Visit the Kill Switch Master List for more information on this and other current known issues: https://forums.bhvr.com/dead-by-daylight/kb/articles/299-kill-switch-master-list
"stats mean nothing" is a lie
Comments
-
I think we would like to see a straight horizontal line, meaning at each MMR value there is a 50% escape/kill rate. I suspect, the line is not straight nor horizontal over its whole range.
If it is a straight line the I think the forums would have to concede that MMR appears to be working pretty well.
Additionally, whatever shape the curve has...it would be awesome to be able to use it as a baseline to compare future changes too.
1 -
MMR isn't supposed to bring a high skill player to 50% in all matches. It's supposed to bring them to 50% against opponents of equal MMR. But high MMR players also go against a lot of lower rated players and in those matches MMR goes down a lot for the higher rated player if the underdog wins and up only a little if the underdog loses. Ideally the ratio ends up being that the handful of times you lose against weaker players balance with the more frequent times you beat weaker players.
So it's very possible for example for high rated players to have a very high win rate and still have a stable MMR because they spend a lot of time in matches against weaker opponents and it takes more wins to gain MMR than it does losses to lose it.
0 -
Personally haven't said they didn't matter, mostly just pointing out they don't mean as much as op thinks. Also Abit miffed they blatantly disregard losing games as simple bad, but that's on me.
As for your point your correct that stats don't factor in mercy(thought that was implied by various factors) but by that measure it also doesn't take in tunneling or camping. We simply don't know and therefore anything we say is speculation.
I'm gonna assume the nurse thing is for others cause I don't remember taking about her
0 -
The thing you describe should apply at both ends of the spectrum, some high MMR survivors should get matched against weaker killers, and some low MMR killers will get matched against average survivors. Both phenomena should lower the winrate and compensate to bring the total killrate to 50%. On top of it, there is the hatch problem, survivors have a way to escape without gaining MMR, so the killrate should not even be 50%, it should be less.
1 -
that is one aspect that i hate about dev's statistics is that they use kill-rates as only statistic to balance and they create perception that kill-rate = win-rate of the killer. 52% kill-rate means that the killer can get 2 kills(draw) in 9/10 of their games and they only need a single 3k to be at 52.5% kill-rate (21/40 survivors died in 10 games). 60% kill-rate means that the killer only needs to kill 24 out of 40 survivors to get 60% kill-rate. a combination of 6 games that are 2k and 4 games that are 3k is one combination which is 24 kills out of 40 survivors(60% kill-rate). Another combination is that close to 60% is 6 games that are 2k's, 1 game that is 1k and only 3 games that were 4k's and only 3 games that are 4k's. it leads to 25/40 kill-rate which is 62.5% in 10 games. Obviously these statistics can be upscale to 100's, 1000's of games.
As long dev balance exclusively only kill-rate only kill-rate itself, the killer will never feel like they're winning more then they are losing because they can secure a bunch of draws and only win a few games decisively to maintain average kill-rate.
I think otz said it best in his video. According to dev's eyes, Pig, Wraith, Twins, Myer's, Hag, Billy, Freddy, Legion and Artist are considered the most balanced killer at top 5% MMR because all of these killer hit 60% kill-rate on the dot. it means these killers need zero changes because they are perfect balance. it takes no consideration on killer gameplay.
the pick-rate shows how confident the killers are in picking up the killer and winning with that killer. the higher mmr for survivor are flooded with nurses and blights. a high kill-rate could be that the killer wins a lot, but it could be that the killer draws a lot(rarely get 0-1k). killer that get close to getting a lot of 2k's have opportunity to get 3k & 4k's since difference between a 2k and 3k can often be a single hook. Once a killer gets a 2k in a match, he is ever more likely to almost near guaranteed get 3k. depends on door rng and the map, a 2k can very quickly turn into 3-4k. killers rarely lose completely. they are often able to draw in worst case scenerios in many games but if they win, they're likely to win big which in dev kill-rate balance is overperforming.
1 -
It does apply for survivors too, and you should expect high MMR survivors to have higher than normal escape rates compared to average MMR survivors for exactly the same reason. It's very possible for high MMR survivors to have an above 50% escape rate and high MMR killers to also have an above 50% kill rate, they're not mutually exclusive since they don't only play against each other.
Also FYI MMR doesn't count hatch escapes, and it adjusts MMR by smaller amounts for the second and third kill than the first since after one survivor is already killed it raises the chances of the others also being killed.
1 -
The issue is that the killrate of 60% is averaged across ALL skills, it's not just the top 5% who have >50% winrate. Thus, we are speaking of a massive part of the killer playerbase having a MMR too high for the matchmaking to be balanced, likely more than 50% of killer players. This to me is equivalent to say that MMR doesn't work (maybe it works mathematically, but the whole purpose is flawed, basically it means that as a killer if you're just in the top 50% the MMR already thinks you're a god that cannot be matched with average survivors).
