Matchmaking Changes [1-16-2013]

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

Ryios

Senior Member

01-24-2013

I say it's broke. Lots of people are getting rofl stomp matches where they win easily. But I'm always on the receiving end of that, I've lost 10 out of 11 now.

Every game I'm in, my team picks the worst comps, gets duo top with no jungle, etc. The enemy team, as if they are all on voice chat, synchronized, wards everywhere, destroying us around every turn.

Before this change I got decent teams almost all the time.


Comment below rating threshold, click here to show it.

Goodguy Hopper

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

Senior Member

01-24-2013

@RiotKidd and Lyte

First off guys thank you for posting this and letting us know what is going on. That kind of free flow of information really helps make me and probably several other players feel included in the community.

I was going to sign in and play league this evening after a little hiatus for the start of the semester. I noticed this thread, and I flipped through it. I was unclear on one major point, is this system already in place for ranked play?

I assume from a couple of your guy's posts that it is, but it wasn't ever specifically said.

My second question would be are you already confident enough in this system that you believe it will only have a positive impact if I get on to play some ranked this evening? I am aware of course that you can't make me promises, I just want your take on how things are progressing with this adjustment.

My concerns about this new system are mostly the implementation. I know you guys are pros and if you have some time to spare would you mind alleviating my concerns by answers this handful of questions.

Do you think it was wise to implement this change in both ranked and normals at the same time?
Do you think that in the future we will see changes occur to normals only as a "safer" check before rolling out the full system?
Was this implemented on the PTR before going live?

Sorry for all the questions the method of rolling this out made me raise my eyebrow even if I am fond of the concept.

My best,
Aly

PS-

Quote:
Originally Posted by Lyte View Post
hold me accountable for change
You pass your ~1 year review, are they calling you the golden boy around the office yet?

For my use later after some games this evening - (please ignore now) -
Quote:
Originally Posted by RiotKiddington View Post
Lyte and I want to know is this, the longer the queue time the better "match result" you will get, but it comes with a cost (of time). Is the current 30 seconds increase in queue time worth the benefit of matchmaking with additional consideration of # of wins?
Quote:
Originally Posted by Lyte View Post
Nope, we only tested it for a few days before the Winter break. It's unlikely the testing affected that player's experiences. This is only the 2nd time we've turned the feature on.
Quote:
Originally Posted by Lyte View Post
The new matchmaking feature does not only use Number of Wins to match players. Number of Wins is now just another parameter (among the many!) that the matchmaker will consider.

When we initially considered this feature, we also believed that skill ratings like Elo were accurate enough that we didn't need Number of Wins as an additional parameter; after all, players typically reach their true skill rating brackets within 150 games. However, when we dove into the research we found some surprising results. For example, some players choose to level up through Co-op vs AI games. Once they hit Level 30, they jump into Normals and get absolutely crushed by players who played Normals to hit Level 30. Because matchmaking now takes into account Number of Wins, these Co-op vs AI players will no longer start their journey into Normals against players with over 1000+ wins in Normals already.

We also found that Number of Wins does make a difference in Ranked Mode--especially at the lower wins bracket. Assuming equal Elos, there might not be a difference between players with 1000 Ranked Wins versus 1200 Ranked Wins, but we definitely saw some lopsided games with players with 1 Ranked Win versus 100 Ranked Wins.

When it comes to matchmaking, there are many dimensions that reflect a player's skill in League of Legends. We could argue that Elo represents the 'mechanical skill' of a player; for example, if a 1400 Elo player faced off against a 1500 Elo player with the exact same runes, masteries and champions, the 1500 Elo player should win a majority of the match-ups.

However, this doesn't necessarily capture the 'experience' of a player. In League of Legends, more experienced players tend to have better rune selections and a larger champion stable for complex counterplay. We find that adding Number of Wins to the matchmaker does a decent job of accounting for 'experience,' so we're testing this feature to determine if the increased queue times are worth the potential accuracy improvements in the matchmaker.

This is the kind of feature that may not actually improve the matchmaker for other games, but in League of Legends where players can build out their collection of runes and champions... it just might work.


Comment below rating threshold, click here to show it.

Décéptìøn

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

Senior Member

01-24-2013

Guiz, um, the patch is still in PBE. wow. Its not even released yet. They didnt tell you when they were going to relrease it but is believed its on the very next patch. Please stop acting like its been patches already.


Comment below rating threshold, click here to show it.

Ryios

Senior Member

01-24-2013

This was in the last patch, 2 ago now I think. If you read back through the riot posts this change is live already.


Comment below rating threshold, click here to show it.

KobeMonster

Senior Member

01-24-2013

Quote:
Originally Posted by Gizmo Era View Post
I'm still on a fantastic "losing streak," it's not probability

I'm still being matched with silver/gold players in normal when my ranked elo was around 1000. I have literally lost the past 30 matches. Even if the match was won, it's because someone carried me; I have lost almost every lane. This is not fun. Please change the matching back to whatever the hell it was before.

Every match I'm a feeding, game-throwing, f***ing-uninstall-the-game troll. I have heard this the almost every match. I'm not being matched with the proper ranking.

PLEASE FIX.

I don't understand why normal has to rank people behind the scenes... isn't that why people play ranked?

I still have not been in a match where the teams felt even since this change, where they both had a roughly equal chance to win. Every time it's easy to see by the 10 minute mark which team is going to win.
How can you possibly be that bad? 30 games in a row...


Comment below rating threshold, click here to show it.

TheGodOfGaming

Member

01-24-2013

I'm confused... am i supposed to be paired up with a draven with 87 wins and go against a riven with 5k wins?

Also, 1600 elo game, 1300 elo jayce on my team.


Comment below rating threshold, click here to show it.

Forchark

Senior Member

01-24-2013

The matchmaking should also take into account (if it doesn't already):
Runes (how many/cost of per page)
Last 5 games kills/deaths
AFK status last 10 games
Won games in last 5-8 games
Disallow replaying the same people within the next 10 games (actually happens every so often that I play the same people again directly following a loss)

You need to look at individual skill versus being carried. The problem with just looking at won games is that you aren't rematching that player with those others that helped win. It isn't a 1v5 game. Above are those items that should help address the issue of individual skill.

And as addressed by KobeMonster... You cannot have a "poor quality player" be matched with another "poor quality player" that is purely in the same neighborhood level wise. It is inevitable that I have a good game, then a terrible game. Take into consideration reports for not cooperating with team and make them play less skilled people. Take into consideration the amount of times one wins with certain characters, etc.


Comment below rating threshold, click here to show it.

Forchark

Senior Member

01-24-2013

Quote:
Originally Posted by TheGodOfGaming View Post
I'm confused... am i supposed to be paired up with a draven with 87 wins and go against a riven with 5k wins?
Exactly my point. A few losses doesn't mean you are bad, and a few wins doesn't mean you are good.


Comment below rating threshold, click here to show it.

Fruitbatpoosnack

Senior Member

01-24-2013

This is nothing new....Am I the only one who notices that the top teams often have very similar Wins to Losses? They have an algorithm and have had one for ages, to work out averages in queues. If this is news to you then you haven't been paying attention.


Comment below rating threshold, click here to show it.

Sighs

Senior Member

01-24-2013

Skill rating -> elo -> doesn't factor in performance statistics but only result oriented approach which is also a function of luck.

Why am I seeing huge discrepancy in team performance? 9 kills to 30 deaths every game is not what a good skill based match making system would produce.

You guys never defend on blatant flawed logic of your system. Because it cant be defended.