MM system is broken

dromb·Sep 26, 2015·1 votes·837 views

Problem 1: MM system uses only MMR to match up two teams, which will be fusturation to the players, who win many games in a row, while being in low division. it leads to a result where bronze player may play hes promotion games to next tier bronze versus top gold players. Example from live: http://postimg.org/image/yjr7b2xet I need to win 66% of games(3 games out of 5) vs challanger team, to get to gold V

Solution 1: MM system should match you up with players who are +/- 1 your current tier, while using MMR to match you with the best/worst players in that +/-1 tier currently online. AND Divisions and tiers will define current tier maximum and minimum MMR

How will soluton 1 solve the problem: **Example: ** Silver V(MIN.MMR=1100; MAX.MMR=1300) Silver IV(MIN.MMR=1250; MAX.MMR=1450) NB! Overlaps are neccesary so you could be able to play with 1 tier higher/lower tier You are silver V Your MMR: 1200 You win 5 games gain 30 MMR per game, you have 96LP

Now your MMR is 1350, but you wont me matched with higer MMR than 1300, because that is the maximum MMR placement silver V can be + all silver worthy players will be matched with eachother.

Now you win 2 games (+60MMR) and move on to silver IV Now your MMR: 1350 + 60 = 1410 You gain faster LP but still get same MMR in the background. now if you win 4 games, getting your LP on 80, it does 120MMR. Your current MMR = 1530 Silver IV max MMR = 1450 You will be MAX placed vs 1450 MMR players. till the tier changes.

Problem 2: MM system currently makes unfair teams, where 1 team has either WAY higher MMR than the other. How problem 2 appears: It will appear when you win too many games in a row and queue time gets long(5-10 minutes)

Solution 2.1: Solution 1 will fix this. because everyone who reaches to the top silver tier, will be placed with top silver tier players.

Note. On Silver I to Gold V there will be comparison between those 2 Note. On silver V if you lose much, you can be matched with bronze I, because of the overlap in MMR.

Why is it better than system which RIOT is using right now? +positive sides:

Divisions and tiers will be describing your skill better.
People will have more competitive plays on theyr level(no "overrunning" from the other team)
Happy players
interesting games, players will learn better(you wont learn anything, if skill levels are too different) -negative sides:
Queue may be longer for some seconds up to 90 seconds during daytime(the time normal people are awake)

Maybe this is more comfortable to read and describes my ideas better, than the one I wrote before.

Sincerely yours, Valter

11 Comments

drombSep 26, 20151 votes

Can you answer me RIOT employee?

ShikigamiSep 26, 20151 votes

To be honest, this is ranked 3's queue. Nobody ever plays it.

drombSep 26, 20151 votes

{quoted}

To be honest, this is ranked 3's queue. Nobody ever plays it.

The logic on the Queue's either they are 5 v 5 or 3 v 3 are the same. You think they wrote the "different" code for every gamemode? you must be crazy. there is 1 method which calculates how the team is put togehter. and it is used on EVERY queue. So either its 5 v 5 3 v 3 or whatever the hell other game modes are, its ALLWAYS the same logic.

Now im still waiting for the response

drombSep 26, 20151 votes

{quoted}

To be honest, this is ranked 3's queue. Nobody ever plays it.

to be honest there are more images than this 1, which was super overdone, just to prove a point.

drombSep 26, 20152 votes

+1 more comment for you. Nobody cares about "you watched 1 pic" read the text i wrote. Those are just simple examples what this system does right now. But if you are 12 y/o just starting to get puberty, then i understand you. Boys like you can't focus on 1 thing more than 30 seconds, like for exaple reading and trying to understand what I wrote down.

ShikigamiSep 26, 20151 votes

Jesus, there's something called sleeping, I do that at 12 AM, just fyi. Anyway, find an example like that in soloqueue or ranked 5's. As the MMR calculation stands, it's a gamemode with limited amounts of players. That being said, in soloqueue you have hundreds of golds lining up to play, whilst nobody plays ranked 3's so the system matches you with the only others who bother to play that game mode. A few years ago, I played with NyJacky in draft dominion, because nobody played it so I got placed with an LCS player.

