How the MMR works, and why what Patrick said makes absolute sense (from a dev POV)

I agree that what causes most of the mismatches is the system needing to find a balance between waiting in queue for closely matches opponents versus expeditiously getting you into a game quickly. That doesn’t make the topic meaningless, though, since it’s useful to think about how to rate killers and groups of four survivors against each other so that, when close matches are possible, the system can recognize it and prioritize those links.

I do disagree that It’s impossible to make a system which can reasonably make those estimations in a 4-vs-1 game. In fact I think the current MMR probably does a good job of making close matches IF all the players it can put into a lobby are roughly the same MMR. I would bet that, aside from brand new players with an artificial rating that might be too high, if all five players in a match are about the same MMR it will be a close game. That said, if I also had to guess where a weakness is, it would be how the system calculates the survivors net rating when their individual ratings are significantly different. If it’s using a simplistic formula like just taking the average of their ratings then it’s probably not getting the most accurate net rating that it wants for a killer to go against them. You would probably want something more like a neural net that is trained to input four survivor ratings and output an estimated optimal killer rating to go against them.

So TLDR, I agree on the main cause of mismatches being queue times which are caused by factors outside the control of the MMR calculations, but I disagree that it’s meaningless to think about how those calculations work and trying to make sure that close matches result when appropriate players are available in the pool.

Look at few minor things:

Grab mechanic - works horribly and can be timingly abused. Where mid grab animation gets canceled into the attack. This could prevent save or doing gen and give a killer upperhand - issue!

Grab mechanic part 2:

Survivor which has been downed at the exit gates most likely due to bad manners can be saved to exit by not being able to perform a grab on a guy who tries to revive. By either reviving or faking revive in intervals, downed survivor will slug out - It has not been addressed to this day.

From all survivor perks, 95% of the time DH BT DS IW COH. BT is good perk and I do not see any issue with it.

DH has been “fixed” and proven way OP. Up until “soon” nothing will change.

CoH has drastically increased surviving rate and efficiency, as you do not need 2 survivors off gens to perform a heal, but everybody can have an amazing selfcare.

QoL changes or improvements on killer side is almost zero to nothing. Not to even mention it took time, to even fix totem spots, where killer COULD NOT snuff the totem.

The game changes are making mostly useless 80% of killers roster - maps, dh, coh.

If others are having their opinion close to mine it makes it communites opinion which express community dissatisfaction which leads to dying game.

Because clearly, some still struggle with staying or leaving and some have left. Yes people come and go. It makes sense. But once there is more people leaving then coming, that is the red flag that something is wrong. Is it not enjoyable for survivors or killers?

Queue times represent ratio, and killers definitely suffer more. Every nerf tries to squeeze more effort from killers to achieve same result. So for every good result you eventually get punished.

I love your reply, it clears up alot.

One question tho, Given the recent harsh dialogue about SBMM on the forum. Have you or the team discussed a rebranding of SBMM to MMR?

I think a lot of people get their underwear in a twist over the word, "Skill"

Top part is an issue that someone who does game development can explain why it's a difficult problem to fix.

2nd part is annoying but it's a strategy. You need people to be alive, usually full health, hope the killer doesn't have STBFL and not have any power that stop survivors exiting. It's annoying but nothing that is a top priority to be addressed.

I'm going to be honest. The whole thing now just seems killer sided. Not one of your points is about the other side that isn't your preferred role, which is survivor. Totems being uncleansable? Survivors had it for years.

My point is we all want this game to be perfect for ourselves. People saying they want the game to be balanced (which is 2k at the highest level) are the same people complaining about the games being to sweaty.

Is the game perfect? No. Do I think they can make some better changes? Yes. Will I blame them for me wanting to take a break from the game or fully quitting? No.

I'm terrible at explaining so my apologies if this reads like a mess but I think we all just want what's best for us, and don't think with a neutral mindset. This is what creates the 100,000 threads complaining about stuff, which ultimately leads to us quitting or annoying enough people to make them quit.

You conflate an individualised rating change with individualised ratings. All team-based games with MMR implementations already have solutions for this. You cannot gain MMR in CSGO while the rest of your team loses it. My critique of the individualised nature is that it treats the game as 4 1v1s happening simultaneously. To my knowledge no other competitive team-based game does that, and it doesn't work. Can you imagine how silly it would be if a match of CSGO tried to define a single match 4 1v1s happening at the same time such that the losing team can still have a couple winners? It's ridiculous because it doesn't reflect the nature of the game being played.

Better players do not have answers to a facecamping Bubba. If you think otherwise you are familiar with only a very low level of play and fail to consider the entire skill range of DBD. A facecamping Bubba will always get at least 1 kill if they play with minimum competence. Even if they are up against the number 1 team in the world. That is the nature of DBD and there are other similar ways to achieve this guaranteed at least 1 kill regardless of skill.

You have failed to address or consider my point regarding how playing better in chase leads to a higher chance of you dying. Instead you just pointed out at best a nitpick in my phrasing of your statemens. You said it was either a fluke or something they can repeat. Well if they repeat it they will just die the same way. Better survivors who loop the killer for too long will often find themselves facecamped and killed directly because they were too good. The system actively lowers the ratings of better players and increases the ratings of worse players. This is a functional failing.

I've been very clear this isn't about MMR increasing being a good thing or losing MMR being a bad thing. I want a functional system. That's all.

I'm not going to propose a win condition for a match of DBD since it's not really too relevant here and you can come up with a few viable options to choose from. Regardless, DBD is a 1v4, so it is a simple obvious necessity that the basis for determining who wins and who loses must be based on the format of 1v4. This complete basic assumption is what's missing from DBD's implementation. The MMR is based on a win condition for a different game that doesn't exist. It is theoretically possible that an MMR system that doesn't reflect the gameplay could still be accurate at measuring skill, though it would be one hell of a coincidence. But along with the examples I've already given of how the system fails by decreasing the MMR of the most skilled player consistently, and increasing the MMR of the least skilled players consistently, it is clear that this coincidence has not happened.

"I literally already answered this directly and you've completely ignored what I've said, whilst quoting the post I said it in. I see the problem here. Sadly it's yours to resolve."

