OhTofu on MMR
I don’t usually post about videos but OhTofu had an interesting video about MMR this week where he breaks down what he considers faulty arguments against the system and arguments against it that have some validity. A lot of what he says I’ve said independently at one point or another so I wanted to give the vid a shout out.
I’m not going to try and bullet point the whole thing, but TLDR you can tell MMR actually does work because of people complaining how they have to “sweat” to get 1-3 kills. Having to work hard most matches to get 1-3 kills as a killer means you’re actually tending to go against survivors your own skill level. Likewise MMR doesn’t force you to play any particular way, but obviously if you play suboptimally then you shouldn’t expect to be winning when you happen to be matched against survivors who play better and have more effective loadouts than you. There are issues with MMR, such as nobody having any idea what their MMR is so everybody complaining about it is grasping at air what the numbers really are, but by and large the system does produce closer matches on average than the old emblem system.
The strongest argument he feels that is valid is that MMR just highlights that top level playstyle and meta isn’t as fun as less effective styles. In other words the more you win the more you encounter the same builds and playstyles and killers, and either you tend to lose to those (which obviously losing is less fun than winning) or you mimic the meta to keep up. Again, this means MMR is giving a good estimate of performance, but it’s making it even more clear where balance issues and similar problems exist at the higher level of the game. Nothing prevents you from running whatever fun build you like, but the more you win with that build the more you come against better loadouts that tend to thwart it. You have to be prepared to “win less” to have “more fun”, that’s the paradox and problem which MMR is bringing to light. It’s not even so much a problem with MMR itself, but more an underlying fundamental issue with how DbD as a game is designed in the first place.
Comments
-
As a survivor with over 1000 hours in the game I've gone against killers with under 100.
As a killer with over 1000 hours in the game I've gone against survivors with 20 hours.
What?
The complaining about MMR is not arbitrary, or random.
20 -
I agree basing the MMR on kills/escapes and time in a match is lazy and terrible. However The point that higher level play leads to sweaty games is valid.
That being said, to quote Choy in one of his recent videos, this highlights the meta is very stale. If your goal is to win at high level (escape as a survivor or kill 3-4 as killer) then your perk choices are limited. This is a problem and why the game is being described as unfun by so many streamers and other players in high level.
2 -
I don't think sbmm is as bad as what we had before but the current system is definitely flawed.
Me and my gf are doing otz's no perk killer challenge atm which I'll post the results of that when we're done and something that's immediately obvious is that kills don't tell the whole story of a game by any stretch. Just to give an example her nurse game was a 1k but she had gotten 8 hooks and had actually done pretty decently but just going by kills which us all the average cares about that was a poor game. Where as in my hillbilly game I got a 3k which involved only 3 hooks and letting a person bleed out on the ground, that was not a game I felt good about at all but for the purposes of the challenge and mmr those things just look at 3k and the spirit of cote goes "you did a pretty good job so far".
I definitely think hooks should have more weight in the current system.
3 -
Actually it’s a bit inaccurate when people say it only revolves around escapes and kills. Escapes and kills determine whether MMR goes up or down, but how much it goes up or down depends on some other factors like relative player rating and the order players die (apparently the first death is weighted more than later deaths because once one person dies it increases the probability a later survivor will also die). It also possibly takes into account the length of the match for the amount things go up or down (I think, I seem to recall that being in a data pull 🤷♂️).
And as far as the other factors, by and large escaping and killing are the most significant ones. Gens are just a tool to escape. Chases are just a tool to help allow the other survivors to finish the gens so ultimately you can potentially escape. So measuring escapes already indirectly incorporates those other factors.
2 -
The devs have already indicated that brand new player MMR is starting out too high. Right now they start around average, they are going to lower that starting point.
Also even if MMR is 100% accurate you will still get mismatches simply by virtue of the game needing to put people in who have waited a while and to quickly fill lobbies where someone left, as well as swfs where one friend is significantly higher level than another.
3 -
And all of these glaring issues debunk the idea that MMR is healthy or in a good place. How do you think that killer felt when me and my friends prevented him from getting even a single hit? How do you think that survivor felt when my Nurse put him down like a '71 volvo in a trash compactor?
Those games aren't just devoid of fun for both parties, they're soul-crushing for the loser.
