Visit the Kill Switch Master List for more information on these and other current known issues: https://forums.bhvr.com/dead-by-daylight/kb/articles/299-kill-switch-master-list
We encourage you to be as honest as possible in letting us know how you feel about the game. The information and answers provided are anonymous, not shared with any third-party, and will not be used for purposes other than survey analysis.
Access the survey HERE!
Survivor escape rates vs killer win rates, a bit of math to understand why the devs are right.
edit: To make this a bit more fun. I've written an app in javascript so anyone can play with the escape rate in a graphical environment.
The app is at:
— Original post —
It shouldn't be so difficult to grasp, but apparently it is.
Here are the measured escape rates offered by BHVR:
These are averages.
In essence, the mean escape rates varies between 40% and 48%
These past few days, many posts have questioned the rates, completely ignoring the past of the game.
Maybe this will help them think a bit.
A killer wins when he sacrifices more than two survivors. The draw and loss are implicitly 2 kills and less than two kills.
What does these rates mean in practice for the killer?
A 40% mean escape rate means the killer will win about 47% of the time.
A 48% mean escape rate means the killer will win about 34% of the time. Yes, in high MMR, against an SWF killers lose about two third of their matches.
You don't believe me? Simulate it!
The following python* code plays many matches and escapes survivors with the given mean escape rate.
It counts the escapes per matches and then computes the killer wins, losses and draws.
#!/usr/bin/python
# by yours truly, drsoontm
import random
import sys
DEFAULT_RATE = 40
SIMULATION_COUNT = 100000
SURVIVOR_COUNT = 4
DRAW = SURVIVOR_COUNT / 2
def dbdsimulate(rate, simulation_count, as_csv):
killer_wins = 0
killer_losses = 0
for j in range(0, simulation_count):
escape_count = 0
for i in range(0, SURVIVOR_COUNT):
escaped = random.uniform(0, 100) <= rate
if escaped:
escape_count += 1
if escape_count > DRAW:
killer_losses += 1
elif escape_count < DRAW:
killer_wins += 1
wr = (100.0 * killer_wins) / simulation_count
lr = (100.0 * killer_losses) / simulation_count
dr = 100.0 - (wr + lr)
if as_csv:
print(f"{rate};{wr:.2f};{lr:.2f};{dr:.2f}")
else:
print(f"r: {rate}% w: {wr:.2f}% l: {lr:.2f}% d: {dr:.2f}%")
def main():
rate = DEFAULT_RATE
as_csv = False
for arg in sys.argv[1:]:
if arg == '--csv':
as_csv = True
else:
try:
rate = int(arg)
if not 0 <= rate <= 100:
print(f"{arg} isn't in [0; 100] and isn't --csv")
sys.exit(1)
except ValueError:
print(f"{arg} isn't an integer and isn't --csv")
sys.exit(2)
dbdsimulate(rate, SIMULATION_COUNT, as_csv)
if __name__ == '__main__':
main()
The program can accept an escape rate in [0;100] and —csv (dash dash c s v).
By default the escape rate is set to 40
Given the csv option, one can also look at the measured distribution:
edit: red: killer win rate, blue: survivor escape rate
All in all, the "power role" (in the theme) only wins about half the time, which mean, as a team, survivors win about half the time too, on average, overall.
I'm also surprised to note that killers really tend to lose way more than survivors as a team at high MMR (against SWF4).
This should make very clear the balancing of the game is way better, overall, that what even the most reasonable among us think.
*python is the new basic. Even non-programmers can try it online (e.g. https://www.online-python.com/)
Comments
-
I must be honest here, my friend: I didn't understand anything.
23 -
😭 Thanks for your honesty.
In summary:
Basically the program simulates matches based on the escape rate.
And for each match, measures if the killer won or not based on the number of escapes.
With it I take the measures at 40% and 48% and give the associated killer win rates (47% and 34%)
Then I draw the graph for the full 0% to 100% escape rate range to show the full distribution.
If you can be more specific about points I need to explain more, I'll try to provide more information.
8 -
In human language, the program does this (in a nutshell, with lots of language abuse)
do 100000 times: (
do four times: (
draw a random number between 0 and 100
if the number is below or equal to 40, the survivor has escaped )
count the escapes and if there are less than two, the killer has won )
count the wins of the killer over these 100000 matches and from it get the killer win rate
And by default, it does it for 40 escape rate per survivor
But it can be told to do it for any other rate
5 -
what a hero
6 -
I always appreciate someone doing the math on the game. This proves the devs "right," but it doesn't mean it's a good goal to set. Even if the 60% kill rate is true, that still means that killer gets little more than a draw most of the time. And possibly, if we look at all match combinations that could lead to that average, a killer could draw like 80% of their matches, and still "win more than the lose" to get that 60%. That, to me, doesn't seem like a good outlook on the role, and doesn't encourage anyone to play it. I know survivors must escape as well sometimes, but you are a "killer," not a "chaser" or a "pallet breaker."