Again, this being said, find an example of this on a popular gamemode, and you'll see that you can't, because the system works.

OhBoyItsaMegamanSep 26, 20151 votes

MM system currently makes unfair teams, where 1 team has either WAY higher MMR than the other.

I think you're jumping to conclusions based on op.gg's estimate of MMR. Riot has not revealed the formula they use for calculating MMR, which means that op.gg's is just a guess. Matchmaking puts teams of equal MMR against each other. When you see a team with higher MMR than the other, that's op.gg getting the MMR wrong.

The Silver vs Challenger 3v3 looks like it was played at 2:30 in the morning. Ranked 3v3 is an unpopular queue and you were playing very early in the morning; there may not have been any other teams queued for a game on the entire server.

Matthias9119Sep 26, 20151 votes

MM system uses only MMR to match up two teams, which will be fusturation to the players, who win many games in a row, while being in low division. it leads to a result where bronze player may play hes promotion games to next tier bronze versus top gold players.

It's very, very hard to stay in Bronze with a Gold MMR unless you deliberately lose your promos over and over. If your MMR is way above where it "should" be for your division, your LP gains become extremely large and your LP losses become extremely small. So basically you'll constantly be in promos even at a 50% winrate, and eventually you'll manage to win 2/3 or 3/5 and get promoted.
At some point, if you get far enough above where your MMR "should" be, the system will automatically promote you. (In the first season with divisions/tiers people thought it was funny to smurf, dodge their promos every time, and work their MMR up to Plat+ while still in Bronze. You can't do that anymore.)

Example from live: http://postimg.org/image/yjr7b2xet I need to win 66% of games(3 games out of 5) vs challanger team, to get to gold V

Is this a newly formed 3s team? If you started a team with established high-MMR players, there may be a limit to how high it can initially place your team, so you might have an MMR much higher than your division at first.

Ranked 3s might also just be wack. Much smaller playerbase than 5s.

Solution 1: MM system should match you up with players who are +/- 1 your current tier, while using MMR to match you with the best/worst players in that +/-1 tier currently online. AND Divisions and tiers will define current tier maximum and minimum MMR

Abusable. Keep dodging or intentionally losing promos and you get to stomp players who are much worse than you. If you eliminate promos and promote/demote based solely on MMR you defeat the whole purpose of the tier/division system.

Problem 2: MM system currently makes unfair teams, where 1 team has either WAY higher MMR than the other.

It does not do this, at least not deliberately. Division/tier is completely ignored by the matchmaker, it always tries to create teams that are close in MMR.

Edit:

I'm not saying Riot shouldn't try to improve its matchmaking, but there are reasons it works the way it does, and you don't seem to fully understand how it works.

drombSep 26, 20151 votes

Problem 2: MM system currently makes unfair teams, where 1 team has either WAY higher MMR than the other.

It does not do this, at least not deliberately. Division/tier is completely ignored by the matchmaker, it always tries to create teams that are close in MMR.

Yes, it tries, but i rather wait for 20 minutes game than 10 minutes and go to losing game allready. op.gg one is example, how you can go. Those algorythms teams are put togehter, are not different from the ones you use on 5 v 5.

Problem 2: MM system currently makes unfair teams, where 1 team has either WAY higher MMR than the other.

It does not do this, at least not deliberately. Division/tier is completely ignored by the matchmaker, it always tries to create teams that are close in MMR.

Is 1800 MMR player with 20% win rate same as 1800 mmr player with 70% win rate? this team i had on op.gg average W/L ratio per player was near 43%(Mine was 83) Enemy team had it near 58% average. If they dont want to try to make the system better, i feel like winning in strikes is not useful. I rather lose 1 game after every 3 wins on purpose, to no "rush" away with my MMR. unless i will skip tier. (if im Plat I, i rather keep my mmr as low as possible if i go to promos) Same logic as you told they gonna abuse mine. + how are they gonna abuse my system to go "further"?? If silver elo is 1100and you can beat then, you are not silver. you dont need to play vs plats or diamonds to prove your not silver V. if you want to reach silver IV you need to beat those players, not WAY above your skill level.

