Which matchmaking 'issue' is the most important to you? (See post for more details!)

1) AFKs in Champion Select Lobby 4,849 36.32%
2) Duo-Queue Elo Disparities in Ranked 1,000 7.49%
3) Skilled Ranked Players in Normal Modes 657 4.92%
4) Premade Matching 663 4.97%
5) Transitioning from Normal to Ranked Mode 1,337 10.01%
6) Free to Play Champions in Ranked Mode 800 5.99%
7) Random Champions in Ranked Mode 644 4.82%
8) Provisional Matches in Ranked 716 5.36%
9) Duo Queue Prevalence in Ranked 416 3.12%
10) Level Disparities 638 4.78%
11) Team Margin of Victory 1,632 12.22%
Voters: 13352. You may not vote on this poll

After Hours with Matchmaking and Lyte

First Riot Post
Comment below rating threshold, click here to show it.

WizzadDeBz

Senior Member

09-24-2012

Professor Dr. Lyte, is there any way a regional parameter could be added to matchmaking? Language differences are super frustrating in a game that relies very heavily on teamwork. Cultural differences can also lead to disunity and misunderstanding among teammates. This is obviously only an issue when there are country gaps (Mexico/Brazil matched with US.) Maybe there could be something that ties language proficiencies to an account for matchmaking purposes instead of basing it on regions. I have no issue with people who can speak English playing with me, it's just those people whose only English is Noob, bads, gg bad team, etc...


Comment below rating threshold, click here to show it.

Mokkun

Senior Member

09-24-2012

Um... so here's my issue, my last normal game, we had a 2200 player on my team.

I'm... 1257 (granted with all of 40 ranked, but I'm no diamond, not likely a gold either)

I was laning with someone with a low enough rating that lolking wont tell me what it is.

In addition, we had another gold grade player. But their team had 1 gold, and some silvers.

Basically, how was there both a rating gap greater than my total rating, and no compensating super-high player to make up the difference?

I'm used to upwards of 6-800 differences top to bottom in normals, but that was absurd. I know it's normals, but something still seems really off in matchmaking there.


Comment below rating threshold, click here to show it.

Durr

This user has referred a friend to League of Legends, click for more information

Senior Member

09-29-2012

I know this is the mega-thread, but I have taken the elo disparities to a seperate post found here http://na.leagueoflegends.com/board/....php?t=2624481 to get votes and suggestions on it.

Thank you so far for the wonderful progress you have made Lyte and the team, and let's continue to make it better for all of us.


Comment below rating threshold, click here to show it.

IIZekoII

Senior Member

09-29-2012

Quote:
Originally Posted by Lyte View Post
Could be late hour, could be who his teammates were (i.e = smurfs but on Level 24 accounts). It's hard to explain any specific 1 game without a bit more details about the context and even then, a nearly perfect matchmaker will have some % of games be lopsided.

yeah ok. but...
ah yeah fix your game,your client and your server please. k thanks.


Comment below rating threshold, click here to show it.

Purgation

Senior Member

09-29-2012

Quote:
Originally Posted by Lyte View Post
Recently, we collected a lot of community feedback on matchmaking.
4) Team Margin of Victory
Some players argue that in epic, close matches, teams should not gain or lose the exact same Elo as a lopsided match. What are good metrics for "team margin of victory" that are focused on promoting team play and not individual performances?
This is a bad idea. A blowout victory is an indication that your matchmaking isn't working the way it is supposed to... this is because:

1. ELO was not designed for teams, and how to correctly adapt it is not obvious (e.g. taking team averages)

2. ELO works best when the pool doesn't have too much churn. You're base has tons of churn. Instead of everything evening out over time, low ELOs drop from the game and new players constantly flow in at 1200, creating a "downdraft"


Comment below rating threshold, click here to show it.

Purgation

Senior Member

09-29-2012

Quote:
Originally Posted by Lyte View Post
On average, if your Duo-Queue is far apart in Elo, you will end up first pick and last pick. On average, if your Duo-Queue is close in Elo, you will end up paired together near the middle or the bottom.