The kill rate and escape rate doesn't mean anything as long as the game is balanced around skilled play, but that's a separate discussion.
4 -
thank you for the chuckle. I needed it.
5 -
I'm glad to help, friend.
…I still don't understand, but that is okay, friend.
Anything that even remotely involves math is extremely difficult for me.
4 -
The big issue with these statistics is that killers often grant the hatch or even let all survs escape. Track your own statistics as survivor and you will notice that the "official" numbers are bullshit.
9 -
And survivors also give up on hook. Or fail to make easy rescues. Or go afk. A lot. (Are you counting these in your 'personal' statistics)
Yes these figures will never be truly representative of players intentions and behaviours, but there's no logical way to draw any conclusion as to which way the scales are skewed, if at all.
An approximation based on data is the best we can get.
15 -
and how often do survivors throw the game they could've won?
devs say they dont include DC/hook suicide stats, but that doesn't mean survivors cant give up / throw other ways.
7 -
Not this again…
A 40% mean escape rate means the killer will win about 47% of the time.
You are not calculating this. You are not simulating this.
You are treating every game like a pure roll of the dice, and the thing is: DBD is not like that at all. The dice aren't even, they're weighted. Badly. And they're even weighted differently from killer to killer.
This argumentation has been tried before, multiple times, and it keeps missing pretty much every factor in the game. DBD is incredibly complicated, which means there IS no simple 'kill rate = win rate' conversion. Hell, sometimes killers with higher kill rates have lower winrates, for example with killers that are better at securing at least one kill in the endgame such as Bubba.
And to quickly demonstrate that they're not directly linked: Imagine if the devs decided to axe the hatch. What would that do for the kill rate, and what would that do for the win rate?
One of them would be changed, the other would not.
These micro-simulations where you just roll dice is not a representation of DBD.
60% is decidedly killer-sided. That one dev remark (Though I think it was Peanits that said it, and aren't they a community manager?) about the 60% target marking even lampshades this by saying that, with 60%, the game is still relatively balanced. That figure is picked because it works for the game's horror aesthetic, without breaking the balance -too much-. It is NOT the target for a balanced game.
20 -
I mean, you're making a few assumptions here to force the system to give you the information you want and declaring victory.
The most glaring is the use of random.uniform since we know for certain that isn't how the game works. The distribution isn't even close to uniform. There isn't the same chance for a 0k as there is for a 4k, for example.
Even just using the nightlight stats, 4k is the most common outcome (33.83%), then 0k (19.77%), followed by 3k (17.34%), 1k (16.49%) and finally 2k (12.58%). That's heavily weighted at the top end, not uniform.
Not to mention your results are highly off. Just defining a killer win to be a 3k or 4k pts killers winning about 51% of the time, and draws occur about 12.5%. so survivors are winning about 37% of the time. Not killers, although that appears to be the result you were trying to shoot for.
If your model doesn't even come close to the known data, then your model is wildly wrong. Which makes it seem like this was done to manufacture an outcome you liked to make a certain point.
The information is maybe interesting, but it isn't relating to DbD in any way.
23 -
OP's approach is a good start, and what you say is also very true. I deal with a lot of identification of trends and understanding correlations and statistics in my job, and I fully agree that going next on hook, friendly killers, and other influences need to be accounted for to better understand what's really happening in people's games.
I'm a novice at Python, so I probably won't run through the code unless I get super curious. But these factors could be coded into a simulation also. Accounting for the randomness of some of the variables you discussed is feasible, but the accuracy of the resulting model can always be questioned until it's inputs and outputs are verified against known quantities.
So yeah, OP: well done!
This comment: good point!
3 -
Survivor are treated 1Vs1Vs1Vs1. Survivor Winrate won't feel the same. Could have long string of bad games. If Killer has string of bad games eventually MMR throws them a bone with easy Survivors.
5 -
Ok so this data is almost a year old
Data never includes DC matches
There is a problem with the data not being entirely accurate
5 -
As a CS guy by training, as much as I love encouraging people to learn about coding, you could have done this with a simple binomial distribution
It's also a bad model because one survivor dying increases the liklihood of other survivors dying, meaning the kind of simulation you've made (where survivors dying are independent events) is not representative of how results actually happen. It's far more productive to look at something like Nightlight's stats for analysis like this
18 -
We stan a Python king/queen/monarch. Lovely post OP. I see a lot of people still aren't happy (it is the DBD community after all lol). But thank you for the effort and explanations. They are very much appreciated.
3 -
Ah yeah, the weekly "math" fallacy trying to pretend killer wins only 50% of their games.
You DON'T NEED to know Python to understand the fallacy.
