The second iteration of 2v8 is now LIVE - find out more information here: https://forums.bhvr.com/dead-by-daylight/kb/articles/480-2v8-developer-update

A reddit user did an experiment. Let's discuss the results.

13

Comments

  • RFSa09
    RFSa09 Member Posts: 818

    That means that he played 100 matches as trapper.. and won a lot of them... nothing more nothing less

  • mizark3
    mizark3 Member Posts: 2,253

    I'm sorry, but read the rest of the sentence that you didn't underline: "… AND (emphasis mine) support the horror theme of the game, where the Killer is a force to be reckoned with and the survival is not guaranteed."

    So the first part that you underlined is to say 'hey it needs to be close', and the second part is saying 'but it does need to be favored for Killer'.

    Now personally I don't take issue with the game being designed around being Killer sided, as long as Survivors are designed to have fun even while losing. My major pain points here are bleedouts, and Survivors Kobing on hook (to die) since that is favored over DCing to give a bot. Rats are too much of a gray zone, since you can rat for the benefit of the team or detriment of the team (eg. death hook hiding or never hooked hiding respectively).

  • NerfDHalready
    NerfDHalready Member Posts: 1,749

    3 pages?? i don't understand what about this post sparked this much of a discourse. like, does this post mean killers should get nerfed? trapper is too strong? what is it?

    you don't go this deep into balance analysis based on soloq matchmaking. yes you can work with perkless and trapless trapper against a soloq team. just find and tunnel that looking straight default feng and tunnel them, if she doesn't give up halfway one of her teammates will. camp basement and noone will ever trade. what's the point being made if there is any, yes an even perkless and trapless trapper playing well is way way stronger than 4 randomly chosen soloq survivors (mmr is a myth).

  • ratcoffee
    ratcoffee Member Posts: 1,475

    The point I was trying to make was just that people who say "it's impossible to win as killer without slowdown, particularly as a low tier killer" are incorrect.

    A lot of people, i think, assumed I was trying to make the point that killer is extremely OP and came in arguing against that point.

    Also, there are some people arguing as to whether you can surmise rates of specific results from kill rate, an argument that happens whenever kill rate comes up. (Incidentally, the answer is no. Any win rate extrapolated solely from kill rate data is statistically useless)

  • ratcoffee
    ratcoffee Member Posts: 1,475

    I've seen his posts before and he understands how to calculate an average, the problem is he doesn't understand why using an average for data that is not normally distributes is statistically speaking a mistake

  • NerfDHalready
    NerfDHalready Member Posts: 1,749

    i see, and agree. on equal skill level killers should have a hard time even if they try but definitely not against soloq teams because matchmaking is an absolute joke.

  • Batusalen
    Batusalen Member Posts: 1,323
    edited August 18

    So when you say 'mathematically speaking' you don't have enough data to arrive at a mathematical conclusion.

    And what more data do you need to calculate that a 60% killrate on average means 2.4 survivors killed per match on average? Because the only thing you are proving here is that you also don't know what an average means.

    The funny thing for me is that you are, again, trying to fight math with "how to get equal win rates in reality" without realizing first that any other killrate would make it mathematically impossible to achieve equality because of how MMR win conditions work (more than 60% killrate, survivors would always lose more on average, less than that, and killers would lose more on average). Second, that you have precisely pointed out the part where I said:

    […] as in reality, getting to an equal 40/40% winrate is more complicated than simply getting to that killrate

    But again, keep trying to fight basic math, it will never get old!

    I think you're misreading what the devs have said. They want the killers to have a higher kill rate because they want killers to feel strong and win more

    Then explain to me, again, how having a 60% killrate would "keep matches relatively even" in comparison with a 50% killrate, according to the devs own words? Because I think it is pretty clear taking in count the phrasing:

    "[…] a 50% kill rate would mean they kill two Survivors per match on average" ⇒ "We try to keep Killers near a 60% kill rate on average to keep matches relatively even"

    Ergo, a 60% means more survivors killed per match on average (2.4 to be exact), which means killers now win as much as survivors, which makes "matches relatively even". It's not rocket science, it's basic math and logic.

    Either way, you didn't explain why that second part is relevant to the 60% killrate and the math behind it, so I'm going to keep it out of the underline.

    So the first part that you underlined is to say 'hey it needs to be close', and the second part is saying 'but it does need to be favored for Killer'.

    But that's the thing, if you do the math and see what they are saying, 60% killrate is not only pretty close, it is in fact the only killrate that allows for an equal winrate mathematically speaking (as again, getting to an equal 40/40% winrate is more complicated than simply having the killrate at 60%, but it can't be achieved without it being 60%).

    But hey, if you want to focus on that second part that they probably said because they would never talk about MMR or win conditions directly, I will apply to you the BHVR go-to answer: Whatever makes you happy.

