Question about the accuracy of the MMR. I'm afraid this System is too inaccurate
hi,
i don't know exactly how this system calculates the MMR. Nobody actually knows. I imagine it's similar to chess and Elo numbers.
I would just like to know how the system is supposed to know which offers I use (e.g. a diffr maps), which items I bring into play (e.g. strong items like red AHS with medkit etc.) and which perks I use.
I would like to remind you that there are strong and weak perks. (hi Meta)
Is all this included in the MMR calculation? I certainly can't imagine that.
Then why is it called skill-based matchmaking? This name just doesn't fit. There are simply too many things that can influence a game. It's not just about time, but also about certain cards, items, perks, offerings, etc.
The name would fit if it were only about pure skill. How is the MMR system supposed to know whether I'm a camper, tunneler, good survivor or good looper? I just don't think it can do that. That's why I've come to the conclusion that it's inaccurate. Pls let me know if I'm wrong…
Does anyone know anything more about this?
Comments
-
- Using a "strong" perk shouldn't make any difference. I can use (for example) "adrenaline" or "pop" (we all agree that they are strong perks), but still not be skilled at the game.Likewise… I can use "red herring" or "hoarder" (we all agree that these perks suck) but be a very skilled player
1 -
It's just kills or escapes. That's it.
7 -
Ever played Mario Kart online with VR? It's a bit like that.
1 -
There IS a fandom wikipedia thats tells alot about the SBMMR system, but even then im personally not sure if this is all valid
0 -
It's kills and escapes, with survivor escapes being modified by the result of the other survivors (others escaping is better), and then the overall result impacted by MMR comparisons. Also, if the game lasts under 10 minutes it has a lower impact.
The system is designed so that most players will succeed 50% of the time. This means that a person using meta perks / iri items can be in the same spot as a better player intentionally running non-meta / no items.
You can keep your MMR lower by intentionally playing a weaker build. Hypothetically you could do this in chess as well by intentionally playing a weak start, its just DbD is the only game where people actually try to do it.
2 -
It's valid. The info on the wiki comes from official sources and data mining the game.
And I hope this comment is not erased again without warning or reason like my first one in this post.
0 -
3-4 kills is treated as a win by SBMM for the killer.
Escape is treated as a win by SBMM for the survivor.
You can get max 20 MMR points per match.
Here comes the big problem now, once you reach 1600 MMR (soft MMR cap) you will be matched with everyone between 1600 and 2100 MMR points, and there are already way too many players of different skill levels in that range.
6 -
I thought the soft cap started at 1800? 1600 seems low, in other games that means basically intermediate.
0 -
it used to be 1800, but due to community complaining a lot over queue times, it got lowered to 1600. Now we have lower queue times and vastly less accurate matchmaking.
1 -
Data miners have figured it out pretty well, if you still need an in depth explanation let me know, but roughly it only counts kills or escapes, but nothing else and is viewed as 4 different 1v1s, the killer against each survivor.
0 -
pretty much this, tho there are many premises that affect the final MMR points gain/loss per match. I was investigating it and actually managed to come up with my own guide on it. Once you realize how it works, it doesn't look complex at all and pretty much explains why everything feels so bad with it rn.
1 -
MMR in this game tells you pretty much nothing about skill.
Nothing metters other than how many kills you got and how many escaped. Its pointless to care about it because it doesnt count any context or important facts that play each match. Killer ablity, killer addons, Survivor items, Survivor addons, power level of perks on each side, map, map size, map RNG, how gen speed was, if killer tunneled or camped, if someone gave up or guit, etc.
Old Emblem system was MUCH better but still far from perfect.
3 -
- On a new account you get a starting value for killer and survivor MMR that is above the very lowest MMR regions.
- From there on, MMR gains are decided by kills / escapes for killers with the MMR of survivors taken into account respectively, so that beating a team with a higher MMR, gains you more MMR and escapes / deaths + suvival time (although I think the latter only applies a positive multiplyer to MMR gain, that would either increase or decrease it further) for survivors with the killer MMR taken into account.
- All survivor characters share the same MMR value. Among killers each MMR value scales in some kind with your highest or average MMR value (I reckon it's probably highest because otherwise it would scale differently depending on how many killers you own / actually play, which kind of defeats the purpose of this mechanic). I also believe that all killer characters have an MMR value in the same bracket. Meaning, you can't be low MMR on 1 killer and high / mid MMR on another.
- For survivors, the MMR gain is calulated individually and not as a team. Meaning 1 survivor could raise their MMR by escaping while the other 3 lose MMR. For the killer, 0-1 kills is considered a loss, 2 kills a draw and 3-4 kills a win.
- There are 3 MMR brackets (high, mid, low). As long as no backfilling is involved, you will only compete with players in the same bracket. As queue times rise however, it is possible for the range eligible players to increase as well. So if you were at the absolute highest end of mid MMR, you could potentially go against players that have a much lower MMR than you but not higher.
