How many games do you expect to play before you reach your true Elo?

< 10 Games 179 1.63%
11-40 Games 817 7.42%
41-80 Games 1,160 10.53%
81-120 Games 1,731 15.72%
121-160 Games 663 6.02%
161-200 Games 926 8.41%
200-399 Games 1,698 15.42%
400-599 Games 1,081 9.82%
600-799 Games 454 4.12%
800-999 Games 170 1.54%
1000+ Games 2,132 19.36%
Voters: 11011. You may not vote on this poll

Help Riot improve matchmaking! Looking for examples of bad matchmaking

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

Trundlin

Member

02-17-2012

Quote:
Originally Posted by Lyte View Post
My concern is if players know that Elo points are distributed onto AFK/leavers, that a portion of players will rage and be extremely toxic with the intention of making someone quit to take the brunt of the Elo loss.
I think this would happen too, but at the same time people would know this is the case and either not quit (pretty simple) or just have a really good reason to report someone, (would need to be able to report post game though).


Comment below rating threshold, click here to show it.

Ajacmac

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

Recruiter

02-17-2012

Quote:
Originally Posted by Lyte View Post
I hesitate to let people choose their 'desired' win%s. Explaining a system like that to a player is complicated and we risk overwhelming players with math when it does not necessarily add value to their experience.

Allowing people to flag their role, or the many posts suggesting 'group finders' are neat ideas; but very difficult to implement. This is not to say it is impossible, but it is an idea that we must tread carefully with and really consider the advantages and added value it provides our players.

One, we have to be careful we do not recommend or push metas or 'suggested team compositions' on our players.

Two, we have to be weary of some roles or lanes being less popular, and increasing queue lengths into unsatisfactory levels.

Three, we do not want to profile a champ as a 'jungler' or 'mage' necessarily. One of the great things about League is that you can build champs many different ways. Some players build AP Master Yi and that works. Some players build AD Twisted Fate, and some build AP Twisted Fate. We never want to stifle creativity. If we label the vast majority of champs as 'viable' for multiple roles... then I am not sure we are actually helping the situation all that much.

There are possible solutions to each and every issue; but, tackling all the problems elegantly is not easy!

I'll Just address each in order as best I can, and I apologize in advance for the lengthy post.

1
[CENTER]"Decide your Challenge!"[/CENTER]

"Easy: 3 wins for every 1 loss"
"Medium: 1 win for every 1 loss"
"Hard: 1 win for every 3 losses"

That takes the math right out of it, and you guys should have someone on staff who could word that more attractively.

2
This is far more tricky to implement, as you stated. Ideally it would need to be simple and fast to use, and yet still effective. I kind of picture a large popup window with 5 boxes, each labelled with a role that can be changed, and a silhouette of a stereotypical champion that would fill said role. The first box is highlighted to indicate that it represents you, and click the label for a dropdown menu (or something more interesting) for selecting roles. Each box could also have an "alternate choice" directly below it, with a much smaller silhouette. The box representing the interacting player could have 2 alternatives instead of one for the sake of flexibility in the MM pool. Finally a little button to start the process.

MM starts looking to fill the exact comp.
After X seconds widen search to any comp sharing 4 of the ideal roles. (indication given)
After Y seconds further widen results to any comp sharing 3 of the roles. (same)
etc.
etc.

Once you get matched up you would receive an indication of what role MM slotted you in with.

This method pushes no meta. The window could even open with every role being blank. X and Y could also be as short as 10 seconds if deemed necessary, but every time the search would be widened the available player pool would grow exponentially.

A big, fat "Any" button for composition and/or role would also be a good idea.

3
The champion itself doesn't have to be the deciding factor per se. For non-junglers it could be as simple as checking what items the player ended the game with. Identifying junglers would be trickier, as simply checking whether or not the person used smite isn't 100% accurate by any stretch. I have no idea whether or not Riot gather statistics of what items were purchased over the course of a game.

I just had an idea, albeit one that might be remarkably less feasible than any I've seen posited thus far (depending on how you guys determine whether or not someone in game is afk), but I'll put it forward anyway. If a champions location on the map while in game could be tracked, it would make identifying junglers extremely easy. The game could simply divide the map into a set of coordinates, and label sections as jungle, river, lanes, base, or it could even be detailed down to separate minion camps and bushes if it was determined it would be useful. If a player spent the first so many minutes mostly in the jungle, he or she was probably jungling.

Tracking individual minion kills would also do the job quite simply.
Player 1:
lg wolf - 1:48
Sm wolf - 1:51
Sm wolf - 1:55
Bl golem - 2:15
etc.

