SBMM DOESN'T WORK. KILLER STATS FROM 100 MATCHES TO PROVE.
Average kill ratio of 2 they said? How about 3?
Average wins same as losses they said? How about 3.65?
And I'm just a mediocre killer with most, other than Cannibal, killers. Individual mmr per killer they said? How about 10 wins streak on Cannibal?
But... Don't take my word for it. Otz had video on the same topic a few days ago. His stats are even more damning. Clear proof that MMR is a failure.
Comments
-
...so even with the MMR that killers are complaining about, killers are STILL winning 75% of their matches? I thought this game was survivor sided lol
34 -
My stats for 200 Survivor games show around a 1.5 kill rate.
It's inconsistent.
39 -
It's not consistent. I've seen plenty of survivors with and escape rate of like 80% but I guess since they kind of even out to 2.0 but the individual matches generally suck for one side.
8 -
I wish the people who recorded these stats kept video logs so we could view the quality of their opponents.
13 -
I can't record that many :(
2 -
People love whining in this community, nothing new. To be fair, BHVR's decisions rarely help.
4 -
I think we all know the MMR criteria are awful and lead to horribly inconsistent matches
1 -
I've been keeping survivor stats for a little while now, with approaching 200 matches on record now (the huge majority of these matches have been either viewable on Twitch or YT over time). These have been a mixture of 2-person SWF groups, 4-person SWF groups, and solo matches.
My personal escape rate (excluding one night where my SWF team threw during the Winter Event doing goofy Snowman things, and I spent a whole night running a goofy build to make Power Struggle work) has been slightly over 60%, with the "team" as a whole just slightly under that at 59%. And while I'm a decidedly decent survivor, I'm nowhere near great at that part of the game.
What does that mean? Very little -- it's small sample size, just as the OP's 100 matches (or even Otz's matches are -- comparing our results with his would be like comparing our basketball skills with an in-his-prime Michael Jordan -- he's arguably in the top 1% of players, so what happens to him will never be typical of the majority of the playerbase). The OP's numbers are an interesting snapshot, but much like my numbers would be, it's such a tiny representation of what actually happens in literally tens of thousands of matches a day, it's not statistically relevant -- simply one person's brief experiences.
For the record, I DO agree that SBMM isn't working properly, as I firmly believe that it prioritizes queue times over everything else, particularly during peak hours, which is exactly what it shouldn't be doing. The devs have mentioned that they intend to tweak SBMM in the future -- it does desperately need changes (I think it's slightly better than the busted matchmaking we had before, but it's FAR from perfect at this point and needs work).
4 -
While 100 games is a decent number for any person to play, its really not that many for determining MMR when you're only playing a killer for 4-5 games, on average.
Even if there is a weighted MMR system, where your General MMR will passively effect your Killer-Specific MMR, constantly playing new killers is going to slow the rate at which you rank up.
Either way, I think we can all agree that the MMR system is too unsophisticated to be any real indication of a person's skill and ability to win. That, or their back-end system cares more about correctly predicting the outcome of a trial than providing an evenly matched one.
2 -
Definitely. I find that the range of quality depends on the time I queue. Late night generally feels a bit more harder than my early games. Skill still varies between but again, late night seems to be so wild.
1 -
So you don’t ever play killer and you have good stats on your little sample size when you tried it, woah. Interesting conclusions.
Also talking about Otz who plays the game as a job, 10 hours a day, could draw each map with each totem + palet + gen location by memory as as if he is the average killer lmao. Sure.
8 -
None of my teammates tonight have had over 200 hours, i am suffering
4 -
I genuinely question whether MMR is actually on for my killer games most nights. My experience reflects OP's. I haven't done an Otz experiment. 4k games at 4 or 5 remaining gens happen with alarming frequency. I don't know if my brackets are bugged or what. I'd genuinely be interested in submitting my steamid64 to a dev for study because it doesn't add up. I would think that, mathematically speaking, I should get good opponents at a certain point and not have double digit win streaks on multiple killers.
3 -
Otz said "if you are experienced, bring the best perks and addons, you are winning most of your killer matches". His killrate was 3.85. The gates were never opened. As perkless he still had a 3.15 killrate.
The game definitely is not survivor sided.
7 -
I feel like it's determined by time of day for some reason. During the day I get really strong survivors that play as a team covering for each other and bang out gens quickly while at night I always get wannabe bully squads that arn't very good at bullying anyone and just potato survivors in general and it's the same nearly everyday it's strange. I'm EST if that matters.
1 -
No one would want to watch 100 matches. Besides, MMR is only evaluating deaths and escapes, so any other qualities are irrelevant.
0 -
Stop torturing yourself. The only way to enjoy survivor is SWF. Join DBD Discord and you will have no problem finding good teammates.
0 -
I started keeping a lot of different (many never asked for) stats with tricksters release I was about to start publishing results after the 5th anniversary but then I felt the resident evil update mess then MMR (working or otherwise) skewed results over a longer period of time.
So I decided to wait until new year & I began the project again on new years day.