However, what I meant by 'advantage,' is that a Duo-Queue close in Elo typically wins more matches than the system expects. A Duo-Queue far in Elo typically wins fewer matches than the system expects.
... which means that the system isn't working properly? For ELO to work in a team matchmaking system it is EXTREMELY important that matches be made extremely fairly. If the matchmaking cannot effectively control for all of the non-ELO factors that is a big problem.


Comment below rating threshold, click here to show it.

Purgation

Senior Member

09-29-2012

Quote:
Originally Posted by Lyte View Post
We have talked about similar implementations for solving #1.
We definitely could do a better job 'introducing' players to Ranked Mode, and a short video PSA might help, but wouldn't be the full solution.

I agree we could raise the requirements to play Ranked, but 200 Normal Wins might be excessive. Right now, it takes about 200 Wins to get Level 30, so I assume you mean 200 Wins after Level 30? Basically, that is about ~100 hours of games which is longer than it takes to hit "cap" in most other games these days!

Agree, I think part of the problem is educating players that Ranked is "bring your best, every single game."
The real problem is that 1200 ELO is the wrong place to start new players (this is a feature of chess-like ELO where the total ranked population doesn't experience churn like LoL does)

This is speculation, but I am guessing that if you looked at your data you would find that new to ranked accounts on average drop like a rock to something much lower far more than the reverse. One possible solution is to explicitly factor number of ranked games played into matchmaking, as well as shifting the starting point downward to try to avoid the "downdraft" problem.

It also avoids two problems - - first, it may help differentiate players who are just new, from those who have richly earned their low ELO from afks, trolling, and other bad behavior. second, it should create a better overall game experience, by letting people start at a level where they can improve from, rather than the experience of getting hammered from 1200 to 600 and yelled at by teammates the whole way down...


Comment below rating threshold, click here to show it.

Purgation

Senior Member

09-29-2012

Quote:
Originally Posted by Lyte View Post
The chances of a 15-game winning streak followed by a 15-game losing streak is extremely low.
If the matchmaker was working properly, the chance is incredibly low - .000000000931... As Lyte points out, streaks do happen in a random system. But Riot has the statistics, it should be fairly easy to find out if streaks occur more or less often then they should. (I'm pretty certain they will find that they do if they look at it, because of flaws in the matchmaking system)


Comment below rating threshold, click here to show it.

Purgation

Senior Member

09-29-2012

Quote:
Originally Posted by Lyte View Post
We are actively working on fixing Ranked issues and player behavior issues! Regarding your comments on Ranked, I think there is a common misconception that some players are just not 'ready' for Ranked.

This goes against the very belief of competitive gaming: all players should be able to participate in Ranked Ladder if they want! However, as I have mentioned before, there are lots of things we still want to do to help transition players from Normal to Ranked Mode. But, I think restricting players from Ranked Mode is not the solution.
What you say makes sense, but it's a player experience issue.

As someone who is spending a lot of time living at a horrible ELO, I know that I am not the greatest player, and probably will not be...

But at the same time, I'm not anything like some of the people I'm matched with.. both my team and other teams - - ragers, afkers, people who don't even understand game basics (who are usually new to ranked and only are in an elo because they they haven't fallen enough yet).

I believe that ELO hell is real, not because I think I'm great... or even above average... but because of the huge skill disparities I see in many games. I am not looking for a fix because I think it will "prove" I'm a better player - - I'm looking for a fix because I'm tired of games where one player on either team (either because they are so unskilled OR too skilled) basically makes the rest of the fairly even match-ups completely irrelevant to the outcome of the game.


Comment below rating threshold, click here to show it.

Purgation

Senior Member

09-29-2012

Quote:
Originally Posted by Lyte View Post
Check a couple posts back, shifting the starting Elo actually doesn't fix any problems!
I disagree with your reasoning, and other game ladder systems have implemented such a shift for exactly the reasons you should consider it - - and it's not just provisional players, there are many related issues (including the issue raised elsewhere in this thread about players with different skill levels by champion or role played).

The problem it fixes is horribly mismatched games where a player with - - say 1000 ELO and an overall record of 0 - 20 in ranked is "matched" to "carry" a team of high 800 - low 900 ELOs against another team with a similar ELO that is far more stable.