    Post edited by Batusalen on
  • mizark3
    mizark3 Member Posts: 2,253

    I'm pretty sure someone did the math before, and a 60% killrate means closer to an average 60% winrate, not 50%.

    I'll try to walk through the idea conceptually. Sets of 5 matches, 20 Survivors, 12 kills, 8 escapes. 60% killrate.

    Cases with Escape/Kill

    1.Most 4E 2.Most 3E 3.Most 2E 4.Most 3K 5.Most 3K with Most 3E (or 2E)

    (If there are more possible permutations feel free to include them in your reply)

    EEEE EEEK EEKK EEEE EEEK
    EEEE EEEK EEKK EKKK EEKK
    KKKK EEKK EEKK EKKK EKKK
    KKKK KKKK EEKK EKKK EKKK
    KKKK KKKK KKKK EKKK EKKK

    Killer/Survivor/Draw %s

    60/40/0 40/40/20 20/0/80 80/20/0 60/20/20 - using these for the averages, same result as #s

    Killer/Survivor/Draw #s

    3/2/0 2/2/1 1/0/4 4/1/0 3/1/1

    AVG. 52% Killer win, 24% Survivor win, 24% Draw.
    No draws is 68% Killer win, 32% Survivor win.
    (260/500 or 260/380 for Killer, 120/500 or 120/380 for Surv, 120/500 for Draw)
    Heck, even being favorable and splitting Draws 50-50 (rather than 60-40) for both gives 64/36 Killer favored.

    IME condition 4 or 5 is probably the most common, but mileage may vary. So while theoretically condition 2 fits your premise, all the other cases still exist and are contributing to the math. Plus I'm sure law of large numbers would converge on a 60/40 winrate with draws thrown out, but it still looks dire with Killer winning roughly double as Survivor on average with or without draws.

    One flaw is we don't know how Hatch is counted. Do they count it as an escape, or do they discard the result as a draw like they said they treat it as such for MMR? Hypothetically if the 3Ks were all Hatch escapes across the conditions, and they throw out Hatches, that could add an extra 7 results. That could add EKKK (Hatch for an 8th result extra) into EEEK (to maintain ~60% kill rate with 4/7K 58% since the Hatch didn't count), adjusting the stats to 52% KW, 26% SW, 22% Draw, or 67% KW, 33% SW with draws thrown out.

    So yeah, overall Killer has double the winrate of Survivor with a quick math check.

    (I figured out 5 more missing cases before I posted this: 4E/3E/1E/0E/0E, 4E/2E/2E/0E/0E, 4E/2E/1E/1E/0E, 3E/2E/2E/1E/0E, and 2E/2E/2E/1E/1E for a cumulative total of 50 matches, with 25 Killer wins (50%), 14 Draws (28%), and 11 Survivor wins (22%) or without draws ~69% KW, ~31% SW. That makes the odds even worse for Survivor wins in both draw or no draw scenarios.)

  • Firellius
    Firellius Member Posts: 4,387

    I don't think he does understand, because the way he responded shows he really does not.

    And what more data do you need to calculate that a 60% killrate 

    on average

     means 2.4 survivors killed per match 

    on average

    ? Because the only thing you are proving here is that you also don't know what an average means.

    You keep dodging this!

    Imma just spell it out, all the way through for you.

    The way YOU, specifically, calculate winrates from killrates here also mandates a 0% average winrate from any killrate under 50%. You have admitted as much yourself.

    Now, you calculate an average by taking all the numbers in a data-set and adding them together, then dividing them by the count of numbers in that data-set.

    In order for an average to be zero, one of the following two MUST be true:

    1. The dataset contains negative values.
    2. There is no value other than zero to be counted in the dataset.

    Since we're talking about winrates here, and it is obviously impossible for winrates to be -negative-, the second rule MUST be true if you want to reach an average of 0.

    So when you claim that, by the method you use, a kill-rate of 40% would result in a 0% average winrate, you are, by mathematical definition, claiming that within that dataset, there is not a single winrate above 0%, which would require that not a single match ended in a 3K or higher.

    So for example: 1000 matches at 40% killrate, meaning 1600 survivors killed, according to your model, can not possibly include ANY match where 3 or 4 survivors were killed.

    Because if there's even ONE match with a 3 or 4K, SOMEONE in that dataset has a winrate above zero, which means that the average winrate can no longer be zero.

    Do you understand this?

  • Batusalen
    Batusalen Member Posts: 1,323

    Then explain. Multiple people in the thread are laying out why you are incorrect and you're just saying 'you don't get what an average' means.

    Because every single one of you proved that you don't know what "on average" means. Not only that, you don't even have a grasp of why dealing with averages is needed in this game and not "real kills distribution" for overall balancing.

    So, what do I have to explain? Why you don't need any more data than what has already been mentioned to make an average?

    I don't care if win rates are equal or not. That's a game design decision. I'm saying your 'math' doesn't make sense.