- Hatch counts as a draw for survivors.
- Given the number of DBD players in each region, we can assume that the MMR distribution looks almost continuous. Meaning, there should not be a huge difficulty spike in matches even when you enter a new MMR bracket. However, because the 1:4 ratio between killers and survivors is almost never given perfectly, it is possible that after reaching a new MMR bracket, your opponents will be a lot better than usual.
- As stated by the developers, most players are in mid MMR and only very few at the very extremes. Meaning most of us probably aren't "high MMR".
- It is very hard to derank in the sense that only a large amount of continuous losses would allow you to sink to a lower MMR bracket. Even if you were to have the absolute lowest MMR value possible in one such bracket.
The main issue with SBMM in DBD is the arbitrary idea, that kills and escapes indicate skill. On a developer live stream one former dev compared the game to hockey. Their reasoning was, that good chases and many hooks give you more of a chance to "win" like in hockey more shots on the goal give you a higher chance to actually land one.
This however does not consider the very nature of DBD. Tunneling a survivor out immediately can give you a kill at 3 hooks, the same goes for camping. So a killer with 3 hook stages (possibly only 1 hook) and 1 kill is considered better than a killer with 8 hook stages (possibly all individual hooks) and no kills. It also doesn't account for the unbalanced nature of a video game such as DBD in regards to loadouts and maps. What is more impressive; beating a good killer / survivor team on a favorable map with the best loadout or beating the same opponent on an unfavorable map with a bad loadout?
In a survivor team there are different ways to increase the chance of survival. One of them being to play careful and hide while the rest of your team tries to escape. At worst you have a chance at the hatch and at best your team mates will carry your load and you escape. Even if you did almost nothing. That means, that it is possible for survivors to get to a MMR where they are hopelessly outmatched against any killer they face, if they choose to play this way. Likewise, this impacts the killer's MMR gain.
Post edited by Xernoton on2 -
I would disagree about the win definition for killer based on kills... As technically the 1 guy you kill couls potentially give you so much mmr that the other 3 that lived were unable to give you enough minus points so you would still "win" with 1k and 3 hooks... The whole system is quite useless for our purposes and a reworked emblem system or something completely different would be nice.
0 -
That cannot happen as far as I am aware. If you gain or lose MMR is decided purely by how many survivors you killed. The actual numbers are calucalted base of the survivor MMR however.
1 -
I mean imagine if you have a 4 men swf... 3 guys have 700 mmr, one guy is a cheater bot with max mmr, let's say 3k for the sake of it. The mmr you gain or lose is based on the difference between you and the survivor. You should get a lot more from someone who is higher than you or lose less, and the other way around. If you yourself have like 1k mmr you should lose not too much from those 3 guys escaping but gain quite a lot by killing the 3k cheater bot. (It is a bot for that example because no human can reach that mmr)
If those are given why should it not be possible to gain mmr with a 1k? It would make no sense at all if it didn't work like that,and it the whole system already makes very little sense.
0 -
There is no MMR, matchmaking is determined by RNG
0 -
The MMR gain is capped at +20 per trial, and as @Xernoton explained the amount you gain or lose is affected by the MMR of the opponent, but gaining or losing it depends only in winning (in case of the killer, +3K) or losing (0-1K):
The strength of the MMR of one's opponents also influences the final amount of points Players gain or lose towards their MMR:
- If the opponent team has a higher MMR, but they lose, their losses towards their MMR are increased and the Player's gains towards their MMR are also increased.
- If the opponent team has a lower MMR, and they lose, their losses towards their MMR are reduced and the Player's gains towards their MMR are also reduced.
- If the opponent team has a higher MMR, and they win, their gains towards their MMR are reduced and the Player's losses towards their MMR are also reduced.
- If the opponent team has a lower MMR, but they win, their gains towards their MMR are increased and the Player's losses towards their MMR are also increased.
By the way, the max MMR is 2100.
Post edited by Batusalen on0 -
I see this brought up a lot. The system could try to account for things like items, perks, chase time, etc. But those things are just going to correlate with kills and escapes at the end of the day.
3 -
For perks and items I agree but chases? Not really. 8 chases >> 1 chase. But 1 kill > 0 kills. Meaning, you can get better results with a single chase than someone else with 8. For survivors the opposite is true. The less chases you take, the less chances the killer has to kill you (at least in theory).
0 -
As long as matchmaking has an emphasis on speed, I think MMR is quite literally irrelevant.
In game 1 you bring meme build against sweat squad,so you sweat too but now you're against potato squad. It's pointless.
The only thing that matters is "intent" or mindset. Dbd is fun whether you sweat or play casually, but just like dealing with a Plague, you all need to be on the same page. That's where I think the problem is.
1 -
There's no real way for the MMR to take perks, items and offerings into account because that would mean they'd need to give every perk, item and addon an arbitrary 'power number' to add to the MMR and, well, that would be way too difficult to do. Different MMRs can get different amounts of value from different perks, and SWFs can make certain weak solo perks much stronger.