Because EVERY time, the trick is the SAME : draws. Instead of counting the wins / loss for both sides (which shows 60 - 40 in favor of killer) , the person will try to split into wins / draws / losses (which gives 50 - 15 - 35 instead, still in favor of killers). Then, they "magically" forget draws, and tell you "Look, killers win only 50% of their game". But the thing is, even under this convention, survivors win only 35% of the time, because 15% are draws. So this is STILL completely killer-sided.
And of course, this time is no exception.
As I said, no need to understand the full code, nor the math: we trust OP for this. The only part that we want to look for, is how they deal with draws. And how, what a surprise: notice that OP's code explicitly computes "draw rate" at line 30, but then never uses it anywhere. Only the "win rate" and "escape rate" in the graphics are shown. Isn't it weird? Why would you bother computing the "draw rate", then in the end not do anything at all with it?
The answer is - and OP perfectly knows it - that if they would show you the draw rate (or survivor win rate, even better) on the graph then it will reveal a completely opposite narrative to the one they're trying to paint. Under the convention that they created, yes killers only win 50% of their games, but survivors only win 35%, and the rest are draws.
Now in this case though we are reaching some particularly disturbing levels of manipulation, as OP added unnecessary complexity to the post in order to "scare" people away and just trust that he has done the math. But no, just because someone used Python does not mean he did "math" nor was honest about it (Python is in fact complete overkill as you could just use Nightlight data to prove the same point).
Speaking about honesty, this quote - from the main post - is particularly questionable:
All in all, the "power role" (in the theme) only wins about half the time, which mean, as a team, survivors win about half the time too, on average, overall.
Now this is a blatant lie. You know very well that YOUR OWN data shows something like 35% winrate for survivors, not 50%, and don't forget I can run your Python script myself if you don't want to reveal the numbers.
It's unfortunate that you did this experiment in an attempt to show that the game is not killer-sided, and it completely backfired: not only does the data it produces show that the game IS in fact killer-sided, but it also directly debunks that "killers win only 50% of the time" fallacy.
Redo the same graph, but show us the survivor winrate as well as the killer winrate, please..
21 -
People typically give up and throw themselves at the hook, I deal with people like that Nea that LilithOmen posted a lot more. People refusing to wiggle in my arms, I know they're going to kill themselves the moment I hook them… I just slug them if I sense they're throwing a tantrum.
TBH, I play at mid MMR and I have noticed this being a thing, Survivor mentality nowadays is really bad, and strong Killer play just makes people give up, even if it's 2 gens left and nobody is dead. It's frequently not because of "Knock Out" or "muh Blight!!!1", it's just pure "Waaah, I'm not winning!".2 -
Well, curious how this plays out. I always wondered why draws were never talked about.
4 -
Wasn't the last time the math came out it the number chart one that showed 60% kill rate was a 70% win rate. As the 60% kill rate over 10 games gave 3 2k games and 7 3k games.
6 -
SIIm always curious about how people call out some math fallacy then proceed to present survivors being able to draw with the killer.
Killer winrate + draw + survivor winrate = 1 doesnt exist in dbd. It's killrate + escape rate = 1... which only indirectly has anything to do with the win rates.
There's only one out of five outcomes of a trial in which no survivor wins. And thats the 4k.
- survivor A wins when A escapes
- survivor B wins when B escapes
- survivor C wins when C escapes
- survivor D wins when D escapes
- The killer wins if more survivors die than escape (3-4k)
BBut then people also get the following stuck in their head, even though its not enforced or hinted at ingame:
- "The survivor team" wins if there are more escapes than kills
- The "draw" happens on a 2k because kills equal escapes.
SSo now we have the problem that the killer win is in direct comparison to two different variables, the individual winrate and the team winrate. So which one of the two should be equal the killers?
IIf you go for an 50% average escape chance, the team wins ~31%, the killer wins ~31% and the rest is draws. But is it really fair for every survivor to have a ~1,6times higher chance to win than the killer?
0 -
You're mistaken. A Killer 'winning' doesn’t mean Survivors are 'losing.' Both the Killer and Survivors can win simultaneously in a match, so when there’s a draw, two Survivors also 'draw.' A 2K is the same as a 2E. The only scenarios where only one side wins are a 4K or 4E. So attributing all 2Ks as Killer wins and Survivor losses is inaccurate.
- 0k (or 4e): Clear survivor win.
- 1K (or 3e): Survivor-favored outcome.
- 2k (or 2e): Neither side exclusively "wins." Both could consider this an acceptable or neutral outcome.
- 3k (or 1e): Killer-favored outcome.
- 4k (or 0e): Clear killer win.
This leaves us with 2 possible win scenarios for the survivor side, 2 possible win scenarios for the killer side, and 1 scenario where both sides win equally. You're making it seem like 2k/2e is favoring the Killer when it's not.
Then you'd have to consider that the players using Nightlight are just a tiny fraction of the community and not representative of the actual stats of the game. According to the official data, Survivors escape around 43.5% of matches, aligning closely with the 40% escape rate (or 60% kill rate) that BHVR is targeting.