    What doesn't make sense is this response to what you are quoting. But again, the kill distribution doesn't have anything to do with the average kill rate, or better said, the kill distribution doesn't contradict the math and logic behind the killrate in any way. That's what you people still don't understand, and that's why you keep using it as an argument to try and fight math itself.

    The relatively even is why not make it 70% for an uber horror theme. Your missing both the word relatively and the second portion of the sentence.

    And talking about not making sense. But the reason to not make it 70% is that then it won't be "relatively even", it would be killer sided. It is so hard to understand?

  • Batusalen
    Batusalen Member Posts: 1,323
    edited August 18

    Except 50% killrate is 2 survivors killed per match on average, and because a win is 3K or more, that means a 0% winrate on average. In other words, that 0% is not the result of doing the average of kill rates, as you are saying.

    That you are trying to give anyone lessons when you have proved again and again that you don't understand even the most basic math is outstanding.

  • Batusalen
    Batusalen Member Posts: 1,323
    edited August 18

    Now "an average" is "super tortured logic" and trolling. My entity…

    A 2K on average means that killers have a 0% win rate on average because there will be people that do more than that 2K and people that do less. That's why it is AN AVERAGE. It is pure, simple, basic math.

    And yes, anything more than a 2K would be a win, as the MMR win conditions state that a 3K is a win, but no, anything less won't be a survivor win because survivors have their own win conditions, and that's one of the reasons why BHVR doesn't deal with win rates or pure kill distributions, because then it will be impossible to balance the game between killers and survivors thanks to the total and complete asymmetrical nature of it.

    And no, a "2.4K" on average won't mean a 100% winrate because you have the ".4" part, and if you do the math, it means that killers get a 3K about 40% of the time on average, and that's why a 60% kill rate / 2.4 survivors killed per match on average means a 40% win rate for both sides on average.

    Seriously, I'd pay to see how some people's brains work.

  • Firellius
    Firellius Member Posts: 4,387

    Dude, a 2K 

    on average

     means that killers have a 0% win rate 

    on average

     because there will be people that do more than that 2K and people that do less.

    Can you get a negative winrate?

  • crogers271
    crogers271 Member Posts: 1,821

    I'm trying to figure out where he's going with this. Like I get the misunderstanding between average and uneven data sets, but this? I'm wondering if there has been some massive communication at some point because saying a killer has an average winrate of 0% is nonsensical.

  • jesterkind
    jesterkind Member Posts: 7,833

    I think the point of miscommunication here miiiight be something like, when they say "average winrate of 0%" what they mean is that the average outcome isn't a win?

    That statement would kinda make more sense based on what they're saying overall, like, the difference between killers never winning and the average outcome being a draw or a loss but wins still happening here and there.

    I'm not good enough with numbers to know if that's anything on its own merits, but it does seem like what they're saying.

  • crogers271
    crogers271 Member Posts: 1,821

    I think the point of miscommunication here miiiight be something like, when they say "average winrate of 0%" what they mean is that the average outcome isn't a win?

    It's possible that he's saying average to refer to the 'median' (the mid point of a group of numbers) instead of the 'mean' (the average of the number). Except I don't know why the median would be relevant to the discussion.

  • Firellius
    Firellius Member Posts: 4,387

    Median wouldn't work either, because if there's any winrate above 0%, the median could not be 0%. Runs into the same issue, effectively.

  • Moonras2
    Moonras2 Member Posts: 380
    edited August 18

    Sorry I don't mean to direct all this right at you. Im not great with numbers or math but it all feels thrown off for me. It doesn't help that considering draws to be a half a win appears to be subjective. If we counted draws as half a win and half a loss would that mean that killrate = winrate or am I wrong on that?

    Something else that seems to throw me off. Some don't want to claim a draw as half a win, which is fine. But then they only want to say that a draw is not a win for the killer, or 0% winrate, but is still a win for 2 survivors. But they leave out the fact that it's also not a loss for killer but is a loss for 2 survivors.

    So then they say it needs to be balanced towards killer to balance out winrate. But does that mean the loss rate for survivors becomes higher than killer? What makes the loss rate for survivor less important than winrate for killer? Losing more than drawing or winning may feel worse than the opposite.

    The whole thing confuses me lol.

  • crogers271
    crogers271 Member Posts: 1,821

    It could if you create a data set that just compromises three variables win/loss/draw. You could get a median result that is not 'win' and then mistakenly call that an average of 0% winrate.

    But I'm just wildly guessing at possible logic at this point.

  • Batusalen
    Batusalen Member Posts: 1,323
    edited August 19

    Like, how hard is it to understand that if the average killrate is 2K, and a 2K is a draw, then on average killers never win, as they draw? I mean, seriously. You guys are seriously not getting that...

    Post edited by BoxGhost on
  • Moonras2
    Moonras2 Member Posts: 380
    edited August 19

    Im starting to think it's all subjective in which way a person thinks the balance should tilt.

