I played a few solo 5v5 ranked games today. That’s been my method of late. A few. Not four. Sometimes not even three. Just one or two in a go, and they rarely last longer than 30 minutes anymore. It’s a strange way to play the game, kinda like taking a step back more than six months to when my ELO was fresh and I was just learning to play.

The one game that stuck out today, though, was a game in which the other team didn’t ban Shaco and I got to have a little fun. It started in their woods when I polished off Warwick at level two. I ganked him again as soon as he respawned and made it out of base, netting him three deaths in the first six or seven minutes of the game. My opponent’s other lanes were struggling as well – one because Sion was incredibly over-confident. I wondered what was going on (why we were winning so easily) until I got to the lobby. That Sion? He had one win. One solitary win, and his ELO was more than 150 points below most of my team.

I will say, most of my games have been with players of similar ELO, with a tolerance of maybe 100 points in the worst case. I can live with that. But when my queues are no longer than 25 seconds, it seems like the MM algorithm should be doing a better job than this. As more games are played, there will be an increasing gap between 1200 ELO players and 1450 ELO players. Why matchmaking would ever put those two together, particularly in solo queue, is completely beyond me.

Matchmaking has long been one of the hottest topics of discussion around LoL. For