But ultimately, debating escape rates is pointless because many Survivors aren’t necessarily playing to escape. Plenty are focused on completing rift challenges, goofing around with friends, being overly altruistic, testing new builds, and so on. Without a dedicated "ranked" mode where everyone plays with the intent to escape, these stats are just one factor among many others, and trying to boil down the game’s balance to them alone is ignorant.
1 -
all this stuff also doesnt account the match dynamics.
killer can lose miserably and win over single moment because the game actually allows killer to bypass their intermediate objective and get straight to the kills. For example by getting noed snowball or simply having survivors throw the game at last gen bc they got too greedy.
survivors can have a lot of inherent advantage, but simply make a big enough mistake and lose the whole game to it irreversibly as it snowballs and goes out of control.
killer in the meantime is a mistake punisher role that certainly can seriously tank their ways to win by making mistakes or not being powerful enough, but still have a chance to snowball at any moment.
survivors only get the mechanic that skips their objective when 3/4 are dead.
it doesnt make much sense trying to compare win cons of either sides and try to balance for them because they're inherently incomparable
3 -
They do win "only" like 51% of their games, the problem is that they draw another 13% or so and thus lose less than 40%. Survivors escape 40% of the time, but that counts both wins and draws for survivors (if you're counting solo results as the win con for survivors)
It doesn't really matter because as I've often asserted, the game is at its bes lt when it favors killers in a fun way! I just think we shouldn't pretend "favoring killers in a fun way" doesn't actually count as favoring killers"
Also I personally feel like it favors killers in a not-fun way but that's personal grievance rather than objective analysis
6 -
I don't know how anyone believes the real escape rate is 40% in solo q when you take all factors into account.
Here are my stats since yesterday.
9 games, 2 escapes
However one of those escapes the killer was afk so it doesn't count. Also there was one game I died where the killer carried me to hatch and I didn't take it because mercy escapes don't count either. So if I had taken the hatch my escape rate would be 33% and devs think everything is fine based on that stat. However my actual escape rate in games where killers are were to win was actually 12.5%. It's abysmal and not worth playing this game anymore as survivor.
8 -
I should also mention there was at least one game where a survivor dc'ed, so since devs completely ignore those games in the stat my escape rate would appear to be close to 40%, when actually it was just 12.5% in games where killers were trying to win. Now the problem with believing that 40% escape rate becomes clear.
5 -
I address draws here, that's what a 2k is.
Since you didn't read it, let me parse it for you:
Just defining a killer win to be a 3k or 4k pts killers winning about 51% of the time, and draws occur about 12.5%. so survivors are winning about 37% of the time
2 -
The fact that your "mathematical model" (which is really just a simulation of rolling dice X times in a row, which has little to do with the game of DbD) predicts completely different outcomes from a real DbD data set (NightLight) is evidence enough that it is flawed, as with the sample size of that data set the mathematical model should at least come close. But it doesn't: With an average killrate of 57.24%, the escape rate obviously is even higher than 40% (42.76), and yet despite that, the killer winrate is still above 50% (51.17). Moreover and even more compelling mathematically, that data set does not suggest a simple, linear trend relation between killrates and winrates, showing various odd instances where killers can have a lower killrate than the average or another killer but a higher winrate, and vice versa. It also shows that all killers within ~1% around the 60% killrate target have above-50% winrates, ranging from 53.1% to 57.15%.
What is also telling is that you simply see everything as a loss that is not a killer win, ignoring tied outcomes. Killers do not necessarily lose if they don't win, there are also draws. Not only does this make statements like "killers lose two thirds of the time against high MMR 4-SWFs" be even more false and biased than the flawed simulation it is based on already does, but it completely ignores the survivor winrate. You know, the equivalent to the killer winrate if we're using winrate as a balancing measure, rather than killrate? And in comparing those, one would then see that certainly, an above-50% killrate will reliably lead to a higher winrate for killers than survivors. In the example of NightLight's overall average, that's the aforementioned 51.17% winrate for killers, versus a 36.26% winrate for survivors. And when looking at killers actually close to the 60% average killrate target it gets to be even more grim, such as with Pinhead: killrate of 60.77%; killer winrate of 57.15% versus survivor winrate of 32.78% - an almost 25% difference in winrates, due to a 10% difference in kill/escape rates.
I really am glad many people have at least stopped claiming "stats don't mean anything", where they engaged in mental gymnastics such as saying survivors give up so much that it skews the rates too much (which we know it doesn't since even for Merchant this was a negligible factor), ignoring instances of killers giving up or letting people escape, or saying they cannot possibly be accurate because Nurse for instance doesn't have the highest killrate, ignoring that that is completely to be expected for a global average given that she has a higher skill floor than any other killer and therefore will be dragged down the most by all the new, intermediate and average players (if you aren't good at using her ability you will lose most of the time, and the same is not true for any other killer where you can always fall back on standard chase dynamics). Not to mention that she is also among the killers most difficult to be played on controller - if we were to look at PC-only stats, I bet she would already be much higher up even on a global average scale.