    To your point, I would say that would depend on the number of matches played combined with the outcome of each match. Average does not always equal exact. An average kill rate of 2k doesn't mean every match ended in a 2k. If the average of all matches is not exactly a 2k every game, then there will be a mix of wins/losses. Could the killer even win more than they lose, or lose more than they win, but still have an average of 2k?

    If exactly every match ended in a 2k then that would depend on how you see it. If you see it as a draw that just nullifies the killers match, then the killer neither won nor lost. While the survivors have 50/50 win and lose. Or you could look at a draw as half a win, half a loss, which starts to balance out.

    If for some reason you see not winning, but a draw, as a loss. Then there is probably someone else who thinks wow, I managed to get a draw, that's a win to me.

    Do both sides stand a chance or is the killer always guaranteed a draw? While survivors always know at least 2 will lose? This could mean 2 survivors constantly have a 0% winrate as well, while the other two could possibly stay at 100%. Or anything in between. Depending on teammates and luck of the draw.

    So at that point is it more fair to even out for wins, losses, or just call it even across the board? To keep in line with the horror theme as they said, it seems to go more towards the win side than the loss side.

    However, that would just be game balance wise. If you go by MMR and a match ends in a 2k. It may nullify the killers gain/loss but at least 2 survivors may gain MMR and 2 survivors may lose MMR. Which just separates the survivor side even more. Because if they were counted as "one", then the survivors would also be nullified. Since they are not, that means at least two survivors are going to raise in mmr even though the other two teammates may have played the bigger role. But that's just part of scoring individually.

  • crogers271
    crogers271 Member Posts: 1,821

    Like, how hard is it to understand that if the average killrate is 2K, and a 2K is a draw, then on average killers never win, as they draw? I mean, seriously. You guys are seriously not getting that...

    Because that's not what you've been arguing/stating. To go back to one of your earlier posts:

    Why? Because of how the MMR works, only a 3K or more is a win for the killer, and survivors only need to escape themselves to win. Ergo, a 60% kill rate means a 40% win rate on average (I repeat: ON AVERAGE and I can't emphasize it more) 

    In your most recent post you are using the phrase 'on average' to mean the most common result, but in the earlier you are using average to calculate a percentage (i.e. a mean). The concept of a winrate calls for a percentage, not a general statement. This is why saying things like 0% winrate are so non-sensical.

    Beyond that, a 60% killrate is not needed to ensure parity. That's still wrong. Technically speaking, you could have a kill rate below 50% and still have a winrate above 50% based on game design (that's not true of DbD, but I'm back to pointing out the importance of uneven data sets).

  • mizark3
    mizark3 Member Posts: 2,253

    The problem is that every average is made up of many numbers. A 2k Average can be 2K/2E, or a 4E and a 4K, or a 3K and a 3E. In those 3 cases, only 1 of them has a 0% winrate, and even then it is a 100% draw rate. Heck, even draws are often considered a half win, or 1 point for a draw and 3 for a win, so even in the most extreme circumstances draws still count towards win counts.

    Then I truly don't understand how you can counter your own points in the 2nd and 3rd paragraphs. "And yes, anything more than 2k would be a win, …" in the context of Killer, then contradict yourself saying "2.4k on average won't mean a 100% winrate because you have the .4 part…". You tried to justify that by "if you do the math", but I literally did the math in the previous post which showed closer to ~70/30 winrates for Killer at 60% killrates, or ~50/25/25 if you include draws.

    You ignored all the math and did none yourself, and are jumping hoops of logic to suit your narrative even when faced with hard evidence to the contrary. That's why everyone is thinking you are the one trolling. When you ignore what must happen to make an average to push the idea that 2K average means 3Ks and 4Ks CANNOT exist, then it is impossible for you to not be joking.

  • Firellius
    Firellius Member Posts: 4,387
    edited August 19

    We're getting it. But you're making a mistake here.

    You're describing the outcome of the average match, not the average winrate.

    Your claim is that 60% KR is the only way to balance it, because that produces a 40% average winrate. Now, due to the volatility of DBD, a 60% killrate can theoretically get as high as 80% winrate, if all outcomes are distributed only across 0Ks and 3Ks.

    Your argument here is that 'some people will get a higher winrate, some people will get a lower winrate, and the average winrate will thus become 40%'. And it's true that 60% KR can go lower. If you distribute it between only 2Ks and 4Ks, the winrate can go as low as 20%, I believe.

    But this is where the 50% KR example comes in to show why the assumption that things will average out to your projected ideal is an incorrect one.

    With a 50% KR, the outcome of the average match is a draw, not a win. So you say that, on average, killers will win 0% of their matches.

    Some will get higher than 0% winrate. That is most certainly true, but here is the problem:

    No one can get lower than 0% winrate.

    Which means that you don't have the counterweight to pull the average back down to 0% in case someone scores higher. No one is sub-0% in winrate, as that is an impossibility. You can't win -5% of your matches. You can't win negative three matches.