Also, I'll believe they will fix it when they actually fix it. Not a moment sooner.
3 -
Aside from new player initial rating being high, those “glaring issues” existed in the emblem system too, and also would exist in the similar systems people say they want.
5 -
Does this person play killer? What killers do they typically play? What are their games like? Builds? Etc
2 -
OhTofu has thosuands and thousands of hours on killer.
2 -
Strawman? I'm not saying the old system is good, and I'm not saying the ideas presented by this player base are good. The MMR we have now is awful. There is almost nothing desirable about it. Full stop.
2 -
What killers does he play on average today? What builds is he playing against? How do his games end?
0 -
It’s not a strawman, you’re saying MMR is bad but assuming it’s because the underlying ratings are wildly inaccurate. I’m saying the issues with MMR aren’t the numbers being accurate, it’s to do with factors outside that like being put into first available slots when you’ve waited to long that would be inherent in any matchmaking system.
1 -
He plays all killers and also plays survivor regularly. He’s one of the top DbD streamers, I’m surprised you haven’t heard of him, he even does commentary on big tournaments.
1 -
A system that struggles to find an equal opponent for you so much that it puts you against someone with a wildly incomparable rating is BAD.
You're phrasing this so weirdly. "Our rocket that doesn't fly to the moon totally flys to the moon, they just gotta fix the engine first. And the fuel injection system. And the faulty clay shielding."
Edit: It's also still a strawman dude. We don't need to draw any comparisons to previous or alternative systems to understand that the current MMR, as we have it, is BAD.
4 -
Usually watching someone else or not at home 😅 I'll follow him
0 -
Survivors with low average chase times will die more often on average than ones with high average chase times so tracking their deaths does indirectly correlate to their chase time.
1 -
I agree with most of his points, especially about how MMR seemed much better in the first month after release and how now it's just throwing all players together with seemingly zero logic.
As much of a pain as MMR has been at times, it's a necessary evil. The devs can't make changes without good data behind them, and they can't have good data without good matchmaking. You need a controlled environment to see these things.
I hammer the lack of nuance regarding skill calculation on the survivor side, but it's better than what we had. And there's unfortunately not a great way to quantify game sense/positioning outside of looking into how often a particular survivor's entire team dies relative to remaining gens. A spectacularly dumb player probably has a high correlation with 4ks at 3+ gens.
I think they're trying to have their cake and eat it too with the system though, and it won't work. I still maintain that hiding personal MMR does more harm than good for the community. And you can't have good SBMM without significant wait times for outliers. Some players are going to have to wait for good games.
4 -
I like tofu, but trying to be an enlightened centrist toward a fundamentally flawed and overly simplistic system is just dumb.
9 -
"As much of a pain as MMR has been at times, it's a necessary evil. The devs can't make changes without good data behind them, and they can't have good data without good matchmaking. You need a controlled environment to see these things."
That logic fails on account of how, for all intents and purposes, MMR is a worse matchamking system than what we had before on account of how it doesn't do its job. The goal of MMR-based matchmaking is simple: Put players in matches with players of a similar MMR level. And it DOESN'T DO THAT. It does not do the one thing it was designed to do. The environment is even LESS controlled than before because players are not getting matched with their own MMR brackets even close to consistently. DBD's matchamking is not good, so it will not give good data.
5 -
That is true. For the first month it seemed much better/tighter than what we have now, which is completely random. But the framework is at least there for this to work if they ever get inspired.
1 -
Yes because that’s the exception, not the rule. Most of the time if a survivor can run a killer around like that they will escape.
1 -
It actually does put more people in closer matches than the emblem system did on average.
1 -
Yea...I'm sorry but I'll gladly disagree for the most part. Especially the 'the mmr system tends to produce closer matches on average'. The system focuses on escapes and kills. Thats it.
I consider myself to be a pretty good survivor/teammate but I play solo. I body block, I take hits, I trade hooks if needed, I go back and try to find NOED if needed, etc. IDK how many games I've had recently (thus I'm taking a break) where my teammates seem to be completely unaware that gens = escape. Teammates that go down 10-20 seconds into a chase/immediately, or who dive bomb others on hooks and snowball the game even more. Just the other night, I had 5 matches with teams just like this in a row. 6th match was decent, 7th match again I got matched with new/bad players. Hopped off after that.