Re: hatch. Say that ~10% of survivors manage to escape through hatch in a normal game. Then as a killer you would need a 45% killrate to stay at steady MMR (45% wins, 45% losses. 10% draws). So the natural killrate of the game should be 45%, that is what MMR tries to achieve. Since the observed killrate is 60%, it means the actual deviation towards the normal result is 15% not just 10%, making the stat even more abnormal.
This flawed MMR distribution implies that MMR must make some compromise to still give you matches otherwise you'll never be able to play, likely this trade-off is your current queue time. Among all the problems this implies, one is that if you want easy matches you can simply dodge any match where you have too fast queue times, because it means the MMR has been able to find opponents that will give you a challenge, in contrary if you wait 10 minutes for a match it will likely be a stomp as MMR has given up by that time.
And actually, I think this is exactly what happens, in particular I had this 10 min queue for killer a while ago, then this match happened:
0 -
Again, you're assuming that MMR goes up or down the same amount every match but that's not the case. It's quite possible, for example, that the average killer MMR is, say, 1100 and the average survivor MMR is 900 and because of that when survivors die they lose less points than they gain when they escape, balancing out them escaping a bit less often than they die. There's no particular reason to think that killers and survivors have the save exact average MMR value.
0 -
Statistics is a field that people spend years in college for, and get degrees for. I'm not going to sit here and pretend random DBD player on the forums is a master statistician who understands the nuances of statistics, and the application of it. No, in most cases here, it's just someone who wants to insist "THIS NUMBER BIGGER THAN THIS NUMBER! SEE! I AM RIGHT!"
These statistics aren't useful for the average person. If they want to give statistics, then they should explain why the stats exist. Just tossing numbers and going "COME TO YOUR OWN CONCLUSIONS!" is honestly just destructive, and leads to absolutely nothing but arguments.
7 -
Did you read my post? That's exactly what I'm saying.
0 -
It is essentially a separate issue to have statistics and to consider why the numbers in those statistics are what they are.
The reasons for the lowest kill rate for nurses have already been discussed. It is not because she is weak; it is a learning curve issue. There is never a need to buff her.
Wesker has the highest kill rate. He is certainly not weaker than the others, but needless to say it is because the survivor has not learned how to counter the latest killer.
It is important to look at the numbers only and how decisions should be made.
If a statistic shows that 100% of criminals were found to eat bread, that doesn't mean that bread makes people criminals, and it's not a reason to ban bread.
1 -
I did read your post, I'm saying that just because killers on average have a 60% kill rate doesn't mean the system isn't finding matches or broken.
0 -
I like Otz but the community manager in this thread confirmed that they don't balance solely on kill rates and they've had previous posts that have indicated they are aware that once someone dies the chance of more dying is higher. And they've also said they aren't balancing for 2k/2e every match as they don't even really consider that possible.
Literally here and in a previous post. Kill rates are likely only shared because they're one of the easiest things to digest and give to the community who is asking for statistics.
Just as an example, they're likely aware that Sadako is on the weaker end when it comes to killers. I don't really think you're going to be able to do a 50 win streak with no addons like Otz always likes to do. But the players who do use her get success at the level they play at and she gets a decent enough kill rate that she's got a place in the game. She's also a very easy killer to use. Probably the best starter killer in the game at this moment because she both has strong perks, a simple power, and map mobility.
1 -
i think its opposite. i think dev only balance off statistics and ignore killer gameplay. for example if you ask otz how sadako feels to play. He suggests that sadako is D-tier. among worst killers in the game but stats do not correlate to her being worst in the game. its just how she feels to play. its the same with freddy. otz has freddy at D-tier as well but freddy does better then like 50% of the killers.
deathslinger got some negative changes and yet his kill-rate is the same as huntress but his pick-rate went down significantly to 2.16% as opposed to what it was in the past which was something like 5-6%. kill-rates means so little in comparison to how the killer feels to play.
0 -
Please invite me into the balancing team. I have so many good ideas for both sides and bet that we can work a few things out that well get und communityexcited. Experience: 20 odd years as dungeon master for various system's :)
No snarky or biting comment's, all communication will be kept professionaland civil. Promise.
0 -
Otzdarva commented on the numbers and had some very good thoughts.
I am not fan boying, but what sounded pretty reasonable was this: the game is very volatile, especially for the survivors. A killer can potentially make a comeback from a very bad start or development in the trial, whereas survivors can sometimes topple the whole game with one blunder.
Also, a survivor who doesn't like the killer/map/whatever can always suicide on the very first hook and practically doom their team from minute one, while a killer can often secure a single kill even in a bad game. Both of this aspects contribute towards skewering the kill rates in the killers favor without the games being "balanced". I guess in order to achieve what we would call balanced, ie having a good shot at surviving a given trial/killing a good amount of survivors, the overall killrate needs to be way above 50%, ie probably around 60%.
