Would ranked actually be better if you where placed with people based on your rank rather than your MMR?
wouldnt that mean people would be put into there respective ranks and ether stay or climb from there or would that be to easy to abuse?
DODGEORLOSE1 votesThat would bring up the issue of lp gains/losses. Although a simple solution would be to start each player out at 20/20 and increase/decrease by 1 every game until it caps at 30 something.
Captainn Ginyu1 votesThat would bring up the issue of lp gains/losses. Although a simple solution would be to start each player out at 20/20 and increase/decrease by 1 every game until it caps at 30 something.
i imagine taht would also bring up promotionals as well
KingVKB121 votesI think ranks are just the covers for MMR. like Something to look at.
Kai Guy1 votesThey do become easy short hand way to convey skill once a player is well established to belong in that rank. Up to that point they are garbage distractions people get far to worried about.
Also they screw up MM quality because of Duoing restrictions are off Ranks not MMR values. Wont break the system but I don't like it one bit.
Kai Guy1 votesThe short answer is No its not a positive change. But its good that your curious and care about the player experience.
Longer answer. To try and address this you need to keep in mind that because Tiers and Rank are instantly applied they may not be remotely accurate with low sample size of data. If you treat every player the same certainty, that Is to say everybody in a tier is treated as identically skilled, you create issues when the system has low data on a player. A player with 500 games in silver 3 looks the same as a player with 2 for a system like this. MMR is about lowering uncertainty on accounts and trying to have MM build tight enough match ranges off each accounts results from every single game on the ladder. Its never 100%... and they are made to account for this. Rn=Ro+K(W-We) or some variation of this is the adjustment formula your going to run into if you google the topic.
A visual to help see what MMR MM aims to do. https://imgur.com/a/4SFSAf9. I think everything I list in that is accurate but encourage you to check into it yourself because I goof up sometimes. This is just a hobby of mine not a profession >.>
The reason you need to care about accuracy is If the assigned values are not accurate you see an increase of Fig 3 being produced by the MM system. The lower accuracy the more inaccurate results are produce, this makes a pretty vicious cycle. Its caused because just using a Tier which is instantly assigned means everybody is assumed to be correctly placed, the result is MM does not put as much effort into trying to control ranges. Its Range of "Fair" Is strictly locked to the Tiers.
Each Tier is made not to be some absolute value of skill. Just because some one is g1 does not inherently make them as good as some one else also in g1. You expect a new account to start under gold, and if they are plat level of skill the expectation is they will be able to make it to that MMR. As that account is making this climb they move thru multiple Tiers. This is something you can see from looking at Player % distribution, Not every single Tier looks to represent the same MMR ranges. Tiers are mostly to feel good, to get a since of accomplishment. To motivate people. The Victorious skin line is a carrot for the majority of the player base to strive for as a benchmark. But they are wildly inaccurate until an account has put in a high # of games, Its out dated information but the last time I saw a Red use a number it was 150-300 matches. Soft rests should make that a little bit lower but that might have been accounted for by the red for that range.
So MM Quality tends to drop down in MMR systems at ranges where players can intentionally get to by lowering their "skill" (throw games to drop down ladder or make a smurf to be treated as a newbie by the system.) and go up as you move farther and farther up the ladder. That's just on paper thou, People suck and are super emotional so that's not always the case, high MMR has its fair share of Toxic players as well. But... well MMR is not about player behaviors that don't impact win or loss. So there's no solution to MM issues from Toxic playerbase. You need Player Behavior systems to run that show. End result in a well made MMR system is the weaker players sink, the better ones rise. Average ones float around in their range because if they get misplaced they become fig 3. Get to low and your imapact on a match is consistently the highest. Get to high and your holding your team back. This only gets exaggerated the bigger the gaps are.
Now... Riots systems workable but I don't love everything the do for theirs. Conversation for a different day however. I get carried away on topics like this. So I think this should be a decent start. Let me know if your curious about Elo or MMR systems. Or just check my discussion history and look for one of my many walls on the topic.
Quepha1 votesAbsolutely not. MMR is designed to create even matches while ranked is designed to be flashy nonsense. Uneven games are miserable so matches should be made to try and avoid them.
Kai Guy1 votesHey! What up Quepha.
Quepha1 votesHello
Kai Guy1 votesBeen 3 days now and i am still trying to understand _We _ variable calculations lol. god dam variable math. Still mired in Normal Distribution models. Have not made it up to Logistical Distribution yet.
The good news is the formulas are starting to look like English which is good but when I turn it into formulas again to apply numerical values I keep making mistakes. ><
I intend to still get back to you when I have a proper grasp on the concept but I don't feel like i am there yet.
The update, is what I see so far strongly supports what you said to me. That in a 1v1 model with Normal Distribution it does not look like the value can be anything worse then 0. out of the W-We Calculation.
So unless there an additional modifier variable to the (W-We) Or something that allows for the creation of a -We value 0% odds of ever seeing it happen. Even if you have things that would result in this, Its feels like a pretty easy situation to avoid as a dev by just using an Assumed Ro value for the outlier accounts if it would impact your systems We. I still don't get the math yet that would be used to make that a reality.
It does look like I overvalued information about negative MMR gains from a win in a Dota 2 Valve post as accurate for multiple MMR systems. So thank you for the correction.
I am not sold slanted LP gains are going to make any impact on anyone's MMR thou. Only seen the Riot Sapmagic post on the topic personally so share a link if you got more info on what Riots system is currently doing.
Or don't, its not like you have to care or help me lol. I'm interested in the topic and intend to keep learning.
Captainn Ginyu1 votesAbsolutely not. MMR is designed to create even matches while ranked is designed to be flashy nonsense. Uneven games are miserable so matches should be made to try and avoid them.
idk i often feel like these days im not being put against people of my own mmr
Quepha1 votesSorry you had some bad games but that doesn't really mean much.