The intention is to play each survivor in rotation using only their adept perks 100 times each in the space of 1 calendar year (actually hopefully before December 1st i'd like to have it wrapped up for xmas) assuming only 4 new survivors release this year that takes the list to 33 survivors & 3,300 individual games.
I actually intend to do 3,500 games because i'll then do 100 games using my own build & 100 with the generally considered meta just to see how they play out.
If all goes well i'll do something similar with killers in 2023.
Every single game is recorded & will be uploaded & linked to along with the stats to verify their accuracy & authenticity.
Round 1 recordings are wrapped up I just need a day & a few spare hours to sit back & watch the videos & accurately note down all the different stats I'm collecting.
I have my own expectations on certain results but I guess we'll see eventually.
For the record I believe I'm a pretty good survivor in my normal build I escape well over 50% of my games, I'm not a god looper probably not even a great one but I try not to put myself in positions where I have to be while also progressing the game & looking after the team I certainly don't hide or do nothing!
While even my own stats can't be considered much more than a sample as there are far too many variables such as skill level on both sides / team mates (all my games are with at least me 100% solo) & even region played in can be totally different games.
But I still think that many games in a relatively short space of time is a fairly solid sample & even if escape rates are still argued about many of the other stats will show fairly significant trends.
Keep on collecting I'm a stats nerd & would love to compare others to my own to look for more growing trends.
1 -
LOL what? Those are my killer matches and I'm 90% killer. 100 sample size is considered to be statistically significant: https://tools4dev.org/resources/how-to-choose-a-sample-size/
0 -
That's not quite correct about MMR only looking at escape and deaths. It's true that whether a survivor dies or escapes determines if a killer's MMR goes up or down respectively and vice versa for the survivor. But how much the MMR goes up or down supposedly depends on other factors, including the relative ratings of the players, the order the player died (i.e. the first death has more weight than later deaths since once one survivor dies it makes it easier to kill a second survivor, etc), and I think the length of time it took to kill a player. The MMR doesn't just go up or down a set amount every match, it varies depending on the scenario.
1 -
Yes, all true. But what does it have to do with survivor qualities? I'm not arguing that it's not important, I just don't understand why it matters. Please explain.
0 -
The ability for people to survive, or for killers to secure kill, is entirely dependent on having redeemable qualities that lead to these outcomes.
How the hell did you respond to my comment while completely missing the point? I don't need to watch all 100 games, but I could certainly watch 5-10 and draw some conclusions about both the quality of the data gatherer, their opponents, and the efficacy of MMR.
Yikes.
0 -
I don't really understand what you are saying. I've had the game for 2 years and MMR system was collecting data for month before it was turned on. I don't "rank up". I stay at the same rank all the time. Also, it's not even about me. Otz has been playing it like a job and you can't say that MMR doesn't have anough data on him. He steamrolls every match.
0 -
I was just pointing out that the statement "MMR is only evaluating deaths and escapes, so any other qualities are irrelevant" isn't accurate because it does consider other qualities of the match besides just whether or not someone died to determine how much MMR goes up and down (i.e. relative ratings, duration of the match, the order they died). It's a common misconception on the forum that all that matters is if someone dies or escapes, but the context of when they died in the match and the relative ratings also matters.
0 -
Out of curiosity, are you including the stats for the games with DCs? Because a single DC typically leads to a full survivor loss and is not indicative of killer performance.
1 -
solo q the killer wins
SWF if the premade is decent the premade wins if not Killer wins again. the second option is less likely one so thats why most killers still win most of their games.
1 -
I apologize if I missed the point, but tbh, the way you speak, it's easy to miss. I have no idea what you mean by "dependent on having redeemable qualities". What are redeemable qualities and why do they matter?
0 -
Well, now we understand where your issue is. You have a misinformed view of the game and streamers. Otz does not steamroll every match. He has multiple matches where survivors walk all over him.
1 -
He does steamroll every match when he plays to the maximum of his ability with sweaty builds. Watch his video. He had 3.85 kills average and not a single gate escape. If that is not steamroll, I don't know what its. Just to be clear: I wasn't talking about his regular streams, where he is being nice and tries not to tunnel and doesn't stack gen regression perks.
0 -
You gonna watch all 200 matches even if they had?
5 -
This game is indisputably survivor sided. You put the best against the best and unless you play nurse or blight you will lose 90% of your matches.
0 -
If you read the thread you'd see my answer to this.
No, he does not. You are commentating on a very small sample size of games he recorded and analyzed. In several matches he got very lucky with a bunch of suiciding potatoes, and in a larger sample size those become the minority, not the norm. And does that mean you admit that one of the best killers in the game has to play incredibly sweaty or they're at risk of being steamrolled? I'm not sure why you're trying so hard to portray the realities of this game inaccurately. The point of having video data for a LARGE sample size is to be able to do qualitative and quantitative scrutiny, this isn't hard.
1 -
I have meticulously taken notes of my past nearly 100 games. Later this tlday I will have my 100th game and post my stats.