    So the only way to get that 0% average winrate, is to have 0% winrate across the board.

    Which means there can't be a single win. On a 50% KR.

    Now, I think you and I both know that that is preposterous.

    Which by extension means that your projection is also preposterous.

    This is why what you're talking about is not the average winrate, but merely the outcome of the average match.

    And the average match is one that no one plays.

    Post edited by Firellius on
  • Neaxolotl
    Neaxolotl Member Posts: 1,477

    Ah look, we are still discussing as if kill rates means anything meaningful

    It's quite obvious survivors are fundamentally easier to lose just because of stupidity(mostly for teammates and challenges) rather than game balance, while not so much for killers, it only makes sense to aim for slightly higher KR especially if they want to make a horror game

    It doesn't really matter if numbers are 50/50 or 60/40 or 70/30, because none of those numbers shows actual game health, if devs decided to aim for 60% there is a reason for it, and it's not hard to understand slightly higher KR is basically healthier for the game

  • Moonras2
    Moonras2 Member Posts: 380

    They've stated why they aim for the higher kill rates as well as the fact that they try not to bring win rates into it. That doesn't mean we can't discuss or give our opinions on the matter.

    I was happy with the game in the state that it was in when I first started. But I also knew it needed the many changes it's gotten along the way. Just like I don't think those changes need to stop yet. Without feedback, combined with their own stats, we would still be playing that game as it was when I started.

  • mizark3
    mizark3 Member Posts: 2,253

    Yeah that is true. The game was at its healthiest for everyone before patch 6.1, and that had ~50-55% killrates on average IIRC. BHVR's attempts to raise kill rates destabilized the game, and gave us the worst metas of all time (Call of Brine+Overcharge+Eruption and camping 3 gens from the onset). Although they did reach their much desired 60% killrates lol (and passed it, once more and more people caught on to 3 genning as their only game plan).

  • Batusalen
    Batusalen Member Posts: 1,323
    edited August 19

    In your most recent post you are using the phrase 'on average' to mean the most common result […] The concept of a winrate calls for a percentage, not a general statement

    No, I'm using 'on average' because we are talking about averages. Averages that can be expressed as other averages, like a percentage win rate:

    • 60% killrate on average = 2.4 survivors killed per match on average.
    • A killer win is 3K, so 2.4 survivors killed per match on average = killers win 40% of the time approx on average.

    Or:

    • 50% killrate on average = 2 survivors killed per match on average.
    • A killer doing a 2K is a draw, so 2 survivors killed per match on average = killers draw 100% of their games on average, or in other words, win 0% of their games on average.

    Or (and this one is for you, @Firellius):

    • 25% killrate on average = 1 survivors killed per match on average.
    • A killer doing a 1K is a loss, so 1 survivors killed per match on average = killers lose 100% of their games on average, or in other words, win 0% of their games on average (no negative winrate, mate).

    This is why saying things like 0% winrate are so non-sensical

    It makes sense, but you either, again, don't understand what an average means, or you are simply refusing to accept what basic math and logic say. Simple as that.

    Beyond that, a 60% killrate is not needed to ensure parity. That's still wrong. Technically speaking, you could have a kill rate below 50% and still have a winrate above 50% based on game design

    The next phrase…

    (that's not true of DbD, but I'm back to pointing out the importance of uneven data sets).

    Now, let's forget you contradicted yourself for a moment, and please explain to me how you pretend to balance a game like DbD while granting everyone an equal chance of winning with just "uneven data sets" (which I'm going to guess are "kills distribution" in this case) when:

    • The win conditions for both sides are completely asymmetrical.
    • You can't directly compare skill levels of both sides, as they require completely different sets of skills and game sense.
    • One side is a team of 4 players that can have different skill levels.
    • That side is not even considered a team per win conditions, and each individual player in it will win or lose individually.
    • You have 5 possible match outcomes, which define the wins or losses of 5 different players.

    The answer is simple: You don't, you can't. Starting with not even having a reference to what would be a "balanced point". Because yes, if a killer is doing a lot of 3/4Ks we could assume he is "overperforming", but what would be the point it should be then? What combination of "Ks" in the kill distribution will make him "balanced"? How would you even guarantee that among all the players that use that killer the kill distribution will be even so he is balanced at any skill level? Etc.

    So, even if "uneven data sets" can help identify if an individual killer needs balancing, it doesn't work in terms of overall balancing and will not help balance killers against survivors. That's why then we look at our win conditions, we notice that the common denominator is "kills", so we use an average kill rate as the target balancing point. And look at that, if you do the math it even provides a "sweet spot" that allows us to guarantee an equal chance of winning for both sides (60%, 2.4 survivors killed per match on average, 40% win rate on average) without even having to worry about skill levels, side, character used, or anything else.