The next step from acknowledging that the stats are real and carry meaning is to acknowledge the really very simple and obvious reality that a non-50/50 distribution of kill/escape rates is "sided", and 60/40 very simply and obviously killer-sided. As the devs have also stated as their literal intention with balancing the game for that average outcome. It's not to start the new mental olympics where a 60% killrate is somehow "balanced at best, maybe even survivor-sided!"…
The real kicker is that the 40% is not even the "real" escape rate: Matches with disconnects are excluded from the stats, which when included are likely a significant factor and it would not be surprising if they moved the rates by as much as 5%, meaning the escape rate would end up being 35%. Furthermore, hatch escapes are included in the stats, and yet they typically happen in matches survivors have lost. That's countless matches contributing with a "25%" escape rate despite having a 100% killer winrate. And for the "high MMR" stats specifically, it can also be noted that due to presumably only looking at the survivor players' MMR values, there are tons of matches captured where these high MMR survivors faced non-high MMR killers, given that the matchmaking is rather lax. So that further diminishes any such balancing arguments.
Also: How about you turn your own bias against yourself, and graph a "simulation" based on killrate vs. survivor winrate? But wait, we don't even have to do that, we can simply take the inverse from your current graph. If we take the blue curve as the killrate and red as survivor winrate, we arrive at a ~20% survivor winrate for the 60% killrate. Wow, so survivors are losing 80% of the time!
17 -
Thank you for serving as "Community Notes"
5 -
If the stats were reversed (60% escape rate and 40% kill rate) everyone would be losing their #########. No ifs or buts arguments.
9 -
why pushing survival rate against win rate? Its not even the same ground.
4 -
It's not 1 to 1 though. A 40% escape rate is per survivor, and there are 4 survivors in a match. You can still lose (you got sacrificed so your escape rate went down) but the other 3 survivors win resulting in a killer loss.
1 -
What does that have to do with what I said. If bhvr tomorrow released 60% escape rate and 40% kill rate people would be screaming survivor sided game at the top of their lungs (they already do but that's besides the point) and there would be no mental gymnastics to say the stats are fake, manipulated, "that's not what they're really say bro" or whatever. People pretend Dbd's still in its 2018 balance state.
8 -
That's not the point. The point is that there have been many discussions questioning the target escape rate.
This script (and now the app in ) shows a good reason for that choice.
What assumptions?
The chosen default escape rate is provided by BHVR a.ka the most accurate data we can have.
But more t than that, it's also the target chosen by BHVR. (40% escape rate, 60% kills).
It appears you misunderstand the use of the random uniform. A 40% escape rate means that, on average, a survivor escapes 40% of the time. So the program draws a random number between 0 and 100 (and it quite obviously need to be uniform) for each of the four survivors to determines if it has escaped or not. That's quite trivial.
As for the killer win, that's has always been defined like that: 2 kills: a draw, more a win, less a loss. But I'm not working on the survivor team wins, only the escape rate. Note that draws are irrelevant as they are just a "meh" game for both sides.
I'm convinced you don't understand what the program does and I don't know how to help you.
I'm not talking about the survivor team win rate. I've been very careful speaking about killer wins and survivor escapes.
The whole discussion that pushed for this is all these posters complaining about the 40% escape rate being too low.
The "throwing a bone" you are talking about is part of that target 40% escape rate. That's how BHVR is quite skilfully on that 40% target.
In this case because a draw seems quite irrelevant. Neither side has won, two survivors are sacrificed. That's a "meh" on both sides.
The program only shows, by simulation, why the 40% target from BHVR makes perfect sense.
Everything else is out of scope.
A 60% escape rate would translate to about a 18% win rate for the killer. Nobody would play killer ever (people were already leaving in droves with a 50% escape rate)
You started well understanding the way computations are made (as simply as I could so anybody could understand …)
Yet you fail to understand what the program does.
You understand that it decides if a survivor escapes by rolling a dice. Right?
So you can understand that doing that for four survivors decide if the match is a win (a.k.a more than two kills) for the killer.
Additionally, I'm assuming you understand the convergence of results for truly random sampling (of which uniform random is a decent approximation)
So if you understand all that then you have to understand the result is accurate. (I'm hoping you understand the difference between survivor escape rate and survivor team wins.)
Do it yourself if you want, you'll get the same results.
As for the arguments about players ruining matches not being counted, it's obviously the one thing to do and … it doesn't even matter.
All the program says is that a 40% escape rate gives about 47% win rate for the killer.
The reason survivors escape at 40% is utterly irrelevant.
The points remains: BHVR targets 40% and, from their own accurate measure, get these results when players aren't rage-quitting.