I am doing pretty well as killer, so I guess that my MMR is at least medium, maybe higher, but without seeing out own MMR this whole discussion is pretty tough to argue any way.
My Kill rate is 2.72 - 2.73 and it's a pretty hard earned number. I would say that I am pretty good at this game, so if you are consistently good, you can eek out a higher kill rate then 2K/2E,but it ain't easy. I think that's ok though, it would be pretty abysmal if you got stuck with 2K/2E no matter what you did and how you played, but some of this matches were fought tooth and nails.
My sentiment about the SBMM still stands: as a good killer player you can consistently show a good performance, but the quality of the overall matches detoriates considerably. It's ok to have a super sweaty game every now and then, but if each match feels like you played the finals of the dbd Olympics, game after sweaty game, day after sweaty day, full of eating pallet after pallet, eating DH after DH, it can burn you out. And I don't think that the answer to that should be a high horse snarky response "that's what you get for playing that good. Reap what you sow. Get destroyed by your own success. Get punished for playing consistently good."
1 -
"You are commentating on a very small sample size of games he recorded and analyzed." - well, yes! As I said before, he is not tryharding in his streams, so there is nothing else to look at other than this specific video.
"In several matches he got very lucky with a bunch of suiciding potatoes" - that's the whole point of MMR: eliminate matches where you get "lucky" with potatoes. And that's why I'm saying it's not working.
"And does that mean you admit that one of the best killers in the game has to play incredibly sweaty or they're at risk of being steamrolled?" You are missing the point. I'm not discussing playstyles. I'm discussing MMR, that we were told supposed to pair sweaty killers with sweaty survivors.
"I'm not sure why you're trying so hard to portray the realities of this game inaccurately." - lol, lean off your conspiracy theory. I'm not "trying to portray" anything. I collect data and I present my interpretation of it. There is nothing more to it. Just to be clear: the point of discussion is not survivor vs. killer. The question I'm trying to answer is "does MMR work"
"The point of having video data for a LARGE sample size is to be able to do qualitative and quantitative scrutiny, this isn't hard." - not sure what you are implying here. 100 is statistically significant. https://tools4dev.org/resources/how-to-choose-a-sample-size/ You keep being evasive in your answers.
"qualitative and quantitative scrutiny" - LOL, you keep doing it :D Working for a marketing department? Let me try that too:
We should create a distributed analysis by proactively enabling synergistic statistical approach leading to big picture revealing mathematical realities of SBMM, this isn't hard.
You like me talking your language?
0 -
I'll take that as a no
0 -
He said that he can do his "qualitative and quantitative scrutiny" based on 5 matches. I'm literally trying my hardest not to laugh aloud so I don't wake up my family. Pure comedy 🤣
0 -
Yeah thats about what I'd expect
0 -
There was some guy who recorded his win streak without any add on or perks on a Killer and posted in the forum. As you can guess people just pinpointed the low quality survivors out of his entire matches and discredited his entire streak and meme'd his topic title for a bit.
0 -
Bearing in mind each individual killer has its own MMR, and with some killers you played games in the single figures, and then totaled all killer results together to give the impression the killer MMR is a singular result, this test has some rather large holes in it.
MMR does need changes, granted (still better than what it was before) but to make this worthwhile it's probably best to play 100 games as the same killer and tally that result, which will create a more accurate subjective set of results.
Unfortunately, as a set of results made to prove that MMR doesn't work, the method isn't the most effective.
0 -
The majority of the playerbase is casual/beginner,so at a lower level yes,kilers get more kills,but we shouldn't balance the game around begginer players,it doesn't prove anything
0 -
Depends what you classify as "winning" for killers.
I look at it this way, 0-1 kills = lose, 2 kills = draw, 3-4 kills = win.
1 -
Play 100 games on one killer. Someone in middle, like Nemesis, or Demogorgon.
That's only way to get some results with how MMR works. Don't keep changing killers.
3 -
I would watch random samples. Just to learn how he played it and how survivors played it.
0 -
System has data, but system doesn't have players equal to him all the time...
It used to work like that and Dowsey had to wait few hours to get tournament squad.
0 -
If you live stream them the vods will stay up. Twitch and YouTube are great for that kind of thing.
0 -
And what if those random samples were not representative of the rest? How many would you watch?
0 -
If it had description. I would watch at least two for each killer he played, one with 4k and one with escapes. Then probably more with my mains, because well, I understand those more...
0 -
If you're using kill rates as a metric to evaluate SBMM being balanced or not, the devs latest stats showed a 53% kill rate. Pretty sure their sample is way bigger than your 4 Trapper games.
0 -
well, if you "use" Otz...
why not also watch tru3ta1ent?
1 -
Wow, that's really weird. I recorded a lot of games (over 100 or so) and am consistently seeing what the OP posted. Also I've been keeping an eye on what other players are posts and they're pretty close to what OP is posting too, for kill rate.
Wonder what makes your data so different than everyone else. Also, I thought you played Killer?
1