MMR system sucks, its faulty as hell, and no one should be defending it. The fact we can't even tell our OWN MMR is even worse and makes even less sense.
6 -
The whole point here is this game shouldn't be competitive seeing how handicapped so many of the killers are compared to that of the survivors, especially in pairs or groups of 3 or 4.
Every m1 killer feels the sluggish movement in chase. Evert m1 killer is a walking simulation on every survivor for the first several minutes of the game bc of pallets being all up, and speed perks like lithe, SP, etc.... then add in dead hard when they finally do catch up.
There's no competition against most groups that sit on gens while 1 or 2 take up 3 or 4 minutes of the killers time for 2 hooks.
Then the game ends. 5 or 6 minutes in. Killer has 3 or 4 hooks total. A hundred broken pallets, and feels like he sucks when in reality he doesn't suck necessarily, it's just the handicap of that particular killer.
Throw in survivor egos and you have a miserable time bc these poor bastards go in thinking they have a real fighting chance at dominating matches bc it's set up to be competitive yet devs slap on loads of obstacles and second chance perks
It's all designed to dely the time it takes for killers to actually down each survivor. It only takes 5 or 6 minutes to pop all the gens.
It's a losing battle for most killers from the start with the way its set up at the moment.
I feel for anyone that goes in thinking they are going to 4k and do well. It's a gamble.
Mmr is good to me for the tier down aspect. (If it works) in theory, if you have a terrible time, and lose massively, the next game should be a little better and so forth
The old system wasn't like that. This in theory should help you have better matches the more you lose. Etc...
Does it work? Who knows? I had a good day yesterday.... will I today? I dunno. We'll see.
All I know is I used to have full days where I got stomped every match to the point it was comical. Mmr should fix that.
4 -
There's no merit in a system based solely on kills and escapes and ignores every other factor that is key to determining skill/competency like chases, hooks/unhooking, gen progression/control etc.
I still can't believe it took them two years to come up with this.
6 -
It usually does. Survivors that can do really well in chases also correlate to survivors who know to do gens and to go to the exits when the gens are done and who can get out in the end stage once the doors are open. The survivors who know how to do those things will on average escape more than the survivors who don’t
1 -
...That's not a strawman, at all.
I can attest that as more of an intermediate player who prefers to play a bit chill, most of my games are way better than they were under ranks.
Actually:
That's a pretty good example of a strawman. Or at least an analogy so faulty that it's pretty damn close to one.
He's saying that it's an improvement. He's not saying that it's perfect. No MMR system is perfect.
The issues with ranks were:
- Killers had zero/little matchmaking protection for much of it. You can confirm this if you look at my very earliest posts, where a dev responded and talked about this.
- Killers would get matched against a much wider variety of skills. This is easily confirmable by Otz's 'streak' videos, where he was able to get 100 wins with no addons while now he really struggles to get 20.
- It was so easy to smurf your ranks down that even fog whisperers were openly doing it.
SBMM isn't perfect. But 'perfect' will always be the enemy of 'good', and the nemesis of 'better'.
6 -
I'll believe that statement when I see it.
And considering I, you know, play Dead By Daylight, I should be seeing it all the time instead of literally never.
3 -
Having the framework there doesn't matter if the framework sucks. It's barely functional, doesn't do what it's designed to do, and, frankly, doesn't even work in Dead By Daylight. Seruosuly, who thought it was a good idea to put skill-based matchmaking in one of the least skill-based PvP games on the market?
3 -
And why would you assume that the survivor who is best at not getting downed by the survivor isn’t one of the three people doing gens most matches too? The person getting downed is usually the weakest runner of the group, not the strongest, leaving the stronger players to be the ones doing the gens.
0 -
His TLDR was that MMR does work. My TLDR is that it doesn't. He responded to me by comparing it to other systems as a direct improvement. I'm not sure what's complicated here.
I never once responded to the idea that it's an 'Improvement'. I'm sure I could devils' advocate in favor of some things that the old emblem system brought, but that was never my point and I don't know why you both are looking at it through that lens. The current MMR does not work. It directly spoils the experience for a large portion of the playerbase, and unfortunately some of those people being adversely affected are the exact players we NEED. The new ones, who haven't even gotten a chance to learn the game.