People have complained about that 40% target and this shows 40% is only giving a 47% win rate for a killer.
Everything else is a moot point.
1 -
Stats are old, all you have to do is play the game to see the devs aren't right about anything honestly lol.
4 -
The stats are only used for the default value of the escape rate. (Besides, from a recent post from a dev/modo (can't recall the name: too tired), the current stats are not significantly different from these.)
You give the program an escape rate and you see how it translates in term of killer win. That's all it does.
I wrote it because of the complains about the 40% escape rate target (a.k.a 60% kill rate target)
(If you look at the top, there is a web app to toy with the numbers)
3 -
Isn't the issue more that your program assigns a comparable weight to each survivor escaping and doesn't reflect that once a survivor dies the chance for the remaining survivors to also escape is less than 40% so the weighting is inaccurate?
I'm just asking for curiousity by the way. I'm fine with the target escape rate being 40%; I just want the game to be fun by making game mechanics favour spreading hooks so the game is more enjoyable. However, I'm not invested in what the escape rate is as long as the game is fun to play for both roles.
3 -
Alright.
Since OP refuses to share the draw rate, I ran the program myself. Result:
- Wins : 47.5% (3K or 4K)
- Draws : 34.7% (2K)
- Losses : 17.7% (1K or 0K)
This is literally the output of the program shared by OP, from which he claims that the game is balanced. Anyone can run it and see these numbers, just by copy pasting the code in a Python interpreter.
So let me quickly summarize.
- Killers win 47.5% of the time. Survivors win 17.7% of the time, 3 times less. The rest are draws.
- OP saw these numbers, and concluded that "Killer winrate is only 47% so the game is balanced".
- When asked to share the full numbers, especially that draw rate, which reveals the trick, OP kept refusing, even saying that draws are "irrelevant".
12 -
Your simulation is proven wrong by taking 0.1 seconds to look at nightlight and seeing that a 2k is the least common outcome, not the most common. Wax lyrical all you want, every piece of hard data taken from actual gameplay shows that survivors making it out within the same game are not independent events. thus your simulation, which treats them as if they were, can tell us nothing about the actual state of winrates.
If you disagree, feel free to go to the hundreds of people who have ever complained about people choosing to go next on first hook and tell them that actually, the phenomenon of going next should have no effect on their chances.
9 -
That's interesting. That correlates with how when I recorded my most recent set of 100 survivor matches I saw that a 4K result was 2.5 times more likely than a 4E result.
1 -
It's a good question.
The data from BHVR says each survivor has a 40% chance of escaping on average, no in-game condition attached.
On the web app, I draw the actual mean escape rate of the survivors (in blue). It converges close to the chosen rate.
The way I see it, the reason survivors have escaped isn't really relevant, only that they have escaped at a rate close to that specified value. And of course the main information is the deduced killer win rate for said value.
I've given the program so people could play with it "refuse to share the value". Conspiracy theorist much?
The draws are irrelevant as they mean that both teams have got a null.
"proven wrong" then proceeds speaking about something I've specifically said was irrelevant.
I don't believe you understand what this is about.
2 -
The simulation also doesn't take in any current tactics of the game or skill of players its just a random roll generator. So of course it will keep the hard 40% floor escape rate no mater what. It is just like the dps simulators of RPG games to sim gear 99% of the time it takes in an unplayable skill level and imposable circumstances for the numbers it puts out.
1 -
You've said it's irrelevant but you're incorrect. The fact that kills per game doesn't follow a normal distribution in practice is very relevant to a simulation that does conform to that distribution.
8 -
So, no comment on the 50% wins - 32% draws - 18% losses part?
Your OWN EXPERIMENT returns a 18% win rate for the survivor team, which most people would agree is killer-sided. Any reason you have "omitted" to share this part of the output of the program?
I feel like people reading your topic would have found this 18% number at least informative, don't you agree?
Any reason you have only shared the 50% winrate of the killer, silently implying that survivors win the other 50%, when YOUR OWN PROGRAM shows the rest if made almost only of draws, and very rarely do survivors actually win?
9 -
I ain’t no math expert, but, the simplest way of looking at it is that each survivor has a 40% chance of escaping, there’s four of those. Killer to win the match needs to surpass the odds of a 40% escaping survivor, 3 times.
Just saying that the 60/40 is not the end of all things. There’s a bunch of factors you have to consider:
1 - Each individual survivor has a 40% chance of escaping, which means that to achieve a “win” the killer has to beat those 40% odds three times.2- Killers are vastly different in strength, survivors can be considered equal because there’s no real difference. So top meta Killers could be carrying weaker killers, which in return could be having horrible games.
Saying that survivors should have a 50% chance of escaping would make the already hard role of playing killer, way harder, because it’s not a 1vs1, it’s a 1vs4 and basically you need 3 to win.0 -
Okay, thanks for explaining.
