Only issue to this is that if you died early, left and started new game, your MMR would change during new game, but I doubt MMR change from single game would affect matchmaking that much...
I wouldn't do it just based on number of escapes. Time of death is really important to this.
Only issue to this is that if you died early, left and started new game, your MMR would change during new game, but I doubt MMR change from single game would affect matchmaking that much...
I wouldn't do it just based on number of escapes. Time of death is really important to this.
Most systems have a delay, say 8 hour lead time for data capture.
Time of death has similar issues. Consistently dying early due to bad play severely hurts your teammates odds of escape (and would therefore affect MMR), but not all early deaths are necessarily bad for your team's odds of winning. If someone is really dying in the first minute of most of their games, both the terrible overall escape rate by itself (and potentially a modifier for "losing streaks" where you lose more if the vast majority of your recent games have had few or no escapes) should put them towards the bottom rather quickly. What's the benefit that comes from the extra complexity and side-effects that death time being factored in would introduce?\
Edited to quote what I was replying to.
Hmm, so maybe instead of real time, it would be better to track number of gens left. 5 minutes with 4 gens left is different than 5 minutes with 0 gens left...
I'd rather they just incorporate actual skill into the MMR - gens, chases, altruism. The problem right now is the potato survivor hiding in corner relying on the team to either die or carry them for the escape because escaping is the only thing that matters. If team dies and the potato escapes the potato moves up to drag down another team. If the team carried them the potato to escape again the potato moves up to either be matched with that team again or another team to drag them down. The players that hide/do nothing all match should be dropping in mmr not rising only to continue dragging down teams.
I agree, a "team"-based performance rating would account more fairly for some important aspects of survivor play. The killer rating could in fact also benefit from being based on the performance against the entire team, so depending on how many survivors they've killed rather than looking at 1V1s against individual survivors.
Although I will say that even with the per-survivor escape/kill rate, the rating will still reflect performance levels well enough over large enough sample sizes, with edge-case scenarios such as a survivor sacrificing themselves to lead to more survivors escaping than otherwise would have escaped averagering out. The actual problem the MMR system has is not really that the per-survivor ratings don't capture performance levels well enough, it's that the rating adjustments aren't harsh enough, and the matchmaker not strict enough. People gain MMR more easily than they lose it, there is an MMR cap in place, and the matchmaker seems to care much more about keeping queue times low than keeping ratings of players close. These things are what BHVR should first be tweaking.
That said, the team-based ratings could perhaps actually help a bit with the issue of killers gaining MMR too easily:
The rating adjustments would scale off of the average rating of the survivor team, rather than the individual survivor ratings. For example, if you have a 1000MMR killer facing survivors with 500, 1000, 1200, and 700MMR respectively, rather than the killer gaining a lot of MMR from killing the 1000 and 1200 survivors, the adjustment would depend on the team's average rating, which is 850 in this case. This makes it less possible for killers to gain a lot of MMR just because they killed 1-2 higher-rated players in an otherwise rather outmatched survivor team.
But I do think that ultimately these changes would not be necessary if the rating adjustments in their current form would just be harsher (losing more for losing against higher-rated opponents, gaining less for winning against lower-rated opponents), the cap would be removed, and the matchmaker more strictly tried to match players closer within their rating brackets. I for one would not mind having a few minutes longer average queue times at top MMR brackets if it meant that much more often I would actually get matched with 4 other players from within those brackets.
The drawback to this, due to the rather miserable coding of this game, you would be forced to wait out a match to it's end for the MMR to actually calculate, even if you were the first survivor to die.
The game can only track MMR on individual players, and only compares other MMR values of multiple players during matchmaking.
Tldr: This is an archived version of the DBD forum, it has all posts, comments and media. The archived data will be published at the end of October. This site is not in any way affiliated with or endorsed by Behaviour Interactive.
Comments
Only issue to this is that if you died early, left and started new game, your MMR would change during new game, but I doubt MMR change from single game would affect matchmaking that much...
I wouldn't do it just based on number of escapes. Time of death is really important to this.
..
Most systems have a delay, say 8 hour lead time for data capture.
Hmm, so maybe instead of real time, it would be better to track number of gens left. 5 minutes with 4 gens left is different than 5 minutes with 0 gens left...
I'd rather they just incorporate actual skill into the MMR - gens, chases, altruism. The problem right now is the potato survivor hiding in corner relying on the team to either die or carry them for the escape because escaping is the only thing that matters. If team dies and the potato escapes the potato moves up to drag down another team. If the team carried them the potato to escape again the potato moves up to either be matched with that team again or another team to drag them down. The players that hide/do nothing all match should be dropping in mmr not rising only to continue dragging down teams.
I agree, a "team"-based performance rating would account more fairly for some important aspects of survivor play. The killer rating could in fact also benefit from being based on the performance against the entire team, so depending on how many survivors they've killed rather than looking at 1V1s against individual survivors.
Although I will say that even with the per-survivor escape/kill rate, the rating will still reflect performance levels well enough over large enough sample sizes, with edge-case scenarios such as a survivor sacrificing themselves to lead to more survivors escaping than otherwise would have escaped averagering out. The actual problem the MMR system has is not really that the per-survivor ratings don't capture performance levels well enough, it's that the rating adjustments aren't harsh enough, and the matchmaker not strict enough. People gain MMR more easily than they lose it, there is an MMR cap in place, and the matchmaker seems to care much more about keeping queue times low than keeping ratings of players close. These things are what BHVR should first be tweaking.
That said, the team-based ratings could perhaps actually help a bit with the issue of killers gaining MMR too easily:
The rating adjustments would scale off of the average rating of the survivor team, rather than the individual survivor ratings. For example, if you have a 1000MMR killer facing survivors with 500, 1000, 1200, and 700MMR respectively, rather than the killer gaining a lot of MMR from killing the 1000 and 1200 survivors, the adjustment would depend on the team's average rating, which is 850 in this case. This makes it less possible for killers to gain a lot of MMR just because they killed 1-2 higher-rated players in an otherwise rather outmatched survivor team.
But I do think that ultimately these changes would not be necessary if the rating adjustments in their current form would just be harsher (losing more for losing against higher-rated opponents, gaining less for winning against lower-rated opponents), the cap would be removed, and the matchmaker more strictly tried to match players closer within their rating brackets. I for one would not mind having a few minutes longer average queue times at top MMR brackets if it meant that much more often I would actually get matched with 4 other players from within those brackets.
The drawback to this, due to the rather miserable coding of this game, you would be forced to wait out a match to it's end for the MMR to actually calculate, even if you were the first survivor to die.
The game can only track MMR on individual players, and only compares other MMR values of multiple players during matchmaking.