You can answer whatever you wish. You are simply wrong on this point, or at least, that's what I believe. When a disagreement of basic facts arises I can't really try and argue the case. Just for the sake of it I can show you what the would look like, and why it wouldn't really benefit anyone.

"Firstly, they're better in chase, so less likely to be caught." Correct but not relevant in this case. To get a 1K you must only get 1 down. There is no way to prevent a Bubba with a basic competence in the game from getting at least 1 down per match.

"They're also more map aware so if they do go down, they'll go down where a rescue is most likely to succeed." Correct but not relevant in this case. There is nowhere on the map on any map where a rescue against a facecamping Bubba is possible.

"Furthermore, better teams can co-ordinate rescues in difficult circumstances." Correct but not relevant in this case. There is no way to rescue against a facecamping Bubba even with perfect co-ordination.

"As killer, I'll camp the hook in the EGC, and I've seen many a time survivors make a successful rescue." If you're not talking about Bubba it's not relevant. If you are and you allowed all four survivors to escape while facecamping as Bubba you made a critical error that is either a massive fluke or suggests you are at a very low level of play.

Now since literally all I have done is said 'no' to everything you brought up, I don't imagine you're exactly 'convinced'. But you seemed to feel like I was ignoring you, so there you go. You can quite easily counter 'no' to my 'no' and we'll be here all day. It is common knowledge in DBD that against a facecamping Bubba you do not even attempt to save because it's impossible. That isn't much of an argument but you must know you're going against prevailing wisdom, right? You seem to think my position here is extreme when it is the mainstream view from people that play the game.

Yeah, I was agreeing with you with that second paragraph and putting forward an example of my own. Maybe I should have been clearer, sorry.

Even if we include the opening gate part, I don't see how you can argue survivors that died before they could escape achieved 100% of their objective. It's not just "walking out of the gates", as if dying before doing so meant they just... chose to die. If the killer prevents them from escaping, even if the gates are already open, they did not achieve 100% of their objective. And again, this is obvious if you look at the fact that in a 2-escape match, the 2 dead survivors can also have already died before all gens were done.

And sure, if you look at the survivors that did survive in this scenario, you could argue "but they achieved 100% of the objective as per per-survivor MMR logic", but that is not an issue, because they did in fact not achieve the objective of (team-)playing well enough for the other survivors to also be able to escape, which means they weren't so much better than the killer/couldn't play well enough that they would always be likely to escape, which over many iterations of matches would mean they too will die in such matches sometimes.

Over large enough numbers of matches, individual survival rate does correlate very much with teamplay skills/performances, since teamplay leads to increased survival chances of everyone in the teams. The teams change constantly, but the teamplay skills of individual players in them affect their survival chances as a team and as such of those players themselves consistently. Sure, you can create a ridiculous scenario in which an SWF is actually a team that only ever plays together, and one person somehow dies constantly and the rest constantly survives, but then their rating being lower than that of the rest of their team doesn't matter much, for one thing because they will be paired according to their averaged rating as a group, which they won't pull down a lot (even if they literally have "0" MMR, their group average will still be 75% of the rest, with the rest also having a boosted MMR since their survival is dependent upon that player's sacrificing themselves, so they will get ever-stronger killers until they eventually find their match where they cannot constantly 3-escape), and for another because they are literally making the decision for that person to die, so that person obviously doesn't mind dying, and so, so what if the rating too predicts they will.

While I don't think the current per-survivor paradigm leads to any real issues in accounting for "team" performances (not least because the scenario of an SWF actually exclusively playing together all the time without ever changing anyone or anything and specific players in them dying notably disproportionally more often than the rest probably doesn't exist even just once), you are right that these ratings take more matches before the team performances have the according impact on individual survivor ratings. Therefore basing rating adjustments on the group survival rather than individual survival can more accurately (over less iterations) predict teamplay skills/performances of individual players. Good for us, BHVR will do this in the future!

I also think basing rating adjustment calculations on the team average could be beneficial. Say you have a group where two survivors have an MMR of 2000 and two of 1000. Their average is 1500 MMR, so let's assume they get matched against a 1500 MMR killer. If the killer targets the lower-rated survivors and manages to kill even just one of them quickly, the survival chances of the other survivors decrease dramatically. This is an issue particularly of asymmetrical games, where one side is a team and the other is not, meaning the "team" side suffers from skill disparity a lot while the other side benefits from it more often than not. And so if the killer does end up killing them all, now they get some +rating from killing the 1000MMR survivors, and a lot +rating from killing the 2000MMR survivors, rising notably in rating despite potentially having had a rather easy time due to abusing the blatant weak links.

Say it's a simple divisional ratio, they get 66MMR for each of the lower-rated survivors, and 133 for each of the higher-rated ones. In total, they get +400MMR, putting them at 1900 and close to the better survivors after this match. Obviously the rating adjustments aren't this dramatic in reality, but this is for argument's sake. Now if we instead take the team's average MMR and only grant the killer the full MMR increase if they kill everyone, they get +100MMR. The killer's rating inflates less from winning against mismatchedly-rated teams. The survivor MMR adjustments are also more sensible with averages: The higher-rated survivors with per-survivor rating calculations would lose 133 for losing, the lower-rated ones 66; now they will each lose 100, meaning the higher-rated ones lose less, which is sensible because they had worse teammates, and the lower-rated ones lose more, which is sensible because even alongside two higher-rated survivors they could not survive against the killer.

If the killer kills the lower-rated survivors but the higher-rated ones escape, the rating adjustments would be +133 for killing the two, -150 for failing to kill the others, in total falling by 17 points. If he kills the higher-rated survivors but the lower-rated ones survive, that's +266/-300, falling by 34. If he kills one higher and one lower-rated one, that's +199/-225, for -26. While that means a 2k still leaves the killer close to where they had been, against mismatched teams the killer will lose some rating in any combination of a 2k; with averages instead, the rating adjustment could simply be 0 in either case, since they killed 2 survivors in a team they were expected to kill 2 survivors against. If he kills 3, that's either 2 low + 1 high or vice versa, in total either +266/-75 for +191 or +299/-150 for +149. With averages, you could simply say the killer won but they did not win decisively, so of the 100 possible +MMR, you give them +75.