1 -
@Saiph isn’t being conspiratorial. It is incredibly suspicious you omitted key information and as a layperson yes the information you withheld completely changes the veracity of your argument. If I hadn’t read Saiph, @ratcoffee @AmpersandUnderscore and @zarr ‘s comments I wouldn’t have considered that by your model survivors only win 18% of their matches. It absolutely seems to me, as a layperson who knows nothing about statistical modeling, that you were implying survivors win as often as killers do when that’s not even close to being true here. I believe you should add some sort of caveat to your initial post so people aren’t deceived.
Kudos to those I’ve tagged for taking the time to provide that extra information breakdown. It made a difference for me.
13 -
The 40% escape rate is the result from millions of matches. That means over many matches, on average less than 2 survivors escaped. In a match where 3 survivors escape, you have an escape result of 75%. If that were to happen more often than not, the overall resulting rate would be higher than 40%… But if you want to have it in individual terms: in these captured stats, an individual survivor escaped on average 40% of their matches, less than half.
People really have trouble making sense of stats in an asymmetrical game. But it really is as easy as 50% kills/escapes = perfectly balanced. The 60/40% target was chosen deliberately to make killers perform better on average than survivors. BHVR literally said "To ensure the Killer feels like an unstoppable force to be feared, we’re making slight improvements to many aspects of Killer gameplay" when they aimed for the 60% from former 50%.
But I could employ the same mental gymnastics for a 60% escape rate (that the people arguing a 60% killrate is totally balanced and needed and if anything too low would absolutely lose their mind over): since survivor groups are 4 players, more players lose when survivors die, whereas the killer wins against any survivor they kill - so the aim should really be a 25% killrate, because in that case only 1 player loses, 3 players win, and the killer still wins against the player that lost! Win-win! But 40% killrate is alright I guess, because that's still on average less than two survivors dying.
Note that draws are irrelevant as they are just a "meh" game for both sides.
This statement not only doesn't make any sense (we are looking at whether the game is "balanced" based on match outcomes, we obviously can't just ignore a huge chunk of outcomes because we feel like it, much less say they are "meh" results when they are what most people think of as a balanced match), but just because you think draws are "meh" or "irrelevant" still does not make them killer losses, which is what you implied multiple times.
I've been very careful speaking about killer wins and survivor escapes.
That you have been indeed. Care to enlighten us why comparing survivor escape rate to killer winrate is a meaningful balance measure? As opposed to survivor winrate versus killer winrate? Or again, how would you like if we make it about killrate versus survivor winrate? Which, again, your simulation predicts to fall around 17-20% for a 60% killrate. Survivors are losing 80+% of the time according to your simulation and your logic, yet that "makes perfect sense" from your balancing perspective.
A 60% escape rate would translate to about a 18% win rate for the killer. Nobody would play killer ever (people were already leaving in droves with a 50% escape rate)
Ah, but an 18% winrate for survivors is perfectly fine. Also: The "killers are leaving in droves" and "enjoy your infinite queue time!" bogeyman has never been real. I've played this game regularly on and off throughout most of its lifetime, and there was never a time where survivor queues were prohibitively long. The longest they've been was 5-10 minutes exclusively for SWFs back in 2017. I fail to remember even having had a >5-minute survivor queue since, though possibly with rare exceptions of course.
I do agree with BHVR that 60% killrates are probably better for the game, because the game trending killer-sided fits its theme.
Additionally, I'm assuming you understand the convergence of results for truly random sampling (of which uniform random is a decent approximation)
It isn't truly random though. The escape "rate" does not say the game is a system designed to bring about a 40% chance for any survivor in any given trial to escape. The 40% simply says: in these millions of captured real trials, 40% of survivors escaped. I think you will understand that if you were to actually look at the trials, you would find that the 40% is completely inaccurate for countless of them because the game is not a neat mathematical system where uniformly random chance exists. If two survivors have died in a trial, do the remaining two still have a 40% chance of escaping?… They can use these overall stats to steer rough balance trends within the game, but it does not accurately depict the reality of any trial, and it simply cannot reliably show things such as the winrates.
The fact alone that different survivor group sizes and different killers have different escape/killrates should be hint enough to understand that the overall 40% obviously does not describe the state of the game in any given trial as if it were a logical, consistent system.
And this is even more easily recognized again by looking at real data sets that capture game reality. NightLight shows that in reality, not only is it possible to have above-50% winrates for 60% killrates, but it happens every time, and none of the real results align with your simulated ones. Real results here consistently add up to something that completely contradicts your "model". I struggle to understand how you could be convinced that if BHVR were to actually share data on kill distributions they would be close to your predictions. Just because overall some 40% of survivors escaped does not at all mean these distributions will follow a predictable, simulatable pattern.
Do it yourself if you want, you'll get the same results.