    When we reach that point, now we can use the kill distribution of each killer to balance those that are overperforming or underperforming until we get a balanced average (as again, multiple players with different skill levels will play the same characters, so part of the playerbase will get less than 60% kill rate, while the other will get more, or some will make more 3/4Ks, while others will make less. That's what an average means, by the way). But we can only do that once we have reached that target 60% killrate, and that is just what BHVR did.

    So, there you go. At this point, if you don't understand it, it's on you.

    Post edited by Batusalen on
  • Batusalen
    Batusalen Member Posts: 1,323

    The problem is that every average is made up of many numbers.

    No, again, the problem is that you don't know what an average is. Let's take 10 players in 10 matches, 5 of them do only 3Ks, the other 5 only 1Ks. The average kills per match is 2 survivors killed per match, which is a 50% kill rate per match on average, which again, means that killers draw their matches on average.

    That's it. Anything else is product of your imagination.

    You ignored all the math and did none yourself, and are jumping hoops of logic to suit your narrative even when faced with hard evidence to the contrary.

    Funny, because that's what you guys are doing. Either because lack of knowledge, or simply because like you said doesn't fit your narrative.

    And I'm done. At this point, you either understand it, or you simply not want to.

  • ratcoffee
    ratcoffee Member Posts: 1,475

    you can absolutely have a kill rate below 50% and a win rate above it. 60% of games ending in a 3k, and the remaining 40% ending in a 4 out - ends up being a 60% win rate, but only a 45% kill rate.

  • crogers271
    crogers271 Member Posts: 1,821
    edited August 19

    No, I'm using 'on average' because we are talking about averages.

    Average, math, is defined as: An average is the result that you get when you add two or more numbers together and divide the total by the number of numbers you added together.

    On average is defined as: If you say that something is true on average, you mean that it is generally true.

    Collins dictionary for definitions.

    Which do you mean because there is a giant difference between them (the difference between the mean and the mode)?

    It makes sense, but you either, again, don't understand what an average means, or you are simply refusing to accept what basic math and logic say. 

    I'm doing my best to be polite, but please consider for a moment, that multiple people are on here telling you that your 'basic' math is wrong, and no one is backing you up saying 'hey, I think I get what Batusalen is…'

    Look at this rationally: we're not trolling you. We think you're wrong. Multiple people have laid out the actual math on how.

    Now, let's forget you contradicted yourself for a moment, 

    I'm not, I'm pointing out a hypothetical math scenario to try and help you understand where your mistake is. You're selectively quoting.

    pretend to balance a game like DbD while granting everyone an equal chance of winning with just "uneven data sets" (which I'm going to guess are "kills distribution" in this case) when:

    None of this is relevant to the discussion. You can say that a 60% is good for any number of possible reasons. But the idea that it is a mathematical necessity for either parity in win rates or an MMR system is provably wrong.

    So, even if "uneven data sets" can help identify if an individual killer needs balancing, it doesn't work in terms of overall balancing and will not help balance killers against survivors.

    So I don't think you get what an uneven data set is. Its not something anyone is arguing the game should or should not have, its a fact that its what the game does have.

    As you are saying, you are doing basic math when you need to be doing slightly complicated math.

  • caipt
    caipt Unconfirmed, Member Posts: 689

    killrate =/= winrate. If the kill rate is 50%, then that means the killer (on average) kills 50% of the survivors in a match. If 50% of the survivors escape, that means the average match results in 2 survivor wins and a tie for the killer, meaning the killer getting a 3k/win is unlikely. With a 60% killrate, the average match still consists of 2 survivor escapes, but its a closer match that the killer could have won if they'd played better.

    Essentially a 53% killrate would mean killers are underperforming and a 65% (though not nearly to the extent described) are overperforming by a small margin.

  • crogers271
    crogers271 Member Posts: 1,821

    I'm not sure if you're backing up what I'm saying or misreading what I said because I think we agree (sorry, being extra careful in this thread at the moment).

    No, again, the problem is that you don't know what an average is. Let's take 10 players in 10 matches, 5 of them do only 3Ks, the other 5 only 1Ks. The average kills per match is 2 survivors killed per match, which is a 50% kill rate per match on average, which again, means that killers draw their matches on average.

    In that scenario you have a kill rate of 50% and a winrate of 50%. More specifically you have a winrate of 50%, a loss rate of 50%, and a draw rate of 0%.

    Also I don't know what would actually be the problem with that hypothetical scenario.

    You could also have a scenario of: 2 4ks, 3 3ks, 1 2k, 1 1k, and 3 0 ks (one of many random possibilities). That's still a kill rate of 50%, but you now have a winrate of 50%, a loss rate of 40%, and a draw rate of 10%. Unless you know the probability of the five different game outcomes (not even getting into hatch), you can't mathematically determine the kill rate percentage you need to determine what would create a winrate parity (not that a winrate parity is even necessary, but if that was your goal that's what you'd need).

