Bummer about the ban.
Yea... I love conversations that result in attempts towards improving the game. I try to discuss QOL changes all the time but many topics just don't get any attention by players. No bullshit here, the fact is that not every thing Riot does is for better quality of matches. They make changes for player perception or enjoyment some times that can hurt quality. Duoing is one such situation, I am pretty clear in my negative opinion on that topic.
There are some benefits to the league system, you can have a higher MMR gain and loss per match then a pure MMR system. This can be used to lower the grind which is funny as so many folks feel its for the opposite reason. Your MMR gains and losses directly impact how grindy it is. When and where K factor changes on a account also directly impact how grindy it is. It does not matter at all if it is the Leauge system or MMR Riot has a massive # of tools to make their game grindy AF if they wanted. Yet as functionally MMR expects the average players to never reach the top then its fair to say that the "grind", even in a well made well balanced and 100% fair system is literally forever. So I tend to dislike arguments related to "rigged system 4 more games" because its pretty fucking difficult to out grind eternity which is the expectation you have for average players in a fair and well made system. (Goes back to Arpad Elos system, your never a chess grandmaster if your unable to play the game well enough to beat people higher then you. this is exactly what you should expect in a system built around skill, not time invested.)
When i say time invested lets look at other systems which can be something simple. Like a leaderboard off # of wins. Skill in systems like this can be a low factor because some one who only wins 40% of their games but is able to have the time to put in several thousands of games can climb to the top, even above folks who consistently would destroy them. This is especially true when the matchmaker in said system does not look for any player skill when it makes a game, just pulls from who ever is around. So a meh player? They can be grossly higher on the ladder then some one who would be able to destroy them 100 games back to back simply because they have more free time to invest into the game. Where the fuck is "skill" in a "ranked" ladder like that? Garbage system imo… I honestly hate them its just a time sink.
Now back on topic for Riot and League.
As for the non-duo matchmaking, I don't know how limiting it to only one division ahead and behind you would influence anything beyond making sure that Silvers on a win-streak don't fight Diamonds, and Diamonds on a loss-streak don't fight Silvers.
Theres a few issues. First up is player populations. Higher up on the ladder we are looking at divisions that represent less then 1% of the games population. It gets lower also because that % does not account for active or inactive, long as the accounts put in enough games to make it out of placements and are in that range they are part of the ladders consideration till season ends (turns out there's a few players with multiple accounts in diamond as well which lowers the # of player by 1 for each account with the same human behind it.) At its worst, You will see issues with MM not being able to make a match if its unable to pull from a wider range when it does not have enough players queued to play. Historically we have seen MM in higher MMR become extremely long at upper tiers when there's low population.
Now to address how to avoid Silver win streaks vs diamonds the solution is about K factors. How much MMR do you gain and lose.
Shit... Right K factor... I have not explained that well yet this thread.
First lets go back to the OG Elo adjustment formula. This is what is used to change an accounts MMR after a game is ended. Its usually listed as some variation of this. _Rn= Ro + K(W-We). _Rn= Rating new, This is what your rating is after the game.. Ro= Rating old. What your rating was before the game (its what was used by MM to build said match.) K = uncertainty... the system expects to not be accurate when it does not have enough data on an account. By multiplying the MMR gains/loses by K variable you impact accounts speed when they are new to move up or down ladder biased of trends in match results. W is the result of the match. WE is the systems expectation for the result of a match baised on R0. ( People forget that some times it does not expect a fair game. Its rather common in chess. Needs to be noted that its very different for games like Lol. Its less about a rating gap in MMR and more about if there is a gap between uncertainty on teams for MMR and if the system did not have enough players in proper ranges to make a on paper fair 50/50 game. Plus what ever other situations the devs decided might impact a matches fairness and quality. I don't know riots, just sharing the basic theory behind it.)
MMR changes really should be game to game for its exact value off how fair the system thinks a match is.Thats why its a variable system then equal value changes like your suggestion was. Quick example, Players in placement gamesin many MMR and Elo systems actually change how much their teammates stand to gain or lose for a game. This helps protect folks from getting dicked over by newbies or boosted by smurfs.
