North America - English

League of Legends Community

League of Legends Community > League of Legends > General Discussion
Help Riot improve matchmaking! Looking for examples of bad matchmaking

View Poll Results: How many games do you expect to play before you reach your true Elo?
< 10 Games 171 1.59%
11-40 Games 798 7.42%
41-80 Games 1,144 10.63%
81-120 Games 1,710 15.90%
121-160 Games 649 6.03%
161-200 Games 903 8.39%
200-399 Games 1,667 15.50%
400-599 Games 1,055 9.81%
600-799 Games 444 4.13%
800-999 Games 166 1.54%
1000+ Games 2,050 19.06%
Voters: 10757. You may not vote on this poll

First Riot Post
 
Comment below rating threshold, click here to show it.
mayipora ?? Senior Member
This user has referred a friend to League of Legends, click for more information This user has accepted the summoners code, click for more information
02-16-2012

o.o since red is still reading this post, I'll give my 2 cents on the duo queue issue.


The problem with duo queue really occurs when two players with different skill level play together, that is the rating (or true rating) differs. But how different they may be, we can always get to an "close enough" estimation of the duo rating with a few games (or few hundreds) if they play a lot together.

So the obvious approach would be have a different rating for each pair that plays ranked, but this is neither feasible nor efficient.
My suggestion is that we add a duo queue rating system. The system work something like this: 1. Rating that adjusts when you play with someone who has a significantly higher rating ( say ~1400 play with 1600+). 2. Rating that adjusts when you play with someone who is about at your level (~1400 play with ~1300 or ~1500). 3. Rating that adjusts when you play with someone who is significantly below your level (~1400 play with 1200-).

The three different rating can be used to match up when people goes duo queue. For example, solo queue rating of 1400 with 1600 may be playing against a 1600 team since the the 1400 actually belongs to the 1600 group but stucked in elo hell or doesn't have time to play. Another group of 1400 and 1600 solo rating may be playing against a 1400 team since the 1600 was carried by someone else to 1600 rating.

Anyways just some of my thoughts. Pretty sure you guys can deliver something better =D

 
Comment below rating threshold, click here to show it.
Tigre the Legend ?? Senior Member
This user has referred a friend to League of Legends, click for more information This user has accepted the summoners code, click for more information
02-16-2012

I would like to see players on a team have similar elos, instead of having people at vastly different ratings in solo queue placed on the same team.

 
Comment below rating threshold, click here to show it.
Ginga ?? Senior Member
This user has referred a friend to League of Legends, click for more information This user has accepted the summoners code, click for more information
02-16-2012

@Lyte - I don't know if this site has any specific emailing function built in or were we supposed to email you normally(Hotmail, gmail, yahoo, etc.), so I sent you one via hotmail.


Will we get a reply to confirm whether it reached you? My email had two screenshots. One with a leaver who refused to listen to a team, and another who threatens to go AFK during champion select. Both far too common, and making getting back up to my original ELO much harder.

Also...

Has there been any ideas floating around among Riot Staff about how to deal with trolls in champion select? Like an alternative to dodging that won't damage the game in anyway, yet still allow players to avoid a 20-50 minute troll game? Like how about a Loss Forgiven to all players in a specific game if the troll/leaver they reported was deemed a troll/leaver by the judges and was Punished for it?(and all the resulting ELO loss be dumped onto him). They still wasted 20-50 minutes in a clearly unbalanced game, but atleast their loss wouldn't feel AS bad.

Unlike my previous suggestion, there will be no incentive for the 4 losing players to rage/force a single feeder to quit if they are merely doing badly, BUT if that one player was truly a troll or ragequitted, the remaining players can just try their hardest, and report after the game. And they can be relieved of their loss if you guys deem they truly are trolls/leavers.

 
Comment below rating threshold, click here to show it.
Ornerok ?? Senior Member
This user has accepted the summoners code, click for more information
02-16-2012

Are you interested in games where the enemy team had no chance of winning? I didnt take a screen because i wasnt sure, but i was just in a game (1400) where we won every lane and demolished them.

 
Comment below rating threshold, click here to show it.
C00LST0RYBR0 ?? Senior Member
This user has accepted the summoners code, click for more information
02-16-2012

Quote:
Originally Posted by CornerPocket View Post
------------LYTE-------------
What do you think about this idea? I posted this earlier, but it kinda got buried.

You've seen a few suggestions about giving increasing ELO gains for a winning team based on their collective KDA (no individual boosts). The problem with this, is that people would try to exploit it by delaying the game and driving up the score. You've personally expressed that concern in this thread.

How about this, though:

Give more ELO to a team with a much larger KDA than the losing team, but decrease that amount based on the length of the game. For instance, say at 45 minutes, all ELO increases based on KDA are gone.

Let's just say you have a percentage, and from the beginning of the game up until 25 minutes, it's at 100%. After 25 minutes, the percentage steadily decreases all the way down to 0 at 45 minutes. At the end of the match, some formula is used to calculate how much additional ELO each player should get based on KDA and then it is multiplied by that percentage. That should prevent players from trying to delay the game and get their score up.