I had the story longer before, but some of the people started whining, so i made it short. The reason i told that it is making unfair teams is this: I playd 10 games, all ranked. I had 0 ranked games.(on NA diamond I) but i needed account on west to play with my friends. So, i finally reached 30 and started playing ranked. First 3 games were even teams(yes, they were easy for me, but it was placement matches, so it will rise very fast) after that i started noticing, that enemy team MMR is WAY higher than our's. (on 4th game our team was full of silvers, they had full of golds.) 5th game was same deal MMR difference was near 140. 6th game mmr difference was 160(won that one too, it was 4 v 5, but i did some plays and carryed it anyways.)

on 7th game it was the worst. Theyr team was full of plats, while we had 1 gold V and others were silvers. I was suprised. and i lost 8th to 10th game coz of this.

If they want to make the system better, they should either do as I sayd at that topic, or change on different direction and do completly oppisite. If you are silver and have diamond MMR, then after winning your last division game on silver I you will skip to diamond V. You are proving yourself vs those players anyways, then why should I stick to silver or gold?

This picture I made from the 3 v 3 team, was just a test, either its possible to skip a whole division isntead of a tier. Its not. + Its super fustrating for a player or a whole team, who has had to prove vs challangers that they are able to beat those and then get promoted to gold V.

Matthias9119Sep 27, 20151 votes

Yes, it tries, but i rather wait for 20 minutes game than 10 minutes and go to losing game allready.

If you're finding matches quickly it's because the matchmaker thinks it has found a close match already. It does start to broaden its search tolerance after while, but that normally only happens if you have a very high MMR or you're on at a less popular time (or both).

Maybe you could let people opt out of broadening the search, but if it's not finding a good match right at your MMR range in 5-10 minutes it might take a long time to find one. Also this doesn't really matter in the long run, because MMR and LP gains/losses are adjusted if there is a significant MMR difference between the teams.

op.gg one is example, how you can go. Those algorythms teams are put togehter, are not different from the ones you use on 5 v 5.

I'm guessing English is not your first language, but I have no idea what that is supposed to mean.

Is 1800 MMR player with 20% win rate same as 1800 mmr player with 70% win rate? this team i had on op.gg average W/L ratio per player was near 43%(Mine was 83) Enemy team had it near 58% average.

AFAIK, no, winrate is not taken into account. Elo rating normally does not do this, and their MMR system is based off of that but modified for a team game.

It's unusual to see players with winrates far from 50% unless they're new. Within 50-100 games players usually seem to be close to the right place on the ladder. Obviously ratings on new players are likely to be less accurate, but there is not much Riot can do about that unless they completely change how Ranked works. There are some ways to modify Elo-like rating systems to try to explicitly take that into account -- Microsoft's Trueskill system, for example, maintains both a rating and a confidence factor for each player. They could also do something more like Hearthstone's rankings where it doesn't have to be balanced -- but Hearthstone completely resets their ladder monthly, in effect, so they don't have to worry so much about rating inflation.

Re: the rest of your post:

The abuse case with what you're asking for isn't that people can game the system to rank up higher. It's that they can game the system to constantly go up against bad players and stomp them. If you really are good enough to be in Gold or Plat, you shouldn't be able to keep losing promos and stomping Bronze or Silver players. (And there are people who would do that.) Riot's current system prevents this because rank has no effect on matchmaking, only MMR (i.e. how good the system thinks you actually are). Anything that lets you artificially manipulate who you go up against is very bad.
I can't speak to your specific placement match experience. Riot has stated that they always try to create matches where they believe each team has as close to a 50% chance to win as possible. "op.gg says we were X MMR apart" is meaningless -- third party sites provide poor estimates of MMR at best.
They didn't design the system to be optimal in cases like "what if I'm in Silver with a Plat MMR?" because it's almost impossible for that to happen unless you game the system to make it happen. If you're more than a division or two below where you "should" be according to MMR, holding a 50% winrate will have you almost constantly in promos, and you'll likely skip at least one division when you do win a promo series. You can also forcibly get promoted at a certain point.