The survivor adjustments in these cases would shift from higher-rated survivors getting +75 for surviving and -133 for dying and lower-rated survivors +150 or -66, to all survivors getting +/-0 if two of them survived, and -75 if only one survived.

If the killer only kills 1 or 0 survivors, that's -234 if the killed survivor is one of the lower-rated ones, and -242 if it's one of the higher-rated ones, or -450 if he kills 0. With average adjustments, he would lose 100 if he kills 0 and 75 if he only kills 1. For survivors this would mean rather than 3 individually gaining rating and the dying survivor losing rating, now they all gain rating: +100 if they all survive, +75 if only 3 do.

In conclusion, with averages you could stabilize player ratings more quickly, by increasing killer ratings less if they win against mismatched groups (but also decreasing it less if they lose against mismatched groups, which is less likely however), actually not changing ratings at all in 2k/2e matches, and not having rating be affected negatively for survivors who die despite their "team" winning (but also not positively if they survive despite their "team" losing).

Mind you, over enough matches, the current rating system will already be good enough in that ratings eventually tell a good-enough story regarding the "team" performances too, and you could bring up similar fallacies to try and undermine the team-based survival metric that people do for the per-survivor one (e. g.: "So my teammate dies in 10 seconds, kills themselves on hook, but is SkiLleD because my two other teammates and I end up escaping? Great system, genius devs!"), so don't take this as me saying the current system is bad, it isn't, conceptionally it is already good. But with team-based survival metrics based on the group's averaged rating, you would in theory more quickly arrive at appropriate ratings, and therefore more quickly stabilize people around the 50% winrate target, in turn more quickly and therefore more often leading to fairer matches. In theory because actual MMR calculations are more complicated, and because you don't get neat matches all the time, the players the system has available to make matches from vary greatly, and queue times have to be kept in mind too.

I only brought up crossplay off (PC for me) to say there factually were enough high-level players out there queueing at the same time in the same regions consistently enough for high-level matches to happen at reasonable queue times in my region. This was merely used as an indicator to say that queue times would not necessarily grow prohibitively long for the idea of making the matchmaker more strict. The queue times/skill of players on cross off now, let alone on other platforms or in other regions are not really relevant for this.

SWF queues have been a little longer for me too (duo queues mostly), but as I told you in another post, my survivor queue times in general and on average are not long, and do not regularly exceed 5 minutes. And I have definitely gotten fast lobbies where ratings must have been substantially mismatched, which leads me to believe the matchmaker is not yet actually strict enough (though it could also be that rating adjustments aren't strict enough and people still gradually climb in rating even if they average <50% kill/survival rates, or that the rating cap is too low).

Guess I'll butt in and be the one to tell you "no" again. I believe you are wrong on pretty much every point here. Let's see:

You are not taking into account that the win condition MMR is based on looks at many matches, rather than any 1 match. As OP has pointed out repeatedly, it is difficult for us to understand this, we want to look at any 1 match and have it be absolutely balanced according to the win condition, and can't imagine that a survivor getting a "win" if they escape despite their "team" "losing" or another survivor getting a "loss" if they die despite their "team" "winning" could be sensible... but the math works over many matches and instates an appropriate matchmaking system that does continually nudge toward achieving the desired balanced win condition.

In a 1V4, what is a balanced outcome based on the kill/survive win condition? 2 survivors dying, 2 surviving; 50% kills, 50% escapes. Now, if you have a system that looks at the individual survivors' survival rates, what would that rate have to be in order for a balanced state to be achieved according to the win condition? 50%. This is precisely the survival rate that the system tries to achieve on a per-survivor basis, on average, over many matches. Let's take a series of 100 matches where all survivor players and the killer player are equally rated. What you think must be the case in this series of matches, is that every single match has 2 kills and 2 escapes happen, because that is the balanced state according to the win condition that the system applies ratings based on. What is actually the case however is that if you look at any 1 match in this series, there can be anything from 0 to 4 kills/escapes... important is only that if you take the results of all of these matches together, they will come out roughly to 50% escapes and 50% kills. This can happen if all matches are 2 kills and 2 escapes, but it can also happen if 50 matches are 4 kills and 50 are 4 escapes, or yet if 50 matches are 3 kills/1 escape and 50 3 escapes/1 kill. Or indeed with any of the many combinations of these results that add up to 200 kills/200 escapes.

Matchmaking systems work - and have to work - over many many iterations, and they cannot achieve an actual balanced outcome in any 1 match, ever. No system can actually foretell the future. They can only make predictions, good estimations, and if its mathematically calculated player ratings estimate a 50% survival rate for matches where players are similarly-rated, and you actually get around 50% survival rates over many matches in those scenarios, then it is making good estimations, and achieving on average balanced matches according to the win condition. You are wrong in your completely basic assumption that the system has to make equal matches result in 2 kills/2 escapes. It can't. No matchmaking system can. It can only do math, looking at many matches and average results.

I don't get how you can still argue this point, which sits at the very core of the fallacy OP has now addressed multiple times and already had in very clear and simple terms in the original post.

The most skilled player dying and getting a decreased rating is not what consistently happens, don't you see how absurd this line of thinking is? The player that can run the killer for 5 gens will not be the one dying in more matches than the player that can't run the killer for more than 10 seconds. Sure, there are individual matches where that can happen, since the 5-gen runner gets facecamped and the other survivors decide not to even try and save them... but this is absolutely the exception. Play 100 matches and the survivor skilled enough to potentially run a killer for 5 gens will be peacing out of those exit gates vastly more often than the Meghead or Lockerdwight. Because they can have long chases, which creates time for people to do gens, which in turn allows the exit gates to be opened through which that survivor can escape.

And the absurdness of your thinking goes even further: If the more skilled player actually did consistently die and fall in rating, they would face worse and worse killers, but yet still die? And the less skilled player consistently survives, climbs in rating, faces better and better killers, and keeps surviving?

