Yes you get an inter or afk here and there back in the day but you carry through it or deal with it, it's rng.
NOW!
You get a team of people all on 50 game loss streaks vs people on 50 game win streaks and the guys on the winstreaks are all lower mmr than the losing streaks.
That mmr difference don't mean shit, they still gonna lose. Stop matching people like this. Seriously
Nea1048 votesIn a fair, legit and overall good system "stomps" should be a tiny minority. For who knows a little, think about a bell/gaussian distribution. Since, instead, they are the majority of the games, and since you can mostly see them happen in the very first minutes of the game (so, no snowballing), something is very wrong.
What is "more wrong" is that they have done NOTHING about for years, and still are ignoring AND denying the issue.
Kai Guy2 votesThat makes the assumption about gameplay not being inherently stompy or snowbally. PvP games of skill always have an inherent potential for snowballing due to skill gaps.
The Distribution for MMR is not built from gameplay, just match results. So targeting MM as the issue for snowbally gameplay is a mistake, as gameplay itself is the issue. As gameplay is PvP regardless of the devs balance you run into issues from players if theres widespread behavior issues then there is gameplay issues.
Riot needs a behavior Red again. Some one whos job is to engage with the community and targets player related gameplay issues. That job sucks thou. People on the internet when upset are... well... unpleasant is a polite term for it.
I am glad you know about Gaussian distribution! If you ever see some one demanding that Riot should start all new accounts at the bottom of iron could you please explain to them that if they are under the 50% marker on the ladder what they are asking is to have their title be moved down to Iron with out making any impact on MM.
Issue with "pairwise" comparison in a team game is it does not feel very fair for the average player. Because they get a negative experience with any below average teammates and above average opponent. But by definition of being the average they are gonna be the majority of a curve so they end up being properly placed by skill which is the entire point of a skill biased ladder and skill MM.
So.... If folks want better quality games then game quality needs to be addressed, PvP gameplay is built by players.
Take a look at team games, folks bitch constantly about MM regardless of the dev or game. No one thinks they are equal to their peers because they take their best performances and compare that to everybody else.
So... TLDR. Boards and players are barking up the wrong tree. Riots MM has issues its just not what you see complained about. I would highlight Inaccuracy of Titles and Duo implementation as negatives, and its certainly an opinion I have but I can walk folks over the logic and math for my view with provable values. Great MM with flawless implementation does not inherently make good gameplay.
Exin03 votes[{quoted}](name=Kai Guy,realm=NA,application-id=3ErqAdtq,discussion-id=7MLKN1ir,comment-id=00010000,timestamp=2019-09-03T18:04:41.774+0000)
That makes the assumption about gameplay not being inherently stompy or snowbally. PvP games of skill always have an inherent potential for snowballing due to skill gaps.
The Distribution for MMR is not built from gameplay, just match results. So targeting MM as the issue for snowbally gameplay is a mistake, as gameplay itself is the issue. As gameplay is PvP regardless of the devs balance you run into issues from players if theres widespread behavior issues then there is gameplay issues.
Riot needs a behavior Red again. Some one whos job is to engage with the community and targets player related gameplay issues. That job sucks thou. People on the internet when upset are... well... unpleasant is a polite term for it.
I am glad you know about Gaussian distribution! If you ever see some one demanding that Riot should start all new accounts at the bottom of iron could you please explain to them that if they are under the 50% marker on the ladder what they are asking is to have their title be moved down to Iron with out making any impact on MM.
Issue with "pairwise" comparison in a team game is it does not feel very fair for the average player. Because they get a negative experience with any below average teammates and above average opponent. But by definition of being the average they are gonna be the majority of a curve so they end up being properly placed by skill which is the entire point of a skill biased ladder and skill MM.
So.... If folks want better quality games then game quality needs to be addressed, PvP gameplay is built by players.
Take a look at team games, folks bitch constantly about MM regardless of the dev or game. No one thinks they are equal to their peers because they take their best performances and compare that to everybody else.
So... TLDR. Boards and players are barking up the wrong tree. Riots MM has issues its just not what you see complained about. I would highlight Inaccuracy of Titles and Duo implementation as negatives, and its certainly an opinion I have but I can walk folks over the logic and math for my view with provable values. Great MM with flawless implementation does not inherently make good gameplay.
but isnt whole point of having MM system to minimize that skill gap? Casuals dont want play with and against skilled players they want to play at same level otherwise there is no point to play. Yes people are toxic but i dare to say major part stem from frustration from uneven game. Less skilled players are stomped and they are frustrated, commonly in this grup is player on higher skill level and he is frustrated also because rest of the team cannot keep up with him and support him.