Also, change nothing about ELO losses. If you give decreased ELO losses for players that get steamrolled, it would encourage those players to throw the game after they get slightly behind.... imagine the trolling, flaming and yelling. "Just let them win! We don't want to lose that much ELO! I'm AFKing in the fountain!"
OR
Just to make it interesting, you could make it the opposite for the loser. If there's a large difference in KDA (in the winner's favor), you decrease the amount of ELO they lose based on length of game. They lose less ELO the longer the game is. If it's a shorter game, they lose the standard amount. That way, if they're way behind, it gives them incentive to play it out and try their best to come back. Or they can just surrender and take standard ELO losses.

I think those improvements could add a lot to the system and they encourage both teams to try to win (in the case of the steamroller) or to try their best to defend (in the case of the steamrolled). I don't think there's any way to exploit this, aside from any ways there are to exploit the current system.
KDA based rankings are bad, because they don't actually rate what won the game. For instance tanks and supports get screwed over hard.

 
Comment below rating threshold, click here to show it.
Lyte Lyte's Avatar ?? Lead Designer of Social Systems

Follow RiotLyte on Twitter
This user has referred a friend to League of Legends, click for more information This user has accepted the summoners code, click for more information
44 of 173 Riot Posts
02-16-2012

Quote:
Originally Posted by dzell View Post
Yes! The area around 1100-1300 range is very chaotic it seems. I dropped to 900, worked my way up to 1100 and skill level is much more varied than the 900-1100 range.

Could you somehow use the # of games played to help smooth out the matchmaking process? I think this would make it easier to climb out of your elo range after your own personal skill level has improved. Both the number of games you've played AND your elo will be used to match you up with people.... I think this would be more accurate than elo alone. Could even factor in normal games played.

----------------
< 150 ranked games played, player skill very volatile
----------------
150-300 ranked games played, player skill more stable
----------------
300+ ranked games played, player skill very stable
----------------
Glicko systems have a 'confidence rating,' which is an extra parameter that represents how confident the system is that a player is currently at his true skill level. I think Glicko tries to solve the concerns you have raised.

Elo systems have K-factors, which are similar, but basically represent how important a match is. Early in a player's career, a K-factor starts high because the first few matches are important for determining the relative tier a player should be at.

League of Legends, however, has additional complexity which is why traditional Glicko or Elo systems are not sufficient. For example, a 1800 Vayne player is not necessarily a 1800 Annie player. This means that player performance and consistency varies greatly from game to game, adding a lot of additional variables that are hard to capture with traditional matchmaking systems.

In saying this, there are a lot of smart people at Riot, and we are going to do what we can to make the best matchmaker possible--it will never be perfect, but we will always keep working on it.

 
Comment below rating threshold, click here to show it.
Iaw ?? Member
This user has accepted the summoners code, click for more information
02-16-2012

Remove duo queue.

Make a true solo queue for ranked and normals

Give premades separate queues.

 
Comment below rating threshold, click here to show it.
C00LST0RYBR0 ?? Senior Member
This user has accepted the summoners code, click for more information
02-16-2012

e

Quote:
Originally Posted by kidco View Post
After 100 games i was 1460

After 200 I was 1150

After 300 I was 1460

When did I actually reach my true elo? I haven't improved since game 100. Is elo a fair judgement of my skill or just it mainly luck? I want to know ok? I really do.

http://na.leagueoflegends.com/board/...ighlight=kidco
There is no true elo, there is a range you belong to. 1150 and 1460 are the upper and lower bound of this range. So your average elo is 1300 with a variation of +-150. Especially with one-sided placement matches you can get off as far as 400 elo from your average, but you will eventually return to it.

 
Comment below rating threshold, click here to show it.
Lyte Lyte's Avatar ?? Lead Designer of Social Systems

Follow RiotLyte on Twitter
This user has referred a friend to League of Legends, click for more information This user has accepted the summoners code, click for more information
45 of 173 Riot Posts
02-16-2012

Quote:
Originally Posted by Vraxx View Post
I noticed that you wanted the win/loss of our friends when queuing for normal or normal draft. Just curious, how precise do you want those numbers to be, because as of now, its not the easiest to even find your own (as its only available at the post game screen) and I imagine it would be fairly difficult to track my friend's numbers as well. I liked this idea a lot and would love to help matchmaking become more balanced, but if there is anything that could be done about this inconvenience (or knowing if that field doesn't require precise information) it would be much appreciated
If you don't have the information handy, just providing estimates of your friends Win/Loss and their Summoner Names should be enough for me to find the games in the data. Oh, don't forget the region you played on!

 
Comment below rating threshold, click here to show it.
GrimsIey ?? Senior Member
This user has accepted the summoners code, click for more information
02-16-2012

Honestly, there should be a way for Elo to be calculated by how YOU do as an individual. Wins/loses 3/4ths of the time mean nothing. Elo is all about luck in having a good team right now, nothing else. It's luck of the draw, there honestly is -no- reason for there to be a ranked for the way ELO is calculated. Honestly if I want to be matched by wins/loses that should be the way normals are. Also, there's more trolls, leavers, and afkers in ranked then there ever will be in normal which really needs to be addressed before anything else imo. That's half of the reason why people lose games.