I think the flaw in your thinking here again comes from imagining singular scenarios and not getting that the system is about the big picture. In your mind, this player runs the killer for 5 gens in every single of these 100 matches, and in every single one gets facecamped by the now-angry killer and dies. This is not the reality. In 1 match or maybe a few matches it can certainly happen that the survivor runs the killer for 5 gens, that the killer then is angry and facecamps them, and that their teammates do not try or manage to save them. But in most matches this won't happen: the player might not be facecamped, the player might get saved despite being facecamped, the player might only be chased for 4 gens, or 3, or 2, or 1, or they might not be chased at all... What matters is that this player is so good that they can potentially run the killer for 5 gens, and this will, on average, lead to them escaping more often than not. Certainly, they will survive much more often than less-skilled players would in their position in their matches, let alone the least-skilled players. Over many matches, the edge-case scenarios where the player so skilled they can potentially run the killer for 5 gens will actually run the killer for 5 gens, and then actually get facecamped, and then actually die, will matter less and less, since they will account for fewer and fewer of the matches.

Any low-skilled survivors that escaped in rare matches thanks to a 5-gen-run-into-facecamp event of another player they were lucky enough to be matched with and "lucky" enough that the killer chased them all game and then facecamped, will over many matches die much more often, since this event will not occur often and their lack of skills will have an impact on the rounds and their survival rates will decrease.

Something more to note: Over many matches, these players will be paired with different players all the time. You might be assuming a series of matches we are looking at would always have to be this group of low-skilled players with the high-skilled player from the original argument scenario (and even if that were the case there is no real reason to expect the high-skilled player to be dying more often than the low-skilled players in them), but that of course is not the case in this game, matchmaking almost never gives players the exact group twice.

Over many matches, survival/kill rates correlate very closely with the skills needed to survive and kill in this game, such as being good at chases.

First of all, it is absolutely possible to get games in which not a single down happens. I'm not saying it is what usually happens, but it is not objectively absolutely impossible that it can happen. If the survivors are so much better than the Bubba, they can potentially not go down until gens are done and gates opened for them to leave through. This already disproves your point that skill can never matter in this instance, but let's go on.

If a survivor runs a Bubba long enough for the facecamp to only result in that 1 kill, their chase skills are pretty good. They themselves did die in this round, but you are again failing to look at the bigger picture. In many of such rounds, this player will not always be the one getting chased first, or at all. Imagine in another such match another player is getting chased, dies earlier such that after the facecamp Bubba still has time to go for more... Now can you imagine that the chase skills of the player we were talking about earlier may allow for them to hold the chase long enough in order to be able to survive? What if prior to that the Bubba already chased and facecamped 2 other survivors? Or 3? You have to get away from your narrow viewpoint of singular rounds when discussing a system meant fundamentally to deal with many rounds, whose entire purpose and functioning indeed is only relevant when looking at many rounds.

Just like the first point, you are just factually wrong here. It is objectively absolutely possible to get a hook rescue against a facecamping Bubba. If the Bubba is bad, this is actually pretty easy for experienced survivors, by tricking Bubba into releasing his chainsaw attack with unhook baits and unhooking during cooldown. Flashlights and firecrackers/flashbangs can be used to force Bubba to look away or blind them, potentially being able to make unhooks happen during those brief moments. For another trick, I have not seldomly made saves against facecampers happen by 99-healing myself, baiting an M1 and healing up just in time. Styptics and Mettle Of Man could also do the trick.

Often it is also possible to simply unhook before the attack is charged because Bubba can't continuously keep revving since it will trigger tantrum after a while. Perks like Leader, Resilience and Desperate Measures can be of further help, since they increase unhooking speeds. If the rescuer has BT or the unhooked survivor DH, SB or DS they are usually safe. The brief invincibility frames after getting unhooked can potentially also save them without accounting for any perks, especially if the unhooked survivor uses the hit sprint well and the Bubba gets confused or smashes into the hook triggering a tantrum. The Bubba might also slug them expecting DS, allowing the unhooked survivor to get healed up or recover with the help of perks.

But even against the best facecamper Bubba in the world, you can make rescues happen: one survivor goes for the unhook, eats the chainsaw, another survivor comes in and saves during cooldown.

Even if that were the case and no counterplay to facecamping Bubba existed whatsoever such that the player on hook will always be the one dying, you are already disproving your own point in principle by saying rescues are possible if "you are at a very low level of play", which is literally what the rating is then also suggesting and meant to suggest.

See, this is something where I would be like, if it's actually impossible to save against a Bubba (it isn't, but it's certainly less possible than against any other killer), and if that's actually a balance problem (it isn't, Bubbas struggle to 2k against good teams, and facecamps will often result in 1ks at that level if they don't get an early down, but it's certainly a problem in the average pub match), then... we should change Bubba, not design a matchmaking system that can somehow account for this outlier. Which we should anyway, and in my opinion we should even rebalance camping altogether, decreasing killer strength around hooks in general, compensated for of course by buffs to other aspects of killer play.

The "Bubba facecamps 5-gen runner" argument is flawed because it only looks at a singular, exceptional match; it is conceptually incorrect because it assumes that rescues are absolutely impossible; it is disingenuous at worst and misguided at best, because Bubba is 1 killer out of 26, with an overall occurence rate of 4.37%, of which only a portion are facecamping Bubbas, and we should not adjust an entire matchmaking system to this, rather adjust Bubba to not pose the problems that he does.

I'll leave you with this: BHVR has (in the Q&A and now in this thread again) said they will actually change the metric for the win condition from looking at survival of the individual survivors to survival of the group of survivors, so basically... little of this discussion will even be relevant anymore, since that will resolve most of these points even more easily.

What's your build?

The flaw I'm seeing in your reasoning here is you've fallen into circular logic. To justify considering the game a set of 4 1v1 deathmatches for the purposes of determining a win and a loss you have argued that the ideal survival rate for survivors is 50%. However, the only reason 50% would be an ideal survival rate is under the assumption that the game should be considered 4 1v1 deathmatches. If we instead consider the game a 4v1 and define some arbitrary team-based win condition it is quite likely that 50% would not be the ideal survival rate. Just as an example, if we were to say the win condition of a match for killer was to kill all four survivors such that if they do not kill all 4 survivors they lose, and if they do kill all 4 survivors they win the ideal survival rate would no longer be 50% as that would result in killers losing much more often than they win. Under this definition the survival rate ideal for balance would be closer to 33%. Now to be absolutely clear, I do not think this would be a good win condition but it does show that the ideal survival rate for a balanced game depends on the win condition we set. You cannot justify treating DBD as 4 1v1 deathmatches with the expected survival rate that would generate. 

