get a team rating. Now, if only all teamplayers would perform the same level, this teamrating would actually be correct. As it is, you however will mostly get some underperforming and some overperforming teamplayers. If you are LUCKY, you get enough overperformers to counterbalance that pesky far-rallybot. If you are not so lucky, you get two or more far-rallybotters (overrated accounts).
So, point one is -> team rating for five solo players results in inaccurate ratings, with players getting either underrated or overrated. Both is toxic for the system, because these error ratings will get more amplified the longer the system is running. Only a very small percentage of players actually meets their teamrating every single match. For those that get overrated, they will have the tendency to always get overrated, simply because they have to play with and against better players. Same for those getting underrated.
Now… on to point two. The same mmr is used for queueing solo, as for queueing as a team. So, you queue with 3 players, 2 of those are really good and play extremely well together. The third one is kinda new to pvp and just does what he is told by his other two teammates. That team goes on a winning streak, with each game’s endrating boosting mmr of all three players more and more. Now that third player logs in on a day when his two teammates are not online, and he queues solo.
Thanks to his overrated mmr, he will have to play against really good players… and chances are high, he should be one of those in team that counterbalances someone with a slightly under-average mmr for this team. He, however, fails miserably, because he now has to play totally different to what he as done with his team so far, and he doesnt have anyone explaining to him anymore what to do…. well, you get the picture.
Now, in theory, the system will slowly – game for game – adjust that players mmr. Thing is… even if he wins only 1 game out of 3, this “adjusting period” will take forever. And he will always be that guy the rest of the team has to carry, until… yes… until he personally adjusts HIS skillevel, to meet his mmr. Funny that… isnt it.
There are several more problems with the matchmaking system that exaggerate the inaccurate ratings even more.
Like, how experience with a profession isnt taken into account at all. Or how close a loss was – a loss is a loss is a loss.
Now… this wouldnt be a good thread at all, if I didnt post my ideas on a solution.
MMR should be based on:
-> personal performance during match (damage dealt, damage received, amount of deaths)
-> team performance (win or loose)
-> ranked games played
As it is, mmr is just a fancy word for your winrate, and that should be changed.