  • ratcoffee
    ratcoffee Member Posts: 1,475
    edited August 19

    oh yeah i'm not trying to argue with you, i'm trying to back up what you're saying with extra evidence - kill rate and win rate only have the tiniest corrolation

    as far as Bathusalen goes, i've found he's overconfident in his knowledge on a lot of topics. he's aware of the process of calculating an average, but utterly unfamiliar with how to apply the concept of averages to anything. he takes people's rightly pointing out he's applying averages incorrectly, he assumes the person in question doesn't understand how averages work, not that he might have used the concept wrong. if I told him the standard deviation for kill rate across games was 38.3%, that is to say, +- 1.53 kills per game, i doubt he would know what to do with that information.

    calculations for the stdev of kill rate being 1.53, btw:

    S_comp is the component of the sum of square differences associated with all matches that result in an n-kill game. STDEV(kr) is for kill rate, (k/g) is for kills per game. Everything else is self explanatory i think

  • Batusalen
    Batusalen Member Posts: 1,323
    edited August 19

    I literally explained how those averages apply to the balancing of a total asymmetrical game like DbD, why it needs to be those averages, and indirectly explained how what you said doesn't contradict any of it, and in fact, it complements those kill distributions (those "uneven data sets" the game obviously has and no one ever argued it doesn't) you are trying to use to counter-argument it in terms of balancing, but I'm "utterly unfamiliar with how to apply the concept of averages to anything".

    Right. But it's me who is "jumping hoops of logic to suit my narrative".

  • crogers271
    crogers271 Member Posts: 1,821

    1.53 is interesting.

    On the data, I would not have guessed the 1k and 3k rates would be so close, that's neat.

    Hey, wasn't this thread about Trapper or something and you were asking for community thoughts?

    I don't think its too surprising. If you play all out/ruthless and are good with the killer, you are going to win most of your matches. Trapper especially can snowball, and if anything goes his way early in the game and you're willing to camp you can set up a pretty devastating scenario.

    I'd be curious over a 100 matches what his worst results were. Like were there games where he only got one or two hooks because he couldn't get the snowball rolling or he hit a SWF that counter played him from the start.

    It's not as eye opening as the no power trapper challenges Hens and others did a while back.

  • ratcoffee
    ratcoffee Member Posts: 1,475

    his breakdown of various different outcomes

    he said most of the 3k, 1 outs were because he failed to control both doors, only 2-3 were hatch escapes.

  • Moonras2
    Moonras2 Member Posts: 380
    edited August 19

    id be curious on this as well. There's been times I've thought I was going to lose on trapper just because I spent the first part of the match only setting traps. Sometimes I did lose lol but often times the traps would have a big part in starting a snowball as well.

  • Batusalen
    Batusalen Member Posts: 1,323
    edited August 19

    In that scenario you have a kill rate of 50% and a winrate of 50%. More specifically you have a winrate of 50%, a loss rate of 50%, and a draw rate of 0%.

    By the definition you posted:

    "Average, math, is defined as: An average is the result that you get when you add two or more numbers together and divide the total by the number of numbers you added together."

    So, let's do the math (with a total of 10 matches for simplicity reasons):

    • 10 matches = 40 survivors
    • 5 matches ended in 3K.
    • 5 matches ended in 1K.
    • 3 + 3 + 3 + 3 + 3 + 1 + 1 + 1 + 1 + 1 = 20 survivors killed in total.
    • 20 survivors killed in total / 10 matches = 2 survivors killer per match on average.
    • Now, using rule of 3: If 2 survivors where killed per match on average, and according to the win conditions a 2K is a draw, then killers draw their matches on average.

    Conclusion: Yes, we have a winrate of 50%, a loss rate of 50%, and a draw rate of 0%. Which, in short, means a draw rate of 100% on average.

    Unless you know the probability of the five different game outcomes (not even getting into hatch), you can't mathematically determine the kill rate percentage you need to determine what would create a winrate parity

    Can't I?

    • Survivors win condition: Not get killed. Ergo, a 60% killrate on average means 40% of survivors win their games on average.
    • Killer win condition: Get a 3K. Ergo, if a 60% killrate on average means 2.4 survivors killed per match on average, killers are doing mostly 2K, drawing those matches, while doing 3K 40% of the time approx. Therefore, they win 40% of their games on average.

    Now, let's try with higher kill rates:

    • Survivors win condition: Not get killed. Ergo, a 75% killrate on average means 30% of survivors win their games on average.
    • Killer win condition: Get a 3K. Ergo, if a 75% killrate on average means 3 survivors killed per match on average, killers are doing 3K 100% of the time. Therefore, they win 100% of their games on average.

    That won't do it. Let's try lower:

    • Survivors win condition: Not get killed. Ergo, a 50% killrate on average means 50% of survivors win their games on average.
    • Killer win condition: Get a 3K. Ergo, if a 50% killrate on average means 2 survivors killed per match on average, killers are doing 2K 100% of the time. Ergo, they win 0% of their games on average, as they draw 100% of their matches on average.