I feel like I haven't been ambiguous about my claims here. I fully acknowledge that an MMR system cannot work perfectly in every game and the most important factor is the overall performance over hundreds and thousands of games. My claim is that this implementation does not and cannot achieve that. My example cases where the higher skilled player gains MMR and the lower skilled player loses MMR are not flukes nor are they exceptions; they are the average experience. Yes an actual 5 gen run would be exceptional but it is a fact that the longer you can run a killer the more likely you are to be the survivor currently running the killer when the last gen is completed. Thusly the less time you can run a killer the more likely you are to be doing a generator when the last generator is completed. For the sake of my arguments I am assuming completely rational players playing to optimise their total kills. I'm not even considering the possibility of killers getting angry or upset at having been looped for 3 gens and just giving up to facecamp even if it is a real factor. This isn't even the only example. When a teammate is hooked the players who act quickly to save their teammates are more likely to end up dead and the survivor who hasn't even noticed anyone got hooked and simply sits on their gen is more likely to survive. The survivor who fails to take aggro from the killer to save a survivor on death hook is more likely to survive than one that does. This isn't just unnecessarily altruism they are the optimal choices to 'win' the game from a 4v1 perspective but they become actively selected against by MMR based on a 1v1 perspective. This implementation of MMR actively rewards bad gameplay decisions and lower mechanical skill and actively punishes good gameplay decisions and higher mechanical skill. And this is true as a general trend not just some niche exception.

For the Bubba issue I'm genuinely shocked this is even something people seem to want to debate me on. At the very least you accused me of failing to consider the big picture but now you want to talk about the extremely unlikely possibility of saving against a facecamping Bubba. I don't even need to engage with you on whether or not it's possible. If you are going to argue that we must look at the big picture then you cannot also bring up exceptional circumstances. In the big picture this strategy is not defeated enough to be statistically relevant. As is true for many other strategies to confirm kills. I will say that while I acknowledge you can rescue against a facecamping Bubba at low levels of play a good MMR implementation must be effective across all skill levels (as I mentioned before) not just low levels of play.

As I've explained in my previous posts Bubba is not an outlier in this regard he is simply the easiest to do it with. It requires no skill whatsoever for a Bubba to facecamp and kill you, and your individual skill is at best irrelevant and at worst actively harms you. It is the ultimate example of my point. Other similar cases have some nuance or exceptions. There is some unrealistic counterplay to STBFL camping, it's kinda possible for someone being RBT tunneled to find hatch, maybe a Plague will mess up their vomit and it's actually kinda possible for a Billy to miss their chainsaw etc. And I really don't think you can try to narrow down my point to just 'facecamping/tunneling bad' it's merely a single example of how this particular implementation of MMR fails. I can and have given other examples on how the system rewards bad play and punishes good play from the context of DBD being a 4v1 game.

To directly quote the QnA here "However, we're looking at some improvements to the current system. An example is to take team results into account with survivors, in addition to the personal result."

BHVR said they are looking into some improvements such as a hybrid system that considers individual survival and team survival. They did not say they would replace the existing individualised system. So this discussion is still relevant because the hybrid system has not been implemented, we have no details on how it would work, even if it was implemented it would be worse than a pure team-based implementation and there is no guarantee it will ever be implemented anyway if BHVR decides they don't like it or it would be too hard after looking into it.

I'm sorry, but the flaw in this reasoning here is that... it makes no sense. The win condition the devs are looking at is kills versus escapes, and 50% kills versus 50% escapes is literally a balanced result for that metric. Sure you could argue one should make something else the win condition, or that we do not want to design the game around balanced kill/escape results and rather want 3 or 4 kills to be the desired result, but that's a different argument altogether. A 2 kill/2 escape result is seen as balanced, and that means every individual survivor had, mathematically, a 50% survival chance, and will, over many matches, average a 50% survival rate in balanced matches.

Your next section is honestly just as much of a mess.

I have no clue how you conclude this is a fact, but you are free to demonstrate this with factual evidence.

Also boggles my mind how this could be a fact.

If anything the less time you can run the killer for the more likely you are to be dead when the last generator is completed, if that ever even happens in that match.

This is also completely backwards thinking. One, the players that actually rescue their fellow teammates are much more likely to survive, since simply leaving people to die on hooks no matter what is regularly not actually a play that increases a team's overall survival chances. Two, in cases where it is more likely to increase the team's overall survival chances to not even attempt rescues but stick to gens (such as in certain facecamp scenarios), the decision not to go for the rescue is the correct decision, and making this decision makes the player better at surviving than a player that would go for the rescue despite this. Three, saving teammates quickly is not always the correct play, often you want to wait a bit, for various reasons. Four, unhooking people does not take any more skill than doing a gen in and of itself. Sure, if the killer is camping it does (and even this could be contested, as "doing gens" still involves a fair bit of decision-making skills, such as which gens to repair, when to do so, when to let go, as well as at least some mechanical skills, since constantly hitting great skill checks does make a difference, and since being good at pressuring gens in killer proximity is a dynamic similar to contesting hooks), but then again going for the save might not necessarily be the correct play and the player doing so is not playing in good ways to survive, has worse decision-making skills. MMR is about skills that lead to victories, survival skills, not just doing anything that is difficult even if it doesn't increase winning/survival chances.

Again, you seem to neglect the simple reality of this game that if even just 1 survivor dies, the survival chances of the remaining survivors decrease dramatically. Players that are not good at teamplay that increases their fellow survivors' survival chances, will also have decreased survival chances themselves, and will survive less often on average. There are exceptions to this, but that's what they are, over many matches this will hold true.

This is especially the case because think your flawed thought to its logical end: Who would populate high-MMR environments if what you said were true, what consistently happens and not an exception? Players that avoid chases and are not good at them, selfish players that do not rescue fellow survivors from hooks, do not take hits for them or otherwise play to help them, just sit on gens. They would be paired up with each other frequently. And they would still survive, against killers that are increasingly better at killing. ???