Kai Guy1 votesbut isnt whole point of having MM system to minimize that skill gap?
Absolutely,The issue is players use subjective judgment and don't create a realistic expectation for skill or how MMR sorts it out.
Averages are naturally occurring. so building a curve is not that difficult. MMR = a Mean average representation of any 1 accounts performances. Its made from the population's average level of skill and how you match up to that.With low data its not an accurate representation of skill. Elo systems are built of statistics and Scientific method logic of self correction. They don't care about moves in a game, just the result. The fundamental assumption is, in a game of skill, your skill impacts your ability to win. To deal with teammates impact you use Standard Deviations, there's a wiggle room, or a range things can exist in. So a few points = same expectations.
Now, using the average player does not really seem ideal to them because it means they exist in a MMR range where the very good and very bad players are expected to exist and impact their matches. There are solutions to this using MM thou. You build games and account for uncertainty as often as possible so that quality of individual matches increases and # of mismatches decreases.
If riot does this? No idea. sorry, but I don't have their systems algorithms to check or test.
But the metrics players use? They are not at all about the context of being on a curve or averaged performances. People use their good games as a representation of their skill and discount off games as flukes or bad luck. Hence everybody and their dog thinks they are superior to their peers. They judge skill of in game plays like did some one land that R I needed or how good is vision. Things irrelevant to a Objective system.
Also folks think some how Riot can magically know an accounts skill from the start and assume stats are potentially usable as a good metric. But subjective stats outside of the context of a match are filled with noise and the only way to sort that out is using WR. If you don't then you need to compare individual plays to the game played itself, so either you need a human or a system that understands every possible move and what is most optimal then grade off that.... so youd need a rather advanced AI.
Then there is player population, its not like there is an infinanly large sample size of accounts to pull from at any give moment on any given day. Mismatches occure but part of the formula is made to account for skill gaps. We variable generation. MMR systems tend to adjust on some variation of Arpad Elo's work. He used Rn=Ro+K(W-We).
Again... if a game is just inherently stompy. It does not matter how good MM quality is, games are going to be stompy.
The common excuse is 1 bad player = loss. yet a 1 bad enemy = a win. So the distribution odds here favor your team as your always on it and consistently should not be a game losing player if your at or over your position on the curve. The flip is true, the enemy team is more likely to have a game winning op strong player.... but.... well that by definition requires you to admit that level of skill is not you.
So its "fair" and 50% assumed odds when everybody's accurately placed, just comes down to whos team overall preformces is better, which means that your not gonna pull weight of folks with weak performances so the games honestly just a dice roll. The solution thou... is to improve and as you get better you start shifting your self away from the average. But... people don't do this. They play the same shit the same way every time trying to get the results of their best performances, games where they are vs weaker players who are high uncertainty or had great matchups or teammates that are overskilled. They don't check numbers and they don't improve by intent and consistently fail to understand why closer games result in their impact being lower.
Saying its just ELO HELL or what ever FOTM variation of that is far simpler then taking accountability.
Exin02 votes[{quoted}](name=Kai Guy,realm=NA,application-id=3ErqAdtq,discussion-id=7MLKN1ir,comment-id=0001000000000000,timestamp=2019-09-04T18:28:49.516+0000)
Good assessment but here is another problem we simply dont know how or if MM system works at all, we have no numbers to crunch and any data we have are some approximations from external sites and we also dont know if they have right data. If you look on your mmr somewhere it may be right number or not. Only tangible thing related at least somewhat to skill and mmr are ranks. When silver players see dia and plat players in their games its hard for them to accept and understand how MM system deems this game will be fair and frankly when you see dia player you first reaction is wow high elo player we are screwd and not hmm boosted unskilled joe freelo. Also there is sharp rise in MM complaints, can be fluke but its also giant red flag that maybe something is wrong and riot radio silence isnt helping at all. Even in case that MM system is working perfectly and all games are statisticaly fair lack of means for people to compare or see on what attributes where teams created its huge problem and infinite pool of resentment, frustration and toxicity. Also lol is here for decade and millions and millions games were played so it hard for me accept notion that MM system have not enough data to somewhat asses players. And let me ask you this do you thing game is now inherently stompy because it that is the case whole discusion about mmr skill and mm is pointless.
Kai Guy1 votesYou study the real thing to identify counterfeits. Most posters don't take the time to do this. I get a lot of shit because I did and can point out the errors people make in their posts. End of the day, MMR distribution can be graded by looking to the tail ends of its curve and You can observe the exact results from Riots system that you would expect for a functional MMR.