To remove some confusion and show it in action I am going to pull some numbers out of my ass with no relation to anything. Don't treat them as accurate for jack shit, I just want to show how uncertainty impacts MMR changes its not gonna be accurate to any system riot or honestly what other devs use.
I will show 1 high K factor and 1 low k factor adjustment.
High. 1314= 1250 + 32(3-1) Won the game. Now 1314 is the value used by MM for next match range selection.
Low. 1270=1250 + 10(3-1) Same as above only 1270 is used.
Alone this is not that dramatic. That is good, you don't want a single match to be some Massive MMR value. But what about a 10 games rather then just 1?
High would be 1872. (this realistically is probably WAY to high)
Low would be 1450.
Now your MM system said that it does not want to make a game with a 200 MMR gap because it feels that represents a very unfair range then Low and High are no longer going to be in the same games barring some extra features like say high range duoing. Not kidding when I say I actively dislike its implementation.. fucking 60-70% of player base ranges are gross.... but only impacts "average" players so its fine long as they are all having fun!
Uncertainty/K factor in most systems gets lower primarily for 2 reasons. Reason 1. an account has played X amount of games. You have put in time the system starts to feel it has enough data points on you to become a bit more accurate about how well you play Vs others in the ranges it used for you. 2. You made it to a benchmark MMR value as You don't want players to climb off luck. I am going off Elos work for this, cuz its documented clearly and he explained why. Grain of salt here as it might be more or different things Riot uses.
Also win streaks as a topic on boards are just freaking idiotic... most folks treat 3-4 games as a "streak" in a system riot claimed needed 150-300 matches. Its just not statistically significant.
As you said, Silver players win streak should not be vs Diamond players loss streak. (Small # of games should never result in massive MMR changes. The system needs to value accuracy so it should be made so that it takes a reasonable sample size of games. Riots last official number I recall seeing was it on average worked off 150-300 matches)
Boards concept of a "streak" is stupid. Reality, Its more about win %. Its about a MMR balance. Like a checking account. Wins add losses deduct. A streak of 15 wins is great... but uh.... if you had say 25 losses prior then your looking at a net loss for your check book. K lowers with games so that 15w streak might not even balance out prior 15L.
Mismatches happen from stuff like soft resets and the fact its very easy to lose MMR (just throw games.) should you want to play vs lower skill players. MMR done well prevents bad players from climbing high off luck over skill... which is exactly what you want in a skill biased competitive ladder.
Riots League system will play catchup for Divisions to assigned MMR. Your MMR gap gets to big? skip divisions. Remember how I was talking about features for player perceptions? Riot made it a bit harder to drop out of Tiers. You have to make it an entire tier below in MMR to get demoted last time i checked + on top of that they also give you a warning your about to get demoted... Apparently players get upset when they demote tiers.
Do you see why Tiers don't do a good job of representing uncertainty? The moment your out of Placement matches you are in a tier. Pulling exclusively from that Tier exclusively has issues with Population size, MMR gap between tier as MMR moves faster then LP. You get some weird interactions between uncertainty and tiers and have to remember that as a developer your MM system will need to still be allowed build a game at the end of the day for any users to actually play a match.
The advantage of Riots LP system is you can have a more dramatic MMR change and higher K values in a team game then a purist MMR system. It also hides that you honestly expect 50% or so of all users to lose MMR from the starting value they had.
Downsides are fucking duo is done bad. Yes... It wont break the game but man it makes mismatches and MM has to deal with shitty range gaps. Other downsides are things like people are distrustful of the system or the fact is that most everybody thinks they are better then they actually are. Trend is People treat their "Good" games like they are their "average" performance.
MMR is meant to represent your Mean average between on and off matches.
MMR when its accurate Is supposed to represent a range of expectations, on or off, Rather then some absolute value.
With a constantly changing meta and game, MMR is also meant to self correct and move players up or down at any and all times depending on performance.