Or its other end: Low-MMR environments would see players frequently get matched with each other that are good at chases and contribute a lot to their team's survival chances, facing off against killers that struggle to kill more and more. Yet they would still die? Don't you see that even if it were for some absurd reason the case that such players consistently die more often than their teammates, they would at some point meet teammates just like them, and end up on the "receiving" end of surviving due to those other players' contributions, and do so on average equally?

Even if I were to accept that this is a fair parallel to draw (it's not, I used "the big picture" for a completely different argument where it is actually relevant; even if it were objectively impossible to save against a facecamping Bubba, that would not at all mean it's always the better survivors that die against Bubbas, and worse survivors that survive, there is no real reason to expect this)... May I remind you that Bubba is 1 of 26 killers, being played in a mere 4.37% of all matches, of which in turn an even smaller portion actually see those Bubbas facecamp?

Against good teams, primitively-facecamping Bubbas average less than 2 kills. In tournaments where good players camp with Bubba in smart ways they barely average 2. This strategy can be defeated from a 4V1 perspective. And again, before you say "but those 1-2 survivors that die, do definitely die, and they are also the best players in those rounds!", statistically, over many matches, the players that die in those rounds will always be someone else. And if there are worse survivors in such matches, they are more likely to die than better survivors would be (note again: those better and worse survivors do not necessarily have to be together in those same rounds, it can also be 4 worse survivors against that Bubba, who are then all more likely to die - they will have individually lower survival chances and therefore survival rates over many such matches than 4 better survivors would in their stead, or than mixteams of better and worse survivors (in which again the better survivors would however still not be more likely to die than the worse survivors either)).

Better players that are more mentally apt at making the correct decisions and more mechanically apt at executing the according plays survive more often than worse survivors, against Bubba and any killer, even if facecamping would be impossible to rescue against, and over many matches, the better players will have higher survival rates than worse players, against facecamping Bubbas too, even if they themselves sometimes end up dying to a Bubba facecamp.

If you want, I will again point to the distinction that "better" players with regards to MMR means "better at surviving/killing" players. I don't think that distinction is very meaningful in this context or much of any context, because while there are certainly difficult things you can do in this game that do not contribute to surviving or killing (or are even detrimental to them), the most important skills that most people actually care about definitely directly contribute to surviving or killing. Such as chase skills. I have no clue how you can possibly assume players that are better at chases will die more often than players that are worse at them. The idea that they would die more often because they are more often in a chase when the last gen pops than players that are worse at chases is so deeply flawed I don't even know how to attempt to convince you otherwise.

Your individual skills will 1. increase the duration of your chases against Bubba, in theory even so much that you never go down in the first place, and as such cannot end up getting facecamped, 2. lead you to make decisions that are less likely to result in you dying against a Bubba, such as going for rescue attempts if he is facecamping someone, 3. allow you to actually make rescues happen against facecamping Bubba.

Your skills will also allow you to unhook yourself more often. KOBE!

...

Bubba is definitely an outlier, no other killer can facecamp like he can. ...But for the other also-problematic killer (or of course survivor) interactions, I too say: adjust them instead of making the matchmaking system somehow account for them! Even if the MMR were to be based on group survival rather than individual survival (which it will be! kinda, apparently...), the player on hook against a facecamping Bubba still is not having such an engaging time playing the game, are they? This should be changed regardless.

So far none of your examples have convinced me. Over large enough sample sizes, individual survival rates (and kill rates) will correlate very closely with good, correct, optimal play, including teamplay.

But regardless of that, you still seem to think of the system as punitive/rewards-based. There are no rewards or punishments, rating changes do not actually harm you in any real sense, they just affect who you will face in the future. Say you are actually consistently the best player in your groups, and yet somehow consistently die... Don't you want to face lower-rated killers that are less killy, such that you can survive more often? They will probably be less facecampy too!

This is fair. DesignDad did also say "Our hope is that we can hybridize the system and use both team results and personal results to generate an even more accurate MMR that contains the benefits of both approaches!" in this thread, so you are right about this.

Would be interesting to know just what they have in the works, and why they won't just switch to group survival entirely, but yeah, for me that doesn't make much of a difference because I already am convinced the concept behind the per-survivor logic is sound and works, only that group-based survival (and rating adjustments made against group average) will more quickly deliver desired results, not however different results.

I think this discussion can end at this point anyway though - we disagree on some fundamental levels, and since what has been said so far is rather puzzling to me, either my thoughts on this are correct or I'm not smart enough to understand that they aren't and likely won't. Ha. So with taking into consideration that ultimately none of this discussion will matter because BHVR will just do what they want to anyway, and because the prioritizing of queue times as well as the reality of the pools of players queueing in respective regions and at respective times will always undermine our striving toward actually having equally skilled, experienced, "sweaty", etc. players in most matches, I think it's not worth continuing this. Maybe OP (or even a dev!) will take you up on this again though.

I didn't really say any of those things you are claiming I did.

I'm not saying you are lying, about anything. You might very well be at high MMR as Nemesis, and are facing high-MMR survivors. Just, we objectively cannot know what our MMR actually is, not without external software. That doesn't mean one cannot make educated guesses as to this, I myself generally assume I'm high MMR based on my consistent survival/kill performances over many matches, and on the fact that not few of the players I get in my lobbies are good at the game (I can see this in the gameplay itself, but also because not seldomly I know the players, from tournaments, streams, or just having seen them around a lot in the game, and I can also look at their hour count, which is often in the 3000+ realm, and at least shows they are very experienced). I use these as indicators to say I probably am high MMR, but I would never say I know for a fact I am high MMR, because I simply do not know this for a fact.

Maybe contact support, who knows, perhaps they'll tell you what MMR your Nemesis is at? Doubt it since that would set a precedent and they'd get flooded with requests, but shoot your shot I suppose.

For things that can at least better indicate whether you are at high MMR:

Look at the hour count of the survivors you get matched against. Ask them whether they are surviving often.

Record your games or stream them. It will be fairly easy to tell whether your opponents are good at the game.