So the issue really just becomes how optimized is their program and what features do they introduce that are negatives for players.
you look on your mmr somewhere it may be right number or not. The actual values? Like if your 1400 or what ever? Don't matter. They are arbitrary and only have value in the context of the system. 8 is low value right? but 8 out of 10 is the same as 80 out of 100 or 800 out of 1000.
Any one can be used to represent the same 80% ratio. The way a MMR/Elo system is made is that you only achieve your raiting if the results of your performances warrant it. Its the same math for everybody despite the fact that the variables are different. If some one overlaps with the same WL and WE they will have identical results. 4+4=8 and 7+1= 8. There's multiple ways to add numbers to achieve 8 but they have to be the right values.
The issue I see is players leap into any excuse to avoid responsibility. Elo hell was a thing for ages. MMR is "forced" "Rigged" is the current popular excuse. Now i will say it again, the examples given consistently in these threads are all subjective assessments that have no fucking relation to how an elo system functions so you can dismiss them as inaccurate and odds are the devs for this game do exactly that. Why I spend the time and energy trying to get people to understand the functionality of MMR. Because proving something needs improvement and showing the data that supports it might actually cause riot to make improvements.Once, when the community was smaller this happened form time to time because there was not as much division. A red posts now in gameplay and some ones going to rage and flame them instantly. Doesn't matter what they work on they get flack. Which is a hassle because communicating issues to devs has become way more difficult.
As it stands most posters are doing the equivalent of telling people that 4+4=44 and any one with a basic understanding knows they are wrong.Why do I say they are doing math so badly? Because Forced and Rigged systems don't work with MMR. The value is built from results of games. If you rigg the results prior to the game then your not getting an accurate value. If you don't have an accurate value how in the fuck do you use that inaccurate value to then accurately rig a game?
infinite pool of resentment, frustration and toxicity.
I am pretty vocal that we need an overhaul on behavior systems so that gameplay improves. PVP = Players drive the gameplay. I can show mathmaticaly the impact on players in the context of a game of league. Lets use 10% as an example. Lets say only 10% of players in silver have negitve behavior that sucks the fun out of a match. How often do you think you would see those players in a set of 10 games? Hell, if you think the % is different go ahead and give me a value you thinks more accurate. Once I get a response from you, I will show you how to test this with RNG distribution so we can look at a visual for it.
When silver players see dia and plat players in their games its hard for them to accept and understand how MM system deems this game
Yet fairly Consistently, You don't see massive class interval gaps in this game very often in the same que type. Que rankings don't carry over to other ques. Skill, often times does. I wish riot took a harder stance about smurfing but players enjoy it. I wish riot changed Duoing, but players enjoy it. I wish the competitive ques encouraged 5 man premade, but players don't understand that removing the RNG of teammates is a positive for competitive environments.
Also lol is here for decade and millions and millions games were played so it hard for me accept notion that MM system have not enough data to somewhat asses players.
The best way is already known and has been in use from before riot even made their system. Its off using W/L if your system is supposed to be objective. Its just a matter of optimization, some games can have additional features that use in game stats to impact Raiting gains/losses. Like Elo has been around from the 60's and was done as an improvement on the Harkness rating system which was in the 50s. So we are talking about well over 50 years of peer review by mathematicians and statistician but some kiddo on the boards thinks cuz their mid tilted and AFKED that's all the proof they need to say that its broken, utterly oblivious to the system not careing remotely about in game data or singular matches.Hell most folks use 3 games as a streak and dont even understand its a balance so order of wins is not important.
What's relevant data? what's proven to work? This shit is already known.
Remove W/L and aim for just Subjective is much harder because it has to treat each game as a stand alone event and use the context of all stats in that individual match. Far as Statistical You need the context of the game if you want to rate plays from the game for skill. That's requires a human who knows the game very well to grade each player, or a really advanced AI. If your using something that does not look at each match but the results and the numbers generated, the most relevant statistic is win and loss. Everything else Is a poor representation and the amount of effort to build a system around stats with out considering W and L as relevant is insanity. Also highly inaccurate and would constantly change on a hour to hour depending on popularity.
The data used to sort out the noise from stats is consistently the things that players cant fake. This is why Arpad Elo used W/D/L and context of assumed skill by players. The amount of data to shift thru to build a good subjective system approaches values so high that current tech lacks the computational powers to handle it.
Binominal distribution is already pretty complex.