3 -
Once again, I disagree with him.
9 -
I agree with the video, MMR is needed.
The devs should however improve how the matchmaking works, because it sometimes happens that I face completely new baby survivors.
3 -
I totally agree with you. If you add to this that many times i face survivors that are way better than me, you complete the puzzle. The last 2 days i face with my 10 lvl killer and brown perks, survivors with full purple perks who knows the game way better.
The game isn't beginner friendly. You get a match with 3 kills and then 10 matches with 1 hook at your best. Same person same killer.
You face survivors that they just mokking you as you described. 1 or 2 take the chase and the other 2 makes the gens - end of the game.
I rarely face survivors who pair my level and my abilities. I know i'm new to the game (30 hours) but that doesn't mean i don't want to learn it.
The resume is we all want to have fun. I want to enjoy the game. And i don't enjoy the game if i face opponents with 1000 hours and full unlocked perks. There is no balance in this game.
i don't have problem to wait 10-20 minutes for a nice balance game with opponents similar to me.
I have problem if i go in on 2 minutes and i hate my life playing this game for the next 15 minutes. This really sucks.
1 -
But this is one of the fundamental problems with the current MMR and the meta that follows on from it: it's entirely down to the end results. Now, you could say it's simpler to model it as a series of 1v1 competitions with simplistic win/loss conditions, but if you're not going to show it to players then you don't really need the simplistic ones.
And this fosters a game environment that encourages minimal game playing. Hooks don't matter. Surviving chases doesn't matter. The ideal for a Killer is 4 hooks = 4 kills, and the ideal for a Survivor is to sit on a gen and never, ever see the Killer, rest of your team be damned so long as you get out. You could maybe substitute in one of the major factors feeding into the results in place of the actual result and have MMR encourage healthier dynamics.
1 -
It's one of those things that's going to be anecdotal, but I've definitely seen a difference.
Another good piece of evidence is Otzdarva's 'streak' videos. Before SBMM, he could do 100 win streaks without addons. Now he really struggles to do 20.
1 -
Mmr is bad if you are above average and know the fundamentals of the game but are just simply bad at chases and die a lot you get stuck in games with teammates that have 20 hours on dbd
3 -
Just FYI the devs actually talked about this exact question in the dev chat today and their analogy was if you talk about hockey, for instance, you talk mainly about how many goals the teams score because that's the bottom line. You wouldn't say a hockey team "won the game" if they had more shots on goal than the other team but their shots missed. Similarly in DbD things like chase time and hooks help you get to the bottom line goal of escaping or killing, but they're not goals in themselves. Sure someone who has a lot of chase time probably did ok, but if they didn't escape did they actually win? Not to mention there's already a correlation between being good at chases and being good at gens and escaping, so counting escapes works "as a proxy" (as the dev put it) for those other factors.
1 -
Imagine that we are in charge of rating soccer teams. Do you determine each teams ranking by the number of goals scored or their win to loss ratio? You could certainly argue that a team that won in overtime with a single goal might not be as skilled as another team that scored 4 goals but their opponent scored 6 goals. But I think ultimately we have to judge success based off of the real win condition. A killers win condition is to kill survivors and a survivors win condition is whether or not they survive. And like they said in the stream, if you create a scenerio in which both sides can win then you end up with MMR bloat.
3 -
I mean I am fine with MMR placing me with equally skilled opponents. My problem is that there are 20+ killers in the game but only like 4-5 are even playable against good survivors. And it's not like League where you can just flip characters no problem. You had to put effort into grinding these characters, and the devs just refuse to make them viable at high MMR.
And again, it's not like League, where you could probably argue that Grandmasters can't use any character they want at top play. The skill requirements for DBD are sooooooo low any above average gamer is going to be playing almost optimally (at least as survivor). It takes almost 0 skill to run safe loops, and it is actually 0 skill for everything else as survivor. Literally, nothing survivor does takes any skill outside looping and on some maps there's so many god pallets you don't even have to be that good to do well sometimes! So everyone who plays the game long enough to learn the maps can play like 90% optimally even if they are mechanically dogwater.