The basic assumption of MMR is actually that matches should be balanced in their potential for the players in them to be able to succeed, with equal chances to kill or survive, not that the outcomes of those matches are actually always balanced outcomes. Other games use MMR/Elo systems too, like chess, and the players at similar MMR ratings do not actually have perfect 33% win, 33% draw, 33% loss ratios against one another, let alone 100% draws.

Take Maxime Vachier-Lagrave, FIDE rating of 2761. Facing Shakhriyar Mamedyarov, at 2767.

That's 13 wins, 19 losses, 30 draws. Close, but no cigar.

Or what about Magnus Carlsen, against Liren Ding, two of the highest-rated players?

20 wins for Carlsen, 8 losses, 29 draws.

Obviously DbD is not chess, it poses unique "problems" for MMR systems since it is asymmetrical and does not have the same starting positions for all players (chess has black and white pieces, but DbD has different killers, items, add-ons, perks, maps...). But as long as you define clear, diametrically opposed win conditions (and draw states), that players actually play for, the MMR system will work well to match more equal players with regards to their ability to achieve those win conditions, both in their skills and performances, but also in their use of the different add-ons, perks, etc.

As opposed to what somewhat argued a bit earlier, a certain perk being behind a "paywall" is not actually unfair in the context of this system, since either a player performs well enough without the perk to be able to compete with players that do have and use the perk, or, if the perk is actually necessary to be able to perform that well, they will simply settle at a lower rating and face players that likewise do not have or use that perk, or aren't good enough to perform better than them even when using that perk.

Since DbD as opposed to chess is not actually perfectly balanced in concept because you have two sides with different objectives (well, chess isn't really either, but if you play 2 games with both players getting to start as white and black once it is), matches of similarly-rated players might not actually lead to balanced results of course... but for that the system can account as well, by looking at many games and finding a ratio between survivor MMR (and be it of a group average) and killer MMR that does on average yield 50% kill/survival rates, and thereby equal chances to succeed.

No, you can't infer that from what was said. Teams would likely perform better with Kobe Bryant on them. That's why teams that had him on them were consistently at the top (in the NBA, even winning the NBA multiple times).

I don't know why the players that make "clutch heals", stealth when beneficial and bang out gens would be expected not to survive more often than players who don't. Or in fact why they could not be players that can also lead killers on long chases. And the better they are at such things and the more consistently they manage to do them, the more often they will survive, even against campers. Definitely much more often than worse players that are not as good at these things and do not consistently or at all manage to do them.

If their actions have actual value for the survival chances of players in their games (which includes themselves), they will affect their average survival rate, and that would be true regardless of MMR existing altogether. Just because we now have an MMR system in place that increases a player's invisible rating in the background if they survive and decreases it if they do not doesn't change anything about the "value" these actions did or did not have, for the players' survival chances but also their gameplay enjoyment, or whatever.

Since you use the word "negated" though, I want to at this point again remind that MMR doesn't actually reward or punish players. If you hide, do gens, but end up getting caught and camped without getting rescued and lose a bit of rating, at worst that means you will face less tough killers against which you can actually survive more often by hiding and doing gens.

But why would you assume to begin with that a team with a player that consistently gets many points a game consistently loses? And sure, say they do lose consistently despite having such a good player, if they would face ever-weaker opponents as a result of losing (such as by dropping out of the top league), they would surely start winning at some point?

This also neglects that in DbD, you have constantly changing teams, meaning that it's not only one team that somehow manages to consistently lose despite you consistently performing very well, but many many different teams, and against many many different opponents, who are also getting gradually weaker...

You brought up basketball, and I'm not an expert on it, but there certainly are many different metrics that people look at for players, many different roles, tasks, skills that go into it... But they still decide whether someone wins by looking at points exclusively, and specifically who scored more points by more often putting the ball through the basket. Shots from afar count for more points, but... that's about it. Teams win on the basis of putting the ball through the basket more often than their opponents. Entire championships are decided on this silly little metric, with no regard to anything else. And the champions are usually widely regarded as the best. Odd that.

The matchmaker definitely has to actually take ratings into regard for it to matter whether the ratings make for good matches, but it's not like the matchmaker disregards them completely, and discussions like this can also help encourage the devs to perhaps tweak the "queue times vs. rating accuracy" button a little more into the latter direction.

They are specifically fixing that slot-filling issue, so that's something. They are also looking at a few more things, like cross-adjacent-region matchmaking.

There are some issues though with your analysis and that of the developers when looking at for instance Hockey and taking the parallel when establishing the MMR that should go up or down. The individuals in the team survivors are not judged upon their teams performance, yet are evaluated on their individual performance as if that is the case.

Take the examples of running a killer for 5 gens, then being the first and only one to die. The overall score for the survivors is a win, 3 people got out and their MMR increased. The killer lost the match and loses MMR based on the outcome of the match and yet the star player of the team goes down in ranking? Meaning that their match up becomes worse, now they have to run the killer for longer as the other survivors they are match up with are less high and therefore their team mates are less sufficient.

If you want to take the idea of it being similar to sports, then you would for the survivor end need to establish a performance based rating system. Even in Hockey or Basketball player ratings and their team standings do not have a 1 to 1 correlation and they have a set team. The better teams have usually more high ranked players, because their team consists of actual high skilled individuals is more likely to win. While the goalkeeper might block 90% of all shots taken, if their attacking teammates are incapable of landing or even taking a shot... does this mean that the goalkeeper is not skillful? No... they are and in sports is highly likely to be bought by another team, as their personal rating is significantly higher than that of the rest of their team. The idea that the contribution to the team is insignificant for your standing is false.

The overall behavior that is rewarded is safe and selfish play and not teamwork, imagine why people don't like solo queue? Let the other person run the killer, do the gens and let them hang on the hook to die while you escape. You claim it doesn't matter to climb the ranks, but that is also false as the goal is to be teamed up with others around the same level of play and now the individual that was the star can get placed in worse teams rather than better ones?

The argument falls flat to compare this to team sports, as the survivors are judged on an individual basis instead of the teams overall performance, while the star player of the team might be punished for their skillful play that secured the overall win for their team. As in they ran the killer for 5 gens, then got left behind by the other 3 survivors.

