Which percentile is considered high MMR by BHVR?
I just read the latest developer update, and this very important information is missing, despite being mentioned several times throughout the text.
Example of answer: "TOP 3%".
Comments
-
I'd guess they mean everyone that is close to or at the internal softcap. I obviously can't tell how many percent of the playerbase this is, but i guess it's more than your example because you gain mmr passively and you will hit it guaranteed after some time.
0 -
I don't have the exact percentage to hand but it's a small percentage at the high MMR level - it's above the soft cap.
The bigger areas are at low MMR and at mid MMR - high MMR is really a very small percentage of players
10 -
Can we know what percentage of the players are at the soft cap?
Due to the fact that everyone at (or above) the soft cap is treated exactly the same for matchmaking, it’s the highest MMR bracket for public games.
9 -
If their system is anything close to elo (in terms of statistics), based on where the soft cap is now, its likely <1% of players.
2 -
Mandy, why doesn‘t the Feedback of the Onryo post appear on top of the other discussions, after someone commented something. It is also stuck with the date of yesterday.
Post edited by jonifire on0 -
how would you even calculate that?
0 -
So Elo, which was invented for chess is the rating system most things use for MMR as well. Usually just slightly modified.
Based on how Elo works, there is a statistical percentage of people that will be in certain ranges. So if they haven't modified it too much. And given that last I heard the soft cap was either 1900 or 2100, I'd guess that it's less than 1% of players.
1 -
Last I heard the tippy topiest of the topest was only like 3-4% of the overall playerbase. Not a big portion of the whole at all.
0 -
Hadn’t really thought of that but it’s extremely relevant. Like I would expect top 20% for example to give a very different result than top 5%.
1 -
ok... but how would you calculate that?
0 -
I'm not sure which part of it you are asking about:
If it is how does an ELO calculate: there would be an algorithm with multiple variables. Every player starts at the same point, you win you game MMR, lose MMR goes away. This though is modified by the MMR of players you are facing (DbD further modifies it by game length and, if a survivor, team performance). How much weight is attached to each variable is unknown.
How that high MMR is less than 1%: BHVR has said that the MMR is on a bell curve (like most competitive activities). Given enough games a very small group of people will reach the top.
0 -
I'm asking how anyone calculates how many people or % of people are at any elo range.
0 -
For BHVR it would be simple they'd just have to look at the numbers and draw a line based on where the highest 1% (or whatever number they want) and call that high MMR. That could be 2200, 2300, 2400, etc.
They also likely base where they set the brackets / caps at where various MMR points have ended up.
You could predict where the upper MMR would be when designing the system, but that would require knowing the player pool size and the theoretical upper skill limits of the game.
1 -
For chess, they just take the full list of official members (they all have an elo score), and they just count how many people are in each elo range.
0 -
so
"If their system is anything close to elo (in terms of statistics), based on where the soft cap is now, its likely <1% of players."
has absolutely no base beyond at most an educated guess?
1 -
http://www.uschess.org/archive/ratings/ratedist.php
The USCF released the data a long time ago (it’s in the above link).
I think it’s assumed the current distribution is still roughly the same.
0 -
has absolutely no base beyond at most an educated guess?Its a little more than an educated guess as elo systems should provide a bell curve if the game has a skill element. If it has skill you will either get to a point where all of the top group of people are all of equal skill and/or they are small enough of a number that they don't get matched against each other.
2 -
why would it be? we have no idea what's the distribution on dbd's elo system.
until we get a graph or something showing it, I don't really see a reason to assume it's gonna be similar to chess
so, yes? nothing you said gives a number, just an expectation. but no way of knowing at what number we should expect it.
0 -
I never said DBD worked the same as chess. I was answering a question, on why we thought a certain percentage of people were in a specific elo bracket in chess.
DBD likely is similar, in the sense that the DBD MMR distribution is based on a formula, instead of having predetermined percentages of players that must be in each MMR bracket.
But we also know DBD’s MMR distribution is way less stable than chess’ elo distribution, because BHVR admitted to moving the soft cap semi recently. This implies that the soft cap itself is based on a predetermined percentage of players.
The actual important questions are “what percentage of players are at the soft cap”?, and “what percentage of players are considered high MMR”?
1 -
I would agree.
I believe it was hens that recently made a video where the soft cap was data mined to be 2100. which honestly means nothing on a practical level because we can't see our mmr...
but that can change on any update, and high mmr is such a nebulous term... top 10%, 5%, 1%, 0.1% of players... would all be honestly perfectly valid answers. or "everyone above X elo".
1 -
so, yes? nothing you said gives a number, just an expectation. but no way of knowing at what number we should expect it.Sure, without us actually having access to the player pool size and specific elo mechanics, saying something like 'high MMR is 2050' is just an educated guess.
0 -
Interesting that BHVR likes to keep things black box. It could be 10%. It could be 5%. It could be 1%. Mandy said it's above the soft cap, so it could be "top 25%", as we can divide: 25% lowest mmr, 50% middle mmr and 25% top mmr.
The results between 25%, 10% and 1% are very different from each other and keeping things black box is misleading players.
0 -
To be fair, they're probably under some kind of ruleset of what they can and can't say, lest they accidentally start a riot
1 -
I'm sure that we are being led to believe that top mmr is <5% when instead is top 25% (fourth quartile).
1
