Welcome to the Forum Archive!

Years of conversation fill a ton of digital pages, and we've kept all of it accessible to browse or copy over. Whether you're looking for reveal articles for older champions, or the first time that Rammus rolled into an "OK" thread, or anything in between, you can find it here. When you're finished, check out the boards to join in the latest League of Legends discussions.

GO TO BOARDS


LOL Matchmaking Explained

Comment below rating threshold, click here to show it.

Amazing MuhKuh

Senior Member

09-16-2009

Well the math behind the system seems to be realy intresting. Another question that came up for me after a quick look over the wikipedia entry is this:

are you assuming that the variance of playskill from match to match will be the same for every player, or do you plan do monitor the variance of a player for giving them a personal k-factor?
(again reffering to the notation from the wikipedia article). Notation is a nightmare


Comment below rating threshold, click here to show it.

Savage

Senior Member

09-16-2009

Quote:
Zileas:
We try to make the teams as close to 50/50 as we can.


Why was that chosen as the goal? Why is it desirable to force things to be closer to 50/50 for solo queues than, well, the natural deviation from 50/50 for 5v5 queues?

I'm not trying to be a jackass in asking this; I think you've added some mechanics with very negative consequences and I'm really not sure what you're gaining in exchange.

Especially in light of:


Quote:
Zileas:
We approximate, and close to 50% is good enough for us. We are not 1 on 1 matching, we are 5 on 5 matching, so we get good results most of the time.


Well you get the best results you can, which is all kinds of difficult with everything mashed together - you don't just have weaker players having their ratings carried upwards by playing in stronger premades, but you have different skill sets being relevant at different levels of play, at different levels of organization. You have uncertainty in each person's rating from shot noise.

In short, while aggregated data gives you useful trend information, individual data points have very low precision, and at some point you're trying to add more precision to your matchmaking system than there is in the data points you have to work with.


Quote:
Zileas:
So uh, are you a student or professor at the university in your town, or do you work around there?


No, though I do enjoy college towns. Business is a much more fun game than academia. *grins*


Comment below rating threshold, click here to show it.

FellowerOfOdin

Member

09-16-2009

I READ THE WHOLE POST GIVE ME GOOOOOOOOOOOOOOOOOLD!

Ok, seriously now. First of all, thanks for the huge post and, of course, all the work behind it. As stated in my very own thread, I highly appreciate seeing a gaming company who cares for their community. The approach seems to be nice, but due to the lack of numbers, it's really hard to predict whether it will work nor not, guess we have to trust you here.

Concerning Pre-Mades, I still do not like the idea of allowing them to play against random teams and I do not know how you can precisely include coordinated teamplay and a synergetic hero / summoner skill choice into a matchmaking formula since those advantages are what screw a random team when they face a Pre-Made.

Furthermore, people in a Pre-Made will never leave, a random teammember might do so when being faced with a Pre-Made, leaving his allies to a guaranteed defeat that lowers their ELO.

Quote:
Q: So, because I won a few games in a row, I’m going to get an impossible match now, right?
A: Not exactly. Your rating will rise, so you’ll get harder and harder opponents – but we don’t really care if you go 50/50 win/loss, just that you rating seems to be an accurate prediction of game results. Eventually, you will hit your limit, and you WILL see roughly 50/50 win loss though. Players who are above average will tend to do slightly better than 50/50 because there are more players below them than above them, so matches, when made, will tend to be slightly “downwards”. For expert players near the top of our rankings, they will often run 90% win rates.


This is not completely valid Zileas. As you get a higher ELO, you will face harder teams, yes, but you will also get worse teammembers. It happens very often when I win a few games in a row. I win 3-4 games and then, pretty often, I face a (partially) Pre-Made and get some really bad teammembers who usually start feeding because their enemies are at a much higher level and sweep the ground with them. In 70-80% of the time, I end up losing a game, still getting 120+ IP with a score of +5 to +10. It's a very frustrating experience when you play at your best and still lose because of your allies.

Again, I would really like to have the option to join a game or not Zileas, e.g:

a) When the matchmaker finds 10 people who could play together, it shows you a pop-up window, telling you the difference to your ELO compared to the other teammember's average ELO, asking you whether you want to join the game or not and in case that you don't, it lets you queue up again.

b) The pop-up shows you the ELO difference between your and the enemy team and lets you choose whether you want to join a game or not and in case that you don't, you get queued up again.

I am aware of of the fact that people might tend to wait until they get a game where the difference between the ELOs is as low as possible, but those people might have to face 30+ minutes queues and if someone is willing to wait 30 minutes just to get a fair game, let him do so.

As of now, I am excited to see that you are working on one of LoL's biggest flaws, but I still have doubts concerning the system, let's see what future may bring.


Comment below rating threshold, click here to show it.

Ambi

Member

09-16-2009

I said this earlier but it was kinda ignored, I played a few more games today and got the same result, I'm on a positive win-loss ratio but I'm still being teamed with lvl 1-10s. How does this fit into the matchmaking system explained here?


Comment below rating threshold, click here to show it.

Dyrus

Senior Recruiter

09-16-2009

Well that helped me confirmed how matchmaking worked. I figured it would be that way seeing how I get paired with near to godlike players now days.


Comment below rating threshold, click here to show it.

Threads

Adjudicator

09-16-2009

so do I have a high ELO?
I got matched up with people who were between levels 18 and 25 and basically fed to our loss against a premade of level 30s.


Comment below rating threshold, click here to show it.

Vindico Atrum

Member

09-16-2009

Really, everyone should have figured it out. Its not hard, anyone with half a brain cell can figure out that matchmaking tries to match you with people at your skill level.


Comment below rating threshold, click here to show it.

Sayan

Recruiter

09-16-2009

Quote:
Ambi:
I said this earlier but it was kinda ignored, I played a few more games today and got the same result, I'm on a positive win-loss ratio but I'm still being teamed with lvl 1-10s. How does this fit into the matchmaking system explained here?

You might have a really bad ELO possibly maybe?!? Hence you might be getting into low level teams that have a similar ELO...


Comment below rating threshold, click here to show it.

Adonis

Recruiter

09-16-2009

Quote:
Match making works for me and that is all I care about, I find myself wining more games when I solo que though...

Zileas, you can make the most mathematically brilliant and accurate matchmaking system known to man. You could then come on the forums and explain it with such lucid clarity that even someone who was a few peas short of a casserole could easily comprehend it.

But it's just going to come down to this: if people are winning their matchmaking games, then they are going to think matchmaking is working. If they are losing their matchmaking games they are going to come on here and say it isn't working.

btw, what does 1844 even mean for elo? I know you start at lik 1200, but we've already got people well above 2,000. Is mid 1800s top 10%?


Comment below rating threshold, click here to show it.

Ambi

Member

09-16-2009

Quote:
Sayan:
You might have a really bad ELO possibly maybe?!? Hence you might be getting into low level teams that have a similar ELO...


If after 60 wins out of 110 games I still have a low enough ELO to play with level 10s then thats a bit... off?