I did, and like @Saiph I saw the "l: 17.79%" and had to chuckle. And I already pointed to this by inverting the implications of your graph. The killer lossrate/survivor winrate is below 20% for a 40% escape rate according to you but this finds no mention anywhere and is apparently perfectly reasonable and benign. Mind you, as much as I would like to take this and run with it to use your "proof" that the game is incredibly rigged against survivors succeeding, this is mindbogglingly low and just another "hint" that your simulation is completely removed from reality; I don't actually think 3/4-escapes only constitute ~17% of real games in DbD (they certainly don't in the real games captured on NightLight). But it's funny that you refuse to mention or acknowledge this result of your simulation. But based on your consistent insistence on distinguishing between individual survivors and "teams" I'm sure you will have some impressive gymnastics to perform as to why survivor group winrates don't mean anything and don't matter at all.
The reason survivors escape at 40% is utterly irrelevant.
Only if you think the game is some mathematical system. Hatch escapes are different in a gameplay sense from gate escapes, most of all of course because they can and regularly do happen in matches survivors had gotten completely decimated in. They skew the idea that "40% of survivors escaped" means that those players had games in which they did well, finished gens, and escaped through gates. Tons of the escapes captured by the "40%" will be hatch escapes in matches they lost, often decisively.
I'm also not only talking about reasons why 40% of survivors escaped in the matches their stats cover, but that the stats don't capture all matches and that it stands to be reasoned that if they did the result would be less than 40%. Or that if you looked at specific subsets of matches they would be (e. g. "high MMR" killers against non-SWFs).
Everything else is a moot point.
Not at all. Pointing out that a killer winrate of 47/37% (even if it would be true) doesn't necessitate a 53/63% killer lossrate/survivor winrate and that comparing escape rate to killer winrate (rather than escape rate to kill rate or survivor winrate to killer winrate) isn't a fair or sensible balance proposition isn't moot. It's crucial to understand the comical levels of bias at play here.
But luckily it ultimately doesn't matter whether you are actually wrong or right about the real numbers, for one thing because I don't personally have a problem with BHVR's 60% killrate target (with the caveat that SWFs should be expected to perform better than solos for obvious reasons and come closer to 50%, and that they should make sure solos don't get left behind too much as a result of only looking at overall escapes), and for another because no matter what we think the numbers are, BHVR knows for sure, and no matter what we think the right approach may be, BHVR does whatever they want anyway.
8 -
What are you even talking about. There is only one information available about the kills and it's 60%, from where are you pulling that gaussian? Try to be more explicit.
I believe you are correct. And I should never use the word "math", it makes it looks way more complicated than what it is.
It's all about the perception of the game.
Killer wins are defined as seen above. (I'm too tired to repeat all that)
Survivors (plural) win should naturally be the opposite but in practice it isn't. The vast majority of survivors only focus on their own escape: survivor (singular) win. Hence that regular ranting about 60% kill rate. That's why I looked at the win the way they are perceived.
But one of the points is that 50% gives a very low rate of win for the killer, which may explain why a while ago when the kill rate was balance for 50% survivor queues went up: it was high-mmr even at the basic killer level.
Draws are irrelevant because they are not a win. I'm not sure in English but in French it's called a "mach nul", one meaning being "the same as if it never happened" (nul et non avenu). They dont really have an equivalent for singular survivor wins as it's kind of all of nothing on that front. Nothing really interesting to take from that.
Yes, I've seen the 17% too, which is interesting, or would be if survivors cared that much about team wins. It's all about perception. (And I realize I've not updated the web app to also draw that curve, I'll do it if I recall tomorrow.)
We keep speaking about 40% escape rate mind you but the 48% escape rate in high MMR is also intersting (and explains why that level of play can be so frustrating). I know this detail doesn't escapes you and you expect it to be like that, as do I.
I see everything as system, that aside you may have a point directly in relation with mine. Let me think …
Yes, you are right. Hatch escapes are inflating the escape rate somewhat. Win perception on the survivor side may be lower than projected because of that. I'd love to be able to discriminate them out. There is no data about this though. If you can find some BHVR stat about hatch escapes, I'd love to try to include them.
47/37 I presume you are talking about the measured win rate results of the matches. It's about perception here again. I understand not everybody sees it like that but, for me, if it's not a win, its a loss (I guess I'm extremely competitive).
I fully agree opinion may diverge about this. I've simply provided a tool to have a better idea of what these statistics mean.
If other people draw different conclusions than mine and don't see why I'm only focused on win, it's all the same to me. (Once I believe I've explained myself sufficiently, and I believe I did.)
Opinions are subjective after all. I'd not call either of them biased though.
I'm perfectly happy with my survivor escape rate (about 60%*) and killer kill rate (about … a lot, I don't dare to say how high, with my main**), neither of which are matching these stats. But as you probably know, personal observations are hardly representative.
*) and I consider myself pretty average: it's mostly luck and stealth.
**) Nurse, relatively decent at it if I say so myself.
0