Exin01 votes[{quoted}](name=Kai Guy,realm=NA,application-id=3ErqAdtq,discussion-id=7MLKN1ir,comment-id=000100000000000000000000,timestamp=2019-09-04T21:42:07.517+0000)
Again great assessment, let me respond with little explanation, english isnt my native language so there is good chance i write my thought i bad way. Firstly iam in no way against you or trying rebuke your response i thing you are right in many ways. You take difficult path to scientifically asses how system works and i am not in position to even try to tell you are not right, but 99% players dont choose this way we simply need something tangible for average joe so even he can comprehend why something happend and not mysterious equations behind scene. HoTS have same problems with MM and they were able to provide some number for players maybe it was only randomly generated number with no substance but even that it at least give something to grasp and make people more calm ale they were able to improve MM system to give more fair games so technology and ways are there and i simply dont understand why riot is so adamant to do nothing, we are at point where even "cosmetic" change just for illusion that something is happening can greatly help.
Kai Guy1 votesRiot tends to do things pretty frequently. Team Elo into Leauge system. Pick order into Role selection. Positional Ques to deal with auto fill. Positional MM to deal with auto fill.
Players just have shitty expectations and don't take a moment to learn the bare basics behind MM systems to actualy set realistic expectations.
Take a look at any multiplayer team game. You will find posts that look dam near identical to league.
HOTS. "yet to hire someone who understands team games skill rating should based on stats (taking outnumbered fights and dieing, trickeling, not using AA at the right times) rather than w/l. And we know they can keep track of all these stats so its ridiculous they dont implement this."
"years pass by and this system still is pretty bad…not even worth explaining myself…im just gonna keep pointing it out till those meaninless placement matches worth or mean something …and even the matchmaking system. I mean, everything is wrong…too sad it kills the game…cya! =)"
Dota 2. "Team sucks 50% of the time So, 50% of the time i am trying hard and winning flawlessly, but other 50% of the matches 3 out of 5 of my team are ♥♥♥♥ing USELESS. Some ♥♥♥♥♥♥♥ may say that its an excuse, but why is matchmaking so ♥♥♥♥ing ♥♥♥♥ed all the time? How the ♥♥♥♥ do i win with a ♥♥♥♥ing PA ♥♥♥♥♥♥ who has blight stone and boots on 25th minutes while being in triple safe lane and i am creating space all over the map as ember. Anh why the ♥♥♥♥ is alchemist able to demolish our whole team, but we cant retaliate "
Halo. "please fix the match making I literally matched some dude who never even played this game in RANKED its so crazy because I knew I could have won that game by myself my teammates were no help at all please fix the balancing, "
"Me and my friends who are either high diamonds or onyx's or champions would go onto balanced search type and when we would get into the game, we would lose because we were given a terrible, unranked, gold, low diamond that should be platinum, or a level 26-59. "
So is it a global industry scam? Or do players just not understand the systems used.
People say USE STATS. Math consistently concludes that any paddable stat is worthless as a metric because its not an accurate value to use.
Don't trip out over your English, your plenty clear. People use other players opinions to form their own, but a massive # of opinions are made completely ignorant on the topic at hand. Frequency does not have any correlation to truth.
DontTypeJustPlay1 votesIn a fair, legit and overall good system "stomps" should be a tiny minority. For who knows a little, think about a bell/gaussian distribution. Since, instead, they are the majority of the games, and since you can mostly see them happen in the very first minutes of the game (so, no snowballing), something is very wrong.
What is "more wrong" is that they have done NOTHING about for years, and still are ignoring AND denying the issue.
Honestly though I miss old league where they had a 16-1 fiora top beating the shit out of everybody on our team and bot lane and mid lane were slightly ahead 4-2 mid 6-3 adc with a 2-1 support, and it took the whole fucking team to come together and take down that fed mother fucker while the rest of the enemy team wasn't the PRIMARY focus but they were still threats. If you commited everything to the 1 fed guy the rest of the team took you down so you had to cordinate somehow. Now?
oh now that 16-1 guy just get's focused while the rest of his team 4 mans 1 guy on your team and still loses 4v1 to that 1 FUCKING GUY because if you fall behind you might as well be a fucking cannon minion.
Also not to mention match making matching losing players with losing players and winners with winners and the only difference is winners have slightly less mmr
Kai Guy2 votesTry again with realistic numbers. The exaggeration might win you some fans from the tilted players here to bitch about the game but no one who truly cares is going to take you seriously when your example is a team with 200 consecutive losses vs a team with 250 consecutive wins.
The biggest issue with MM at the moment is player base related and subsequently we need Riot to step up their behavior systems. The lack of a community red for behavior is a negative thing imo, thou I admit that job probably sucks.
MrEkkkkkko1 votesoh, i guess ur playing the beta version of tft? this game was designed as a rng game!
BradBen1 votesUmm i kind disagree matchmaking is quite good these days but the ping is bad or just goes crazy but again i did get a 70% winrate after the patch so maybe u are right lmao
CervixGrappler1 votesMatchmaking is truly fucked.