Also, the killers that surprise a lot of players with their high ranking, ie Pinhead, Sadako and even Pig and Freddy, all got innate slowdown and/or very strong and useful base perks. Remember that new players in low MMR won't be having a lot of perks to chose from, so killers with strong base perks (Deadlock, Hex:Plaything, Merciless Storm, Call of Brine) will fare better then killer's with mediocre base perks, just by making their loadout more meta.
2 -
while a killer can often secure a single kill even in a bad game. Both of this aspects contribute towards skewering the kill rates in the killers favor without the games being "balanced".
I agree with you about this consideration.
The kill rates are 0k, 1k, 2k, 3k, 4k because 0%, 25%, 50%, 75%, 100%.
If the killer has not been able to execute anyone at the end of the game, he will try to execute at least one survivor somehow. It actually works, because it is not worse to be at 25% than it is at 0%.
I agree with your point about this 25% potentially affecting the statistics.
0 -
It reminds me of Cammy in Street Fighter 5. Worst win rate the entire lifespan of the game, but was undoubtedly a top 3 A+ character. She's just popular and easy to learn so noobs who sucked dragged down her win rate. Didn't mean she wasn't dominating tournaments.
2 -
He already addressed that. They help and are an overall indicator of general game health to a certain extent. The point he's trying to address is that while they help, they also aren't everything. And you also fail to acknowledge that roughly 60% is their target for kill rates, not 50%. The reasoning for this is that the vast majority of the time, killers usually get at least one kill. The ability to secure a kill during EGC serves to inflate kill rates a little, so if the overall kill rate is 50% then in evenly matched games killers still lose more than they win, since 0Ks are inherently less common than 4Ks.
0 -
Can the devs explain why the game has not shifted to the more obvious objective of hooks for the killer? This would ensure more interactive game play for survivor and killer. Especially if there was a marginal depreciation for 1 or 2 hooking someone because you didn't get all 3 hooks. @Peanits this would give killers incentive to leave hooks and go for more hooks.
0 -
Most people don't appreciate just how massive these data sets are. They talk about anecdotal stuff and their experience of the game to try and argue against stats failing to realize how astronomical the differences in representativeness are. On Steam alone, some 30 million hours of DbD are being clocked in a month. Even if we subtract an entire third of that for idle and queue time and whatnot, that still leaves 20 million hours, which at an average match duration of around 10 minutes come out to 120.000.000 / 5 players = 24 million matches. And again, this is ignoring all other platforms, which are bound to at least double that number. To further put this point into perspective: Even if you are veteran of the game, your entire playtime spanning years and years is matched on a global scale within a mere handful of minutes. And the stats capture results from all of it.
Stats absolutely mean something, to dismiss them is ludicrous, and there's a reason BHVR looks at and shares them and has for years. People just often confuse "base game balance" with "live game balance" - these stats do not mean for instance that Wesker is the strongest killer in the game, just that he is currently killing the most survivors in the actual matches people are playing. The difference between that is easy to understand: base game balance would be looking at a hypothetical controlled environment where every player is equally experienced, skilled and trying equally hard to win, and then also accounting for maps and loadouts on top of that; live game balance is looking at the actual playing environment of the game in its public matchmaking mode, where there are great and ever-changing disparities in skill, experience, "sweatiness" and maps/loadouts - no two pub matches are ever the same even among billions of trials. And there are a sheer infinite number of possible conditions and things that can happen in a match, which cannot and do not have to be accounted for for live balance because the live game itself does not account for them: matchmaking does not look at maps or loadouts, and given the low MMR cap, it doesn't look at performance much either. Whatever can happen and does happen comes out in the wash of these large numbers of matches, and the resulting stats show trends with massive reliability.
Again, that doesn't mean they show us base game balance facts, they show us the reality of how the game balance unfolds in its live environment. If you were to for example change the matchmaking, such as by removing the MMR cap, the stats would change, despite the base game obviously not having changed. But they are designing and developing for the live environment, because the latter is what actually happens and what represents the real playing experiences real players are really having. Strong add-ons and items, Moris, map X and Y, SWFs with voice comms, survivors that hide all match or kill themselves on hook, killers that let survivors go, endgame suicide parties, and on and on - all of that is included in these stats, and that does not make the stats mean anything less about live balance, on the contrary, since all of these things actually exist and actually happen and actually affect the playing experience of actual players, they have to be included among all the other things, otherwise the stats would not tell us about live balance. Sure more specific stats to get more insights and trends for the realities of more specific subsets of that live game reality would be neat to have and can help inform game design decisions much more precisely, but the global kill rate is sufficient to generally get an idea for the live balance state of the game, i. e. how people are doing in their actual matches on average. And to that end, players playing killer are performing well, including in top MMR brackets.
I don't think 60% kill rate is too much, it may well be the sweet spot, but I do think solo survivor is being left behind too much (one example where specific stats would be neat), and that the MMR cap is allowing for top players to win decidedly too much (particular problems being certain killers/add-ons/camping/tunnelling/SWF, but changing the MMR cap would already go a long way to alleviate these problems).
2
