The Amanda's Letter add-on for The Pig has been Kill Switched due to an issue with incorrect RBT count.
http://dbd.game/killswitch
What would Useful Data look like?
I think I'll be running a couple of experiments on what limited data I can, which got me thinking a bit harder about the problematic collection and presentation of yesteryear regarding trial data. I really wish devstream data were more complete in how it presented data, and BHVR were more willing to acknowledge the differences between new and veteran play.
What do you, fellow killers, survivors, and people who've exited the Fog but are still lurking, think good data sets would be, even if collecting that data might be slightly implausible? Here's a couple of my own thoughts on what I'd like to see from BHVR:
- How many hooks, kills, and 'escaped chases' per match, per "rank/MMR"[tier]?
- Which perks, items, addons, and offerings were brought per match, including "N/A", per tier?
- Which survivors and killers were chosen proportionately per match, per tier?
- What maps/killers saw the most kills per match, per tier?
- What maps saw each killer receive the most/least average kills per match, per tier?
- How many average pips were gained per survivor/killer per match, per tier?
Think any of these could be refined, or have any of your own?
Comments
-
Quick clarifying question- useful for what purpose? Just general-purpose data collection for monitoring the state of the game from a developer perspective, or for something more specific?
1 -
Useful according to a game-centered metric of your choice. A couple of my own (survivor pick and offerings per match) are fairly superficial as far as game design goes, unless the other ones got more concrete stuff to pair with that info - but I think they'd be interesting as far as "what do people want to play" queries. The others are more focused on balance - but I'm genuinely interested in pretty much any direction others would want to see represented in some idyllic dev data dump.
0 -
All the data you mentioned is interesting, I don't know how you can actually collect though just on your own but have at it!
Personally the main stat I'm interested in is kill rates by killer specifically for matches where all players have high MMRs. Unfortunately you can't view MMRs, so the best you can do is take stats where all players are high Grades, which isn't quite as good since Grades aren't actually used for matchmaking and are more of a measure of how often someone has played in a month versus how well they've played that month. Hypothetically, though, in games where everybody is the same MMR the median results should be 1-3 kills supposedly with 0 and 4 kills the least common and an average around 2 kills. (When the devs released some kill rates for all matches across all ratings the average was something like 2.15 or about 53-54%, so not too far off on the overall. The question though is how do the numbers look at the high end?)
Being able to get similar stats by map or filtered for certain perks, etc, is interesting too. I'm not sure how data one person can actually get just from their own matches, though. The best you might be able to hope for is to look at your own personal kill rates as a killer over a long string of matches and see if you're actually getting matches in that 1-3 kill range most of the time.
0 -
The problem is that even with 100, 1000 or even 5000 games, your data is "useless". Don't get me wrong here, i would love to see useful statistics for this game, but useful is the key-word. With 30.000 players daily (let's just go with that number) we would need the amount of survivors, their MMR, play-time and if they actually want to win.
We have no way of focusing on any control group or to use any data in that matter. If we say "oh look we have 80% DH users" it says nothing if we don't know if they are just taking it because everyone says it's broken. These data sets are also useless when we need to consider "what happens if we turn off crossplay". Do Pc gamers change the meta? Or is the mix of console and pc giving us any different data.
We also would need hard-data, that means we need to include data that is confirmed by every player ever. To my knowledge they have no real tool for that on their end. As the MMR is very fresh we would need to wait 2-3 more months to use the data, as such a fresh change means that everyone needs to get to their MMR level and we need to see how that changes first.
At the moment all this data we have "is ___ op" or anything like that doesn't say ANYTHING about the game as new or changed perks falsify such data sets. We would need the data for the old perks and the new perks to see how much they changed. And then we would need to have data of at least 50% of ALL games played (from survivor and killer side).
End to bring it to an end: The data will help noone, even if you find a build where the survivor gets an escape every single time, as every game has so many viables (let's say you have bad game, or they throw) it only gives you one sided view. Or imagine you die every round because you are just not that good yet, what does it say about the game?
Most data we have comes from high MMR or "hear-saying" about this game. And even verified data from the devs just tells so much.
0 -
Since there is a killer that ignores all the basic functions of the game, it makes data collection nearly meaningless. This doesn't mean nurse doesnt have a counter, but because there is so much she can ignore it makes balancing really impossible.
0 -
@dugman & @lavars : I'll not be able to collect all that on the whole of DBD, but I'm not sure how to more eloquently word "Dangit I'm a curious person, think the dev team's previous data is incomplete, and my own experiementations won't amount to much: I wonder what other people are curious about and wish BHVR could shed light on" - I'll try to edit the original post to reflect that better, though.
I'm typing under the pretense that BHVR would be the ones releasing the hypothetical data dump, so they'd have access to MMR and other relevant bits. If I wanted to do that I'd have to enlist a hacker since they can view MMR, which I'm not quite willing to do.
0 -
The data IS incomplete, but like i said such data-sets need way more info than you can give. If you are a very good player and only play super high MMR, you have no control group, you would need the reason WHY someone takes certain perks, for example: what if windows of oppurtunity would be the most used perk ever if it would be a normal perk? many people don't evne have it unlocked.
same with other perks on killers, you would need to unlock every perk for everyone and then run the data in every "elo" and with every player, plus against the data what happens if certain perks would be blocked off. What are alternatives and what makes sense. we all know that two exhaustion perks don't make sense, but running "lithe" and "balanced landing" could be useful in 1 out of 100 games. We would need to set up every natural combination against every synthetic combination, that means "what is acutally the real data in every MMR area, and what is the best theoretical scenario"
0 -
Please reread my first response if this doesn't clarify enough: I'm saying "what would you like BHVR to release?"