2 -
I have to say that is indeed a bit weird to just ignore each individual survivors mmrs and just average it out, despite calculating it for each of them seperatly. Like if you caught the top survivor in that game but had to let go the weaker ones that should count for more in my mind, but whatever the system is already quite bad, so this does not change much.
I d really like to know the wikis sources though, given that they are obviously not official, because the last time I got information on that the data miners gave the numbers I mentioned earlier... So that's quite interesting. I guess that is one of the few times where you cannot fully rely on the wiki... That and the old distance you got by deadharding.
0 -
0 -
I know that people give that dev a hard time because of the hockey comparison, but in the context of DBD, I see what he was trying to say. First off let's not discuss the topic of whether or not someone deserves to be known as "skilled" or not, that is not the focus here. Objectively, the MMR only considers the final result of a match.
Assume two different killers are playing DBD. One is a Wraith who plays as hard as he can, giving no inches to survivors and securing hook stages by camping or getting people out early by tunneling. The other is a blight main who doesn't go out of their to camp or tunnel but gets a lot of quick downs due to Blight being a strong killer and plays without really caring how his matches go.
If we are just going off results, which is how many kills they get, on average these two players would be on a similar level in regards to MMR. The Wraith player should be given harder opponents if he is playing the best he can every match, and the Blight player shouldn't be shot up the ranks if he isn't playing seriously or with a competitive mindset. In this context, the chances of a survivor team beating them are very similar, and so they should face survivor teams that have as close to a 50/50 chance of beating them.
If the Blight decided to start playing seriously for wins every game, or the Wraith player swapped to a stronger killer, both of them would increase their MMR even higher. Which is how a good SBMM system should work, and probably the only way it could ever work in a game with so much variety as DBD.2 -
Thank you for the link that answers all my questions. I hope we can trust this source
0 -
You can't really cater a game built around casual gameplay to SBMM/MMR. This game either needs to go back to the old rank system with a few fixes, or have two separate modes: ranked and unranked. Some casual players are just good at playing, but when they are forced to sweat to win, they are being forced to play in a way they do not wish to play, thus pushing against player agency. Many killers resort to slugging, camping, and tunnelling because they are general casual players who are tired of losing and found a simple way to win matches with the least amount of effort. I bet most players are casual, thus why SBMM/MMR is not adequate.
Also, skill is not determined by wins/losses. We don't look at a hockey/football/basketball player and say they aren't skilled because their team won a lot of games. When players are traded, they don't look at their wins. They look at how many shots/goals they make, how well they play, etc. Basing skill on kills/escapes is just lazy development and needs to be removed or changed back to the original rank system which actually determines skill. Right now, skill is limited to two options, where some players are skilled at looping, others are skilled at healing, doing gens, etc. Player agency is lost because SBMM/MMR only looks at 2 variables instead of all the variables that make up skillful plays. It's just blatantly lazy and the devs should have listened to the community when everyone said it was not a good idea to implement, but they went against their CONSUMERS and did it anyways.
So I ask, why do you all still play if you think one side is better than the other, or the game is unbalanced? Do you really like losing a lot of games? If you really don't care, then why complain about SBMM/MMR or how unbalanced things are? Seriously, the devs won't do anything until they see the player count drop significantly because that's all they're going to look at is the numbers.
So yes, SBMM/MMR is wrong, in fact it shouldn't even be a thing in an asymmetrical game, based off of team wins/losses. But nothing gets done about it because the player base continues to play and keep the player count in a decent range which is all they want to see. Either quit playing or stop complaining because it's inherently useless. Even other games base skill on what players do, not whether they win or lose, such as KDR. I could play 50 matches of COD and lost more than half but because I have a good KDR (not tracked by wins/losses), my MMR would be high. This game doesn't track that, except for pips/bloodpoint ranks which is inherently useless.
1 -
It first and foremost uses kills and escapes (the stuff people overwhelmingly actually care about and play for) as metrics and that is perfectly fine and sensible. In chess or really any functioning Elo/MMR system you look at who won to determine Elo, not how they played. Keep in mind however that Elo/matchmaking rating is an ever-changing value that by definition has to be understood in the context of many matches, just because any one match may not work out all too well doesn't invalidate the entire system.
The primary issue the matchmaking system has in terms of accuracy is that there is a cap in place that is entirely too low and captures and matches among one another a wild range of players with completely different win/loss ratios and ratings. As well as a matchmaking algorithm that heavily prioritizes short queue times over accurate rating matches, even at the extreme ends. These things can and have often been tinkered with, it's just that BHVR errs much too decidedly on the side of leniency and casualness. And they may have good reasons for that, although at the very least at the more extreme rating ranges I really can't see a justification for this approach. Winstreaks of hundreds and thousands of matches just aren't an acceptable state of affairs.
0