If BHVR doesn't want to expand the criteria for the survivors, it most likely would be better to at least treat them as a team and have their results be a reflection of the teams performance. As it currently stands, they aren't identifying skill in survivors to climb the ranks, they are identifying safe players. If you are the first to be found, you are more likely to die than being the 3rd person to be found within a match? Regardless of whether that first person ran the killer for longer or not.

For a killer you can argue that kills are perfectly fine to use as a benchmark, they are the team and the final result of kills vs escapes falls solely on their shoulders. Yet for survivors the term "skill" is not applicable to the analogy used.

If the whole MMR system is just reducing us down to a number to boost our MMR then why can’t it in another way that doesn’t account for kills/escapes only?

How is there no way to take the same thing that determines your pips, and implement that into MMR?

Each medal has its own set of numbers, kill/escape medals can be set at a higher number value than the others, and the MMR system could average out the number for the whole game and reduce you down to a number in that way, no?

The current MMR system rewards selfish players, and tunneling/face camping (don’t get me wrong, it’s a completely valid strategy, but killers should have the option to play without doing that to achieve kills).

Altruism is a big aspect of the game, so is looping, and obviously so are objectives. So why have you guys designed your whole MMR system around escapes/deaths, when you guys repeatedly say the game isn’t meant to win, it’s supposed to be more of a “draw” type situation.

The game isn’t fun when you’re constantly being tunneled out as a survivor, and for a lot of killers it isn't fun to have to tunnel one person out just to have a chance either.

Nor is the game fun when you get left to die on first hook, because locker Dwight made it into high MMR too because he hides and escapes often.

You guys say it eventually evens itself out and that it doesn’t happen as often as your player base is saying it happens.

So when does it even itself out? Because myself, and a ton of others have multiple experiences back to back, and it’s not “evening” itself out.

Also, why does the reasoning for not completing nerfs/buffs/changes always consistent of something along the lines of “it’s too hard, why would we do this when we could do that?”

Your player base is reaching out to you with ideas, and while a lot are based on opinions, or are clearly favorable to one side over another, a lot of suggestions are actually very well thought out, neutral suggestions that could really help.

Even within your very detailed description, you yourself threw in, “low blows”, towards your player base, when all anyone is doing is expressing their frustration and making suggestions.

I am 110% on board with an MMR system, I just wish you guys would listen to your players feedback and adjust your improvements based around that.

I won't address everything here since I've been repeating myself enough as it is, but I can give you the major answers in a dev's stead:

"Why not base MMR on pips or something?"

An MMR system cannot (sensibly) be based on metrics that are not diametrically opposed in the win conditions. MMR systems have to compare players against other players and apply rating changes based on their performances against each other. You have to clearly define what a win and loss is in order to decide who gains or loses rating. How could you ever decide whether one person was actually more skilled than their opponent in a chase, for just one example? Do you say, "Well, if the survivor held the chase for... I guess 60 seconds, and only used... hmm... 3 pallets in this time, and did not vault the strong window on map X, Y, or Z, and if they did not use Dead Hard or a Styptic to prolong their chase, yes, then they won the chase and are more skilled and the killer is less skilled".

...Even disregarding how absurd that already begins to be, how do we even measure whether something was actually a continuous chase - the chase state mechanic does not and cannot do this, so would you just base it on player proximity between survivor and killer and whether the survivor is running...? I hope you see how that would also lead to tons of issues.

And then even beyond that, the major issue becomes and with the old pip/emblem-based ranking system actually was that both sides are able to "win" at the same time without diametrically opposed win conditions, both rank up, and with time, you have everyone piling up at the top rank regardless of their actual skill and performance levels, with completely disparate matchups being the norm.

The current MMR system doesn't "reward" anything. If selfish players and facecamping players actually consistently win and their rating increases, the only thing they would "get" for that is tougher opponents (and more selfish teammates).

Altruism, looping and efficient, objective-oriented play are absolutely huge aspects of the game... because they make a huge difference between escaping or dying, the things the game is literally about and designed around. If you look at players that are good at altruism, looping and achieving objectives (completing gens), you will find that they on average will escape more often than players that are worse than them at these things.

For the "fun" argument... This has nothing to do with MMR. The things you argue aren't fun would still exist without and had already existed before MMR. If they are to be changed, it has to be on a base-game level. The MMR system does not actually encourage these things one bit, because there is no MMR incentive for doing those things, you can't even see an arbitrary number go up or down. The fact of the matter is that people that want to win (to kill or survive) always have and always will do everything possible or necessary to do so, regardless of much of anything. And the people that prefer other things, that find things other than simply winning by any means fun, always have been and still are entirely free to play for those other things. In fact, the MMR system benefits them, because now they will less often be paired with and against the players that care much more about winning than them and play very hard to win, instead they will be more likely to find matches in which the other players also aren't dead-set on winning, or at least not very good at it.

The old matchmaking system on the other hand with time pooled all of these players within the red ranks, and what did that lead to? Even average matches saw 70-80% kill rates. That's not fun. Now we not only have closer to 50% rates across all matches, but it can also be assumed that many more of those matches are actually more desirable playing experiences, not only because they are more "balanced", but because the people in them are more like-minded and on more even levels of how they approach the game and how well they manage to do so.

For me and various of the people I know and/or watch, things are at least notably more even than they had been in the rank-based matchmaking past.

I won't try to deny your experience, and I won't say the system is working perfectly (the matchmaker logic has confirmed issues they are looking at and it will probably always be disregarding ratings more over favouring queue times, and the ratings themselves might also not yet be harsh enough, the cap might be too low), but if you take an objective look at your kill and/or survival rates, perhaps you will find they aren't actually as bad as you might assume they are?

Maybe your rates are too low (or too high), and of course, killing and surviving is not everything that matters... But if it actually doesn't matter that much to you, why do you care that the system looks at those metrics? Just play however you want and for any goals you personally deem worthwhile, and if you then die more or kill less, chances are you will less often face tunnellers or survivors you would "have to" tunnel in order to have a game on your hands against.

Guess I ended up repeating myself again after all...

Sounds like you should be a dev with these guys. Maybe they are hiring?