I am in no way saying "What should I, personally, try to gather data on?"
0 -
Same, we don't have the data in the first place. And my answer is the same. They would need data in the first place, to know what data is important. We don't have this data, and they only have a very tiny spectrum of data that is useable. they prooved it more than enough.
at the moment we rely too much on what content creator say or what we experience. From a pure data-point, we have zero information about perks and any data that would be useful.
We would need data in the first place to know what is useful data. And we need to know that this data is reliable "killers have more than 50% winrate" says literally nothing if in 2/10 games someone trolls or disconnects. Or if people are new to the game, they got lucky etc.
If YOU get the data or BHVR doesn't matter, they need the same amount and the same data. And to be honest do i think they don't have it.
0 -
You not understanding the point or intent of the thread is no longer my problem - have a good evening, morning, night, or day.
0 -
No, i used bad phrasing, sorry for that, let me explain:
To know what would be useable data for them, for us, for the game, for... whoever... We would need data we don't have right now. The most informations or data we have is from what others say. And even this is falsify data.
as long as we don't have data, there is no answer to begin with.
0 -
To reiterate and paraphrase something in the first post: "what would you like to see as far as data goes, even if the ability to gain that data weren't practical or available?"
It doesn't matter what we currently have access to, because the question is about what you'd like access to and what you'd do with it.
0 -
Basically everything, literally raw data.
but like i said, this data is useless, even if we know what perks are used in high MMR and which gives the most escapes or kills, it useless data.
I played a game called "GTFO" even was mod and tech-support for them, and i saw raw data, which was useless as we didn't had "finish a level per person" and "escape per level".
Let me give an example. Imagine we play with two friends, we all take dead hard, ds, bt and lithe, doesn't really matter what perks...
and you loop him for 5 minutes, we get all the gens and we all escape. This data is useless if we want to know which perks are useful. We would need a ground-data-set. What that means is basically a control-group. But that only says so much and from their we could work our way up to know WHICH data actually is useable.
EVERY data-set that has a use, needs to have an already existing data-set we can somewhat rely on. And with changing perks, maps and killers it's not useable data, no matter what data we have. The amount of data we WOULD need includes stuff that is impossible to get.
0 -
In that case, it would seem your answer would be data which is impossible to get - which falls within the parameters of my prompt. What would some of the impossible data be?
0 -
As I mentioned in another thread is that BHVR have said (and demonstrated) that:
- They 'don't balance around high levels of play'.
- They'll nerf a killer (Wraith) because he was too easy for new players to do well with, and too hard for new players to deal with, despite him being fine at high levels of play.
- They won't, however, buff a killer (Nurse) that is seemingly impossible for new players to do well with, and is seemingly very easy for new players to deal with (hence her hilarious kill rate in that dataset), probably because she is strong at high levels of play.
3 demonstrates a pretty crucial inconsistency. BHVR only care about the new survivor experience. I wish I was being hyperbolic - but that's what these statements and design decisions boil down to.
By their own logic and established precedent, either Nurse needs a buff or Wraith needs his nerf reverted.
As to useful data, I'd want to see a graph with killer kill rates on one axis and MMR on the other axis.
1 -
Agreed.
0 -
We don't know what useful data looks like, because DbD players tend to take any data and throw it out saying it's bad
0 -
You would need to get playtime against MMR and what perks are used and WHY.
you would need data on how many are SWF and how much that effected that data, we would need data if a voice chat would change any of that or not, and how much that would change data.
we would need exact data on how much the spawn point of a survivor effects the game and which perks, and then the exact scene with different perks, then the exact game with a different spawn.
These data sets would be so enourmous they would need to consider EVERY little thing to say anything about balancing. We would need every position at every moment and every info that is that exact moment but with different perks and then with every killer with every perk. It's impossible, except had some kind of ai-simulated.
everything else is literally saying nothing except "oh most people escape with Dead hard" and that only says that most people who escaped had dead hard. And even if we have info about MMR and the exact escape/kills we don't have control-data to verify if they are accurate.
0 -
Unfortunately, the only data we've been given is bad.
It's too context-less and generalized to be useful, but still becomes fodder for the endless 'this game is killer sided because kill rates' arguments.
If BHVR gave us a proper graph, showing kill rates vs MMR on killers, and possibly include addons too (if community driven sites can do this and more for infinitely more complex games like Path of Exile, it can't be that hard to do) now that would be good data that would be very difficult to discard. You could point to something like All-Seeing and say 'well, that addon is clearly a problem, but the killer is actually fine'.
1 -
How long Ruin lasts
Times of first Gen is done
How long the first chase is (including the first hit then down)
1 -
Ever since the start of SBMM I have been keeping data on my games as killer and survivor.
As Killer I keep:
Killer Name, Match #, Map, My Killer Perks, Survivor Names, Survivor Perks, Outcome of the Match in Kills/Escape/Hatch, and random observations of what happened.
As Survivor I keep:
Match #, My Perk Loadout, Map, Killer, Killer Perks, Outcome of the Match in Kills/Espcae/Hatch, and observations.
I plan on plugging all of these numbers in a spreadsheet to be able to examine my experience in data form.
1 -
That's pretty dang cool!
1