Alternatively, you could just chalk up all minion/monster kills simply as lane/neutral and give each a time stamp.

After the algorithm does it's work the game could just be given a label as mage, jungle bruiser, etc.

I think that's everything, and I hope something I said will be of benefit, if indirectly.


Comment below rating threshold, click here to show it.

IS186afbf367d955

Senior Member

02-17-2012

Lyte : Once again please do some survey about game length vs fair game. From all the things I've read in the redtracker this looks like the simplest but still effective potential fix.


Comment below rating threshold, click here to show it.

Kenpachiiiiiii

Senior Member

02-17-2012

there should be an option for OVER 9000!!!!


Comment below rating threshold, click here to show it.

dKtiGERR

Senior Member

02-17-2012

Finally, a solution to this crappy system.
Ten ranked losses in a row shouldn't be happening.


Comment below rating threshold, click here to show it.

SubDraco

Senior Member

02-17-2012

Quote:
Originally Posted by Lyte View Post
Streaking is an interesting phenomenon. For example, everyone considers a coin toss as an event that should result in 50% heads and 50% tails over a large number of events; however, in the short-term, if all 30+ million League of Legends players started flipping coins, lots of individual people will see streaks of 5, 6, 7, even 8 or more of heads or tails!

What I am trying to illustrate here is that even if matchmaking was perfect (and I don't think any matchmaking system can truly be perfect), it will still have streaks and lots of them. However, perhaps we can think of creative ways that will help reduce a players' chance of getting 'extreme' losing streaks that are emotionally demoralizing.
flip flip flip winning streak im awesome at this game


Comment below rating threshold, click here to show it.

Razy4evar

Member

02-17-2012

No, matchmaking isnt fine, at least not in the early levels.

i hope the OP reads this post, here im gonna explain what breaks the matchmaking.

Low level + high w/l or normal elo = get paired with full level 30s.

Even against known players...

not to mention the 8 to 20 minutes queues, at level 21 (but it starts happening around level 12).

and this is solo que too. then you find out that everyone else in your team, even those 30s are solo too... so why did those levels 30 get paired with a level 21 out of the blue? something is wrong.

those 30s usually have 1k+ wins.

im tired of facing people with full runes and masteries that easily have 20x my amount of wins. last time i checked my normal elo with Lolstatistics.com (i keep track since level 1) it was above 1800, and my w/l is only 55-28.

i thought months ago matchmaking was tweaked so it would NOT put <30 players against other 30s.


Comment below rating threshold, click here to show it.

blogg

Senior Member

02-17-2012

As long as everyone gets the same amount of Elo or loses same amount of Elo (except people that are still in placement matches) in case of a win or loss the system will be flawed.

Even an imperfect system that tries to quantify the impact of each player in a game is better than the current one

I know it opens the potential for a lot of QQ but will help a lot to better differentiate between players (especially in the middle ranks)

My 2 c


Comment below rating threshold, click here to show it.

3v3ryone

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

Senior Member

02-17-2012

Can you add some stats that are hard to abuse but in the same time improve team cooperation?

for example option to match player "A" with players who buy similar amount of wards within certain time.

How to prevent abuse where ppl would buy "X" wards at the end just for the score? make it so that you get invisible points for placed ward but only once per 3min, with the limit of 1 point per ward.

so if game would last let say 30 min you could get max bonus 10 invisible points if you'd place 1 ward every 3min.(doesnt matter if enemy destroyed ur wards after you placed them) This would separate those who understand what map control is from those who doesnt and it would prevent abuse in the same time.


Comment below rating threshold, click here to show it.

Whambamkablaam

Senior Member

02-17-2012

An issue I have with the ranked game system more than dodging or elo penalties, is the concept of having your first 20 games or so weighted so that if you happen to be unlucky and have a 30 or 40% percent win ratio, at 300 games you'll have to have far more wins than losses to make up for your mistakes early in your ranked career. It completely alienates players who genuinely experience an increase in their own skill as players.

Because the preseason is more important than your last game as a competitive person?

Improvement should be rewarded and not discounted.

If anything, I would suggest a system of measuring skill in the same way many performance tests are designed. For instance, a test might have questions increasing in difficulty through out the test, with the last question being the "hardest." Placement would be more accurate in my opinion if you have the chance to play 5 games at several different elo ranges, with elo ranges increasing every set of games. This way you might actually be placed at a respectable range to reflect your skill as a player. (imo) Obviously a true "1500" player would "carry" 900 elo games but experience significantly more difficulty in carrying 1400-1600 games.

This way people don't end up squating on 1500 elo with 6 games won. That is simply outrageous.

Eh, probably better in theory. but w/e, my opinion.