How is a game considered even by the matchmaker

brunoais·Oct 8, 2015·2 votes·380 views

I just had a game where I was playing in team builder as a Morgana support. During the whole laning phase, I noticed that the partner marksman was playing way worse than I play as with that champion.

When I checked the skill level of all players, I saw that the players on my team are all mid or low Silver and all players (1 exception) on the enemy team were gold level players. Their skill was following that rank, which is about as good as I play (I may be wrong but believe I play as good as a gold player in the support role). Why are clearly uneven teams matched together? How does teambuilder matchmaker does its calculations that leads to such discrepancy in skill between both teams? On a related note, why is that most games where I play are stomps? It is either my team or the other team stomping the other and ending the game too fast or deciding the game's outcome too early.

I'm not simply complaining. I really want to understand it!

Ask anything you feel that I didn't explain well or that I should mention to help us discuss and understand how all this develops.

6 Comments

tseiheiOct 11, 20153 votes

The matching and mmr are pretty much the same between game modes, but they are separate from each other. This means you can be at a silver level in ranked, but at what would be the equivalent level of gold in normals. Of course golds can also be at lower levels in normal, so it doesn't mean you should be a higher rank.

The reason for games that are stomps is simply that people have become better at playing over time, and thus are better at closing out games. I think it also has to do with that teams behind can't contest dragons.

brunoaisOct 18, 20151 votes

That's quite true! But I also believe that it is related to uneveness. By your explanation, sounds like TemBuilder still only takes normal's MMR into account and not the ability of that summonor on that champion performing that specific role in a certain lane.
I know it can be abused, specially in 5-premade but I think that, in order to keep games as even as they can be, it would be better if there was a grading somewhat like that. There should be a grade related to a pondered average of the sum of role + lane + champion to decide someone's MMR level when queuing on TeamBuilder. What do you think?

tseiheiOct 8, 20151 votes

How much lp on a win? How much lp on a loss?

if you are gaining more on a win then you lose( ie you win and get 30, but lose and only lose 9) then you are way ahead of your mmr, and the matchmaking is basically challenging you to try to push you up faster so you can basically climb faster to where you should be at.

if it is the other way around where you are losing more then you gain, then you may be at a higher division/rank then your mmr may reflect and therefore the system wants to push you down.

If you are around even, as in gaining 18, and losing 18, then it is likely that what was said above is basically happening to your teammates. Meaning that while they may be in a low elo, they may be better than what that elo may justify.

brunoaisOct 11, 20151 votes

As I wrote, this was a match using TeamBuilder :). I was hoping it would not use the same matchmaking rules as non-TeamBuilder normals or rankeds. That also means I have no idea how my mmr is varying. I've tried websites such as euw.op.gg but they still don't tell me my mmr.

BTW, I basically stopped playing rankeds with the intent of having as much even games as possible. I'm still trying to overcome the once per 20 matches average.

tseiheiOct 18, 20151 votes

Well there is this part of Teambuilder with new champions.

http://i.imgur.com/4dddtT8.png

But remember that there are a couple issues when it comes to matchmaking:

if you add to many things for the system to look for based on a players "skill", then it will increase, exponentially, the time it takes to find players. This means that higher mmr's who alerady have +10 min q times will have an extremely long wait.
What exactly are going to be the criteria for "skill", and how will the system be able to judge a persons? This can be extremely difficult due to everyone's perception of "skill" can be different, and it leads to some people, like you, who believe that it isn't a proper measurement of a players skill.

brunoaisOct 23, 20151 votes

I believe the ELO calculation is a good enough way to measure someone's skill in the conditions he was placed in.

As a "maybe solution" I believe having one ELO calculation on each of the main elements participating:

Position
Role
Champion
Intrinsic position advantage

With a weight of about (it needs testing), respectively:

towards the pondered average calculation, it should be able to do it right.

Think like this:
A jungler has way, way higher impact on the outcome of a game than a top laner, right? So a jungler it should rank high on those 10% and a top laner should rank low on those same 10%.
You may mention that a top laner has quite some impact on a game. That is true. That's why it is only 10% up there but even on professional play, the top laners are still the least impactful until the teams group up to get objectives. No matter how good you are as a top laner, a jungler that plays as well and has as much luck as such top laner will get better results.

What do you think? Do you have anything to add? Did I miss anything? Am I being too unrealistic?