0 -
I think killer MMR would work if there were more killers. Now survivor MMR (cough soloq) is so broken that MMR doesnt even do anything, it's a catch-22 situation.
1 -
Of course the Canadians talked about hockey... 😂
That aside, the problem is quite simple: they set the win conditions. You don't get to make analogies to professional sports when you're the ones deciding what a win is. You also don't get to make analogies to professional sports and ignore that most team-based sports, when you have identical scores, will fall back on some other way to sort teams rather than considering just a win/loss metric.
And the problem is that their notion that kills/escapes work as proxies is false. If they want them to be proxies, then it shouldn't be the case that the most efficient way for either side to get their actual goal is to minimise the number of hooks/chases they get into. Either the entire game needs to be heavily restructured and hope that the correct behaviour is emergent, or at least for now you need some sort of two-factor step in MMR adjustments.
That, and if you want to make analogies to games with total scores, start making MMR based on the overall match results and treat it as a 1v4, rather than making it into "we will give you a win for every goal you score as an individual".
1 -
My opinion.
The SBMMR is better than the old rank system, of course it need a lot of changes but for an average players like me it's way better.
As I've seen most of people who complain are high level players, specially killers. They usually complain about facing sweatfest SWFs with purple medkits, meta perks, etc. If you don't want to face such teams just don't equip full meta gen regression perks with strong addons and you'll lose some matches and eventually you will face worse survivors and will get more chil matches. If you keep using your top build you'll face good and well equiped survivors all the time and it's a vicious circle. With the old rank system you couldn't do that, because losing ranks were very slow and you were forced to stay afk a lot of matches in a row for deranking. For example, when I hit purple ranks I usually went down to green ranks where I can play more chil without steamroll most of the teams I faced.
Also, a lot of the complains of popular streamers is because they were too used to their fancy "win streaks" by exploiting the old rank system which made them face bad survivors who play very bad but play a lot and hit the red ranks...
2 -
Actually it sounds like you're the one trying to redefine the win conditions, not the devs. I get that in your opinion you want the players who do the best in chases and get the most hooks to be considered the winners, but they're not. It's the people who escape or gets the kills. If the devs wanted chases and hooks and such to determine the winner then everybody would just be going by the endgame scoring screen since that's exactly how Bloodpoint scores at the end of the game are calculated. But you probably won't find very many people nowadays who consider the person with the highest Bloodpoint score to be the winner of the match.
1 -
Wait, what?? So when every time I finishing been tunneled or facecamped u think it a right game experience??? Lol, no, MMR is reeeeeealy bad. Hell, no.
1 -
My games got way more inconsistent after sbmm, I got way more one sided matches one way or the other and after a month I couldn't take it anymore and haven't played since September and from what I've seen it's only gotten worse, at least the pip system tried to quantify the various things that happen in the game instead of only caring about the result.
The bigger problem was that sbmm didn't even use itself to matchmake, hence the abundance of rainbow lobbies when it was a thing. Full red rank games or full purple games or green games always felt a lot more balanced then the vast majority of my games with sbmm
0 -
Thats a much better analogy the the Hockey shots on target one. If you got 8 hooks and zero kills, you lost and you need to reconsider your strategy.
0 -
The only thing I want to point out about the video is Thale argument that we don't *see* our MMR. I see A LOT of people in the forums saying "as a high MMR player" just because they have... Difficult matches? But assuming the match making works, that means they're match with and against similar skill level players, which balances it and doesn't hand out a free victory around that level of play. People having meta perks unlocked doesn't mean they are "high mmr" on either side.
And if we assume that MMR doesn't work then the whole "difficult matches" argument isn't valid either, because those players will be matched with and against random players, giving them imbalanced and extremely random results.
I just wish forum killers (sorry, I meant, users) would stop using the high MMR argument as a poor excuse for their loss, because they DON'T KNOW their actual value, nor the one of the players they were matched with
1 -
I have over 1000 hours in the game and im still #########, my friend has about 200 and is far better. Play time != skill
2 -
So remind me again, what happens if you dont have any system in place, at all? Everyone gets mixed in with everyone else, even 3k+ hours and baby players, right? But are there that many high end players? I know that the game is 6ys old by now and has a big playerbase (some would say it had), but are there that many thousands of hour players in relation to mid- and low-level players?
0
