You basically just said that you gain or lose it by winning or losing. Wininng and losing are just the end result of all the actions by both teams. That is not a representation of ANYTHING.
Its a representations of everything! You just said it yourself, Wins and losses are the end of All actions taken. What is skill in the context of a game if not your ability to take actions? We want to sort skill? Then positive skill is your ability to take actions that result in a win. Negative skill is your actions that result in a loss. The context of who your playing is important because Skill is not equal or universal between players. Rating players is the act of sorting skill gaps.
Thats why You only gain/lose MMR by losing and winning. That's how its generated. You accidently highlighted the reason why, which I bolded.
Arpad Elo's system has a few basic assumptions. I will be paraphrasing some relevant ones below.
- In games of skill you see skill impact ability to win or its just a game of chance.
- Skill is best measured objectively by wins and losses given that skill impacts your ability to win.
- Who is playing who matters if you want to rate skill, because subjective judgment becomes difficult if not impossible when the gap between skill is to large.
- Humans are inconsistent.
So.
"If the mmr system worked then you would not have games be so lopsided so often."
Not really. Its not looking at actual gameplay which is why this kind of system can be applied to any game of "skill" Where ability impacts outcome. But the in game experience? that's not covered by an objective system. If games are snowball from mismatched teams, or If games are snowball because the actual gameplay is snowball, they look the same. All it cares is if Ability impacts wins, its uninvolved with the actual gameplay. If gameplay itself just generates stomp games then that's just the games gameplay and not an issue for a MMR system or an issue because of MMR.
Lots of competitive games besides league are stomps when momentum goes one direction even at close levels of skill.
In my opinion the average player is garbage at comeback mechanics and there is a very high % of idiots in a majority of the games matches who just give up the moment they have a weak score. Which is stupid but hey its their raitings that get impacted most by this. end of the day i am gonna play to win as much as possible off my own performances and aim to make impact always. Knowing how to play from behind is a good way to improve your rank.
I call these players idiots because they are always overvaluing themselves. I admit there's better players then me. I also know its a team game. If you cant fill in the context of how you should play to win in the situation your vs some one stronger in their role then you deserve to lose games by pretending that you magically can beat them solo.
You can have games with people who belong 5-7 tiers above where they are but don't have the skills to carry people so their mmr is not where it belongs
This violates the principle that skill impacts outcome. You can have some one 5-6 Tiers below where they belong in Elo and MMR, because their rating is not yet accurate. The expectation is that once they have provided enough data points from playing games they will make it to that Tier. Core part of a self correcting system is it moves you to where you belong off the total sum of your performance.
What you cant have is someone 5-7 Tiers below not be able to climb to the correct range of expectations. The entire system is made with probability expectations which are used to build and adjust ratings. If its not showing up in your consistant ability it does not exist. Here is another opinion of mine. If you do not win games off your skill then its not a game of skill. If its a game of luck you cant claim that bad teammates impact your ability to win. Because luck is not going to allow for good or bad to impact outcomes.
To claim some one deserves a rating or rank they have not proven they can achieve is biased. Simply put, you saying some one can be 7 tiers higher then their rating but its impossible to make it there would be exactly the same as if I was to tell you I am the worlds greatest's player of league of legends ever. That I Deserve to be challenger and position # 1 but my teammates just stop me from making it there. If you want to ignore past data, and make an assumption for skill that's separate from what's been shown this is a fair claim.
Realistically, We both know its untrue because any one with a basic level of skill and understanding of this game can look at how I play and check my results to show this is not the case. If proof of ability is not required to rate or make a claim for ability this statement is completely fair for me to make because its just as inherently valid as any claim that X deserves Y skill with 0 proof.
The concept is class intervals, when a gap of skill gets to high between the average it shows up directly in win%. When a gap of skill is so low it cripples teammates it shows up in win %. Outclassing the average, being out classed by the average. There is not an infinite # of people so by default there is an average level of skill.
This is why generic systems like rank v rank work better.
Ranks are made from MMR ranges for them to represent an expectation of skill. Population size systems exist and are less accurate then MMR. So using Rank for MM when its off MMR ranges is less accuracy. Its treating each player as having a wider range of MMR and increases the # of skill gaps MM creates.
Also a system that starts all players at the bottom is bad for league because your bottom tier would be your majority % with no place to sink for the weaker players. Your taking the range of skill from iron to silver and smashing it together and thinking yea that's gonna make for some fair fun games?
You also will note that these Generic rank X system tend to use a points gain and loss. Most of always that's built off Elo theory.
Being average is rough if that's your skill because bad and good players have a high impact on your games. But.. making the average represented by everybody under the 50 percentile marker is shitty for quality MM
And removeing any type of raiting then making games 100% RNG full spectrum of skill is also shitty for quality MM.
Riot can do better. Lets be clear. Riot can do better. But not using an elo style system is terrible for any team game. You have more options for 1v1 ladders. This is the best solution for the 5v5 with RNG teamamtes. If you had a Rooster, set teammates? Then you could do a 1v1 ladder system becuase there is a constant for all teams. There is not, so theres no Zero sum system that's going to withstand the pressure league MM will throw at it.
Because hey, its math. You can test these systems by checking the base probability expectations for accounts to find out your # of games for accuracy. You can then build and average # for system accuracy. Then you can check accounts at that # and graph them out for distribution models and Literally SEE if its functioning properly and that your MM has rules that are functioning properly for building teams.
There will be noise but your deviation from expectations should be a functional Range for fairness between skill when accurate and the system adjusting when inaccurate is how accounts climb or sink. Self correcting is awesome like that.
