MMR requires a good # of datapoints, usually 100+ minimum for it to have an account calibrated properly. A weaker MMR system can take as many as 400-600, Riot's last claim was something like 150-300 for theirs. Microsoft claims they can achieve accuracy for teams of 8 players in 91-273.
End result, a Streak of 3 is insanely common even at 50% odds, something you can easily see for yourself looking at 100 coins
So basically everybody who uses 3 as a "streak" or thinks 3 games has a large relevance have unrealistic expectations. Any game after, regardless of the # of games in the streak "ends" it soon as it shifts. Also MMR is a balance so streaks are worth basically nothing apart from K factor and We Variable impact. WWWL = WLWW. (15+15+15-15)=(15-15+15+15) 30=30. Or if you want it represented by variables (A+A+A-B)=(A-B+A+A). 2A=2A
Folks think Hey! I lost, I had bad teams, my streak (witch amounts to no real significance even if your at "Fair" odds) was ruined because of teammates. That's not fun. I blame teammates why did riot give them to me.
but... how often do you think, Hey I lost because every player on my team was better? By definition when a team has weaker preforming players you would expect that team to lose. Vice versa for stronger accounts.
I know ima get downvoted, folks will make all kinds of fanboy riot lover comments etc. But they are idiots... I am so tired of the mentality I see in this game generated by this roll over,give up, its rigged, I cant make impact, bla bla bla, move on to next. I try to correct this imbecilic opinion because these stupid son of a guns KEEP GOING INTO QUE PRE TILTED. One lane does badly.. so folks roll over and ensure its a loss. No accountability in impact, people honestly expect that the only way they can win is if all 4 of their teammates outperform their counterparts? You think that's fair? What kind of stupid fucking logic is that? (Ranting. This is not intended to be read as if its directed to you Panda. Just the idiots who don't do a single google search of research and decided they know everything.)
Look... The concept is class intervals. In MM a Class range means that when properly assigned The worst player in a match could beat the best player in a match. Their good games are better then the weak games. You don't give them fair odds, Example. in Chess A class interval ( C) = 200 E (Elo points) Depending on the Probability Distribution Model used there is some small variance but its basically 75% to 25% odds AKA 3 to 1.
Riot Might have a different expectation for their class intervals, I don't know their values nor have I ever seen it provided. My Expectation is they use Verhulst and thus are running a Logistical Distribution Model for the game, which would usually set a an interval of C to be 76% to 24%. My secondary assumption for Riot is that each "Title" Represents a range of C. Iron, Bronze, Silver, Etc etc. They actual Point value does not mater. C could be 5. could be 1000. Its an arbitrary value that has no meaning outside of the system. On a scale of 8 out of 10 8 is big, but 8 out of 1000 8 is small.
Realistically, when your in your proper MMR range. Yes, the game is pretty much luck of the draw. You represent an average level of skill so weak allies or strong enemies make for a larger impact on the result of the match then your individual contribution. This is frustrating… I have been stuck in every MMR from Bronze To Plat. The odds thou, they always favor a consistently good player.
"Its Rigged" MMR is made from Match results so the data used to rig games would be destroyed because that data was generated improperly.
You cant use MMR to Rig MMR.
"everybody should start in Iron" Ok... So iron just becomes the new silver, at no point do bad accounts look to be weaker then the average player, climbing just got insanely harder for most players because your class intervals are built to assign titles only for the above average players. Raiting is just probability expectations and how well your account matches them. Back to chess for examples, 1200 v 1200 = 50%, 1 to 1 odds 1400 V 1200 = 76%, 3 to 1 odds 1600 V 1200 = 91% 9 to 1 odds. Call 1400 silver? Start folks there? Ok that's the average for the game. A majority of players will stay there because a majority will represent 50%. By having it set to the start point you have made that the Mode. So when you use a Mean average (literally MMR) to represent Accounts the Mode for the Mean is the starting point. Make 1200 The Mode? Bronze is now your majority of accounts.
Now its not a Zero sum system so There will be potential for inflation or deflation depending on # of transient accounts but you solve that pretty easy with a soft reset removing inactive accounts and redistributing the player base closer to the value you want representing the Mode.
The game needs better behavior systems, to many toxic babies who give up the moment they don't do well or they see a weaker teammate and just compound the difficulty in turning a match around. Wont play for a win? Stay the fuck out of ranked.
Now for some MM issues riot does have in my opinion.
Demotion from a Rank is insane. The range is gross. there is a reason accounts pool into Division 4s. They don't move them down so folks don't feel bad and quit playing ranked. Does not impact MM but it affects players opinions. OMG WHYS GOLD IN MY SILVER GAMEZ! FUUU RIIITOOOOO! Opinions affect behavior.
Titles (Rank + Division) are given instantly, even at high system uncertainty. A min # of games is needed to expect accuracy, That should be reflected in some ones rank.
Duoing implementation is stupid. Range is off titles? Not MMR? Fuck that noise. Silver 4 can hit bronze MMR and Gold 1 Can hit Plat but the two accounts can duo and MM has to compensate for something like 80%+ of the entire player base worth of skill? Good luck... population size in any given day for any given que is gonna produce some stupidly one sided games.