    Well, look at that. It looks like the only average kill rate that allows for average win rate parity is 60%. And I invite you to try to do it with any other kill rate you want.

    Is it clear now? Or shall we keep going with the fallacies and mental gymnastics? Not that I'm going to really keep with this, because at this point, there is definitely nothing else to explain. So, consider anything else you post answered with the BHVR go-to response: Whatever makes you happy.

    Post edited by Batusalen on
  • Batusalen
    Batusalen Member Posts: 1,323
    edited August 19

    you explained incorrectly. the fact that there is such high variance in the data set is a reason specifically to Not extrapolate based on the average of all those data, not a reason you Have to.

    Then again, explain to me how you would balance DbD, granting every player the same winning chance no matter their skill level or character played, with only those data sets when:

    • The win conditions for both sides are completely asymmetrical.
    • You can't directly compare skill levels of both sides, as they require completely different sets of skills and game sense.
    • One side is a team of 4 players that can have different skill levels.
    • That side is not even considered a team per win conditions, and each individual player in it will win or lose individually.
    • You have 5 possible match outcomes, which define the wins or losses of 5 different players.

    Come on, tell me exactly where I'm wrong and explain the correct way of doing it without taking in count anything else than those data sets, if I'm so wrong in my explanation of how and why.

    this is what I mean, you fundamentally don't understand why just the average kill rate is not a helpful stat for the kind of analysis you're trying to do

    Maybe because I never, here or in any of the past posts about this topic, argued that only having the kill rate is helpful to anything, and I'm not trying to do any "kind of analysis".

    What I have always said and argued is that "For balancing reasons, the kill rate should be 60%, and having it be that number is the only way the game will be balanced between killers and survivors" in response to everytime you people have claimed that it being 60% means the game is killer sided. And I have already explained in detail how and why it has to be that way multiple times by now.

    Just that. Like I said to @mizark3 earlier, anything else is a product of your imagination.

  • ratcoffee
    ratcoffee Member Posts: 1,475

    I simply wouldn't balance DBD with only kill rate.

    saying "you either have to use my way of balancing the game using only kill rate, or come up with a different way using only that stat" is a false dichotomy. I would simply opt to use stats with a higher resolution - such as the kill distribution stats on Nightlight.

    I assume when BHVR says they want to achieve a 60% kill rate, they mean that kill rate has a casual correlation with some other internal stats that they don't share with us, stats that contain more helpful information but for whatever reason (for example, trying to protect trade secrets or their own technologies) they don't feel it would be prudent to release to the public, and that in getting those numbers to what they think is fair and fun, they estimate that kill rate will hit 60%. I know Peanits has said they look at far more than just kill rate, which would seem to support that theory.

  • Batusalen
    Batusalen Member Posts: 1,323
    edited August 19

    I simply wouldn't balance DBD with only kill rate.

    Neither is BHVR, just as I explained:

    When we reach that point, now we can use the kill distribution of each killer to balance those that are overperforming or underperforming until we get a balanced average […] But we can only do that once we have reached that target 60% killrate, and that is just what BHVR did.

    But good dodge to avoid answering there.

    saying "you either have to use my way of balancing the game using only kill rate, or come up with a different way using only that stat" is a false dichotomy

    What is false is that I said to you "do it only with the kill rate" at any point, specially after explaining that not even BHVR uses it as the only data for doing it.

    In fact, you are the one who said that "there is such high variance in the data set is a reason specifically to Not extrapolate based on the average of all those data, not a reason you Have to". So, it is you who said that the average kill rate is not needed to do balancing at all, not the other way.

    I assume when BHVR says they want to achieve a 60% kill rate, they mean that kill rate has a casual correlation with some other internal stats that they don't share with us, stats that contain more helpful information but for whatever reason (for example, trying to protect trade secrets or their own technologies) they don't feel it would be prudent to release to the public

    Yeah, let's just simply forget all the reasons, including the math and dev statements, that explain why it should be a 60% kill rate on average, and let's start making unfounded assumptions just because you don't want to agree with it. Totally logical approach.

    and that in getting those numbers to what they think is fair and fun, they estimate that kill rate will hit 60%

    Totally agree, as to get any other number to a place where it is fair and fun, like having an equal winning chance to everyone no matter their skill level or character used, the kill rate should by 60%.

    I know Peanits has said they look at far more than just kill rate, which would seem to support that theory.

    Or, and hear me out in this one, he meant that they look at a lot of other data because even if the kill rate is helpful for overall balancing (among other things), it is complementary to other data that indicate other, more specific things. To give an example, even if a killer has a 60% kill rate, if his kill distribution has a lot of 3/4Ks, it will indicate that he is winning more than he should and needs to be nerfed. Again, just as I explained, which casually coincides with what he said in the quotes @Moonras2 shared.

    But hey, yet again, "whatever makes you happy".

    Post edited by Batusalen on