Introduce premade MMR penalty for TMM.
-
Agreed, a % could be used so it doesn't overinflate the rating of low rated players or become useless with high rated players. Just like 100 is 10% of 1k but just 2.5% of 4k. Therefore the extra should be made a % rather than a fixed number. But deff agree communication adds an insane amount to the level of play.
-
This is one of the trueskill 2 features. They add an offset to every players skill estimation based on the size of the party that player is in.
Being able to accurately determine the size of the offset required for each party size would require determining the actual win rates of parties vs expected win rates. Does anyone know if FAF stores the information of which players are in a party for a particular game on the serve? I do not believe replays currently have this info.
-
A MMR penalty for teams is feasible. It should be a flat amount and not a percentage, because the zero point of the rating scale is arbitrary. The hard question now is: how big should the penalty be? We don't save the teams of a game anywhere to my knowledge.
-
One solution could be to add team information to replays, wait a few months and then do the analysis. An arbitrary value of +100 per player in a party could be used in the interim. +100 feels about right to me, but it's hard to say without data.
-
I feel like +100 per player might be too punishing at the lower end of spectrum. That's 300 rating addition when the full team already is pulling around 3-3.2k rating for 4 players. That's 10% increase of the team true skill.
Instead maybe it would be possible to group the teams into divisions by their overall rating and then assigned the penalty based on what the overall team division would be.
For example a Silver team of 3000 overall true skill would be given +35 per premade. A Diamond team of 6500 would be getting +75 per premade player and GM 8k rated team would be getting 100+.
-
I'm not sure having it vary based on skill is desirable. It would imply that 2 GMs in a party gain a larger % chance of victory relative to baseline than 2 Silver players. If I had to guess I would assume the opposite is probably true. This is because higher rated players are more able to infer what their teammates are doing with limited communication + they are more adept at communicating via pings and team chat. But overall, I expect the difference between performance improvements is probably small between high and low rated players from being in a party.
-
MMR stands for?
-
@brutus5000 said in Introduce premade MMR penalty for TMM.:
MMR stands for?
Match maker rating. More or less trueskill equivalent, or in some cases Trueskill. Basically the thing that hides behind the Divisions etc.
-
Ngl my performance decreases by 10% for each person in voice with me
-
I question if this is really a top priority.
I do agree that being a premade team and on voice does improve the overall team strength quite a lot. But then it also seems to me like it's always the same people that play together (at least in my 1600-1800ish rating bracket), thus their rating naturally represents them being pre-made already. I have found this to be no issue for me in TMM.
A factor that I think causes way more problems to the quality of games is a premade team with a very large rank disparity. Like 2200+600 queuing together, and getting matched against a pair of 1400s. These games I always find very non-enjoyable, and the outcome is also very volatile depending on the size of map. On big maps, the even-ranked team has basically no chance, as the higher rated player has time to take some of the lower rated player's resources and outscale. On smaller maps, the team with the large rating disparity still has a very large advantage if on voice, as in my experience the coaching/help on major strategy calls from the better player can make a large difference.