[mattmaroon]

So after playing LoL for awhile, which has a much sloppier matchup algo (it's harder to do accurately in 5 v 5 than 1 v 1) I have to tell you, better is better. The intensity of matches is part of what's made me switch to SC2.

The problem with the sloppy algo is that it's rarely fun. 45% of the time you're crushing the other team so easily that it's kinda boring. 45% of the time you have 0 shot of winning, and know this in a few minutes but it takes a half hour or more. 10% of the time it's close and then it's actually fun.

[mawhidby]

I agree completely regarding LoL. Their matchmaking system is supposed to give you a 50% chance of winning any given match (according to their wiki). I usually pre-made with 2-4 friends, and we almost always lost. We were low levels (10-15) being matched up against five level 30's. Needless to say, we were getting crushed every time. I think I had a 15 game losing streak or something when I pre-made.

Their wiki says they make the matches harder when you pre-make with friends because they figure that you will be harder to beat. I got so tired of the uneven matchmaking that I can't bring myself to play it anymore.

[mattmaroon]

They do. They add 120 to each player's ELO when queuing as a premade. The actual advantage you get from playing together ranges from much more than that (with a full team of players who coordinate well) to much less (a 2 man and a 3 man combined who don't coordinate well). We win almost every 5 man we play because we coordinate well, and none of us are that great individually so when we solo play our ELO drops.

Also, as I told Tom "Zileas" Cadwell, you can easily have a matchmaking algorithm that ensures you win 50% of the time but is not fun. If 50% of the time you're playing against hopeless newbs, and 50% of the time you're playing against the top guys, you'll win half but have fun on 0. LoL is not far from this.

LoL's matchmaking has 3 main problems that I can tell.

1. Radically different advantages to premades. Some premades get much more benefit than others. Starcraft 2 takes this into account by giving each unique premade its own rating. So if I play a 2v2 with my friend John, he and I will get a unique rating that doesn't affect matchups when I play 2v2s with my friend Chad. I'm not sure if this is an option for LoL given that it's often 5v5 (might work better on Treeline, which is 3v3) and a team is often comprised of one small premade and a few randoms, or 2 smaller premades.

2. Traction. They don't have enough people in the queue at any given point to create good matchups in a reasonable timeframe.

3. Elo. Elo ratings were meant for deterministic heads up matches. In chess, Elo works because the player who won simply played better.

LoL tries to compensate by determining why your team won or lost (and what part of that was your skill or lack thereof) when handing out the points. This is VERY hard to do well.

[mawhidby]

I understand the philosophy behind adding to a player's ELO when doing a premade. But I feel that part of the problem is that relatively new and inexperienced players (which I consider myself and the people I premade with) are going to be matched with higher ranked players. This makes it really hard and discouraging for you to play and get better when you are constantly losing to better players.

Personally, I enjoy playing the game with friends, it's more fun that way ... just not so much when you are losing badly every time.

[roc]

> I'm not sure if this is an option for LoL given that it's often 5v5 ... and a team is often comprised of one small premade and a few randoms, or 2 smaller premades.

Wouldn't the matching algorithm be calculating ratings for each distinct premade, regardless of whether it's a complete team?

So whether [You+Chad] are playing 3v3 with a pug, or 5v5 with any variant of three PUG+pre-made players, shouldn't you always be working on the [You+Chad] rating?

[cynicalkane]

Citation needed. I read that they changed premade ELO adjustments to be negligible for bad players and high for good players. This was like a year ago or around that.

[mattmaroon]

Well, my citation is an email sent to me by Tom "Zileas" Cadwell. I've talked to the Riot guys quite a bit, they actually play our Facebook game. I even got to check my elo :)

He didn't get into too great detail though so it's possible I misunderstood and it actually is a sliding scale, though I've seen no evidence of that (and quite a bit of the opposite).

[cynicalkane]

Zileas is a pretty smart dude, and they've been working on this for a while. I can't find the post I was talking about, but this one hints at the adjustments: https://www.leagueoflegends.com/board/showthread.php?t=12029

"This bonus is calculated from a bunch of research we did on hundreds of thousands of game results to figure out how much of an advantage being in a team is. We do some behind the scenes adjustments as well for stuff like beginners paired with pros, etc."

Some opaque, out of context statistical data: http://www.leagueoflegends.com/board/showthread.php?t=16705

Personally, the matchmaking seems to me to get better and better the more players pick up the game. I still see unbalanced games, but I can't remember the last legitimate curb stomp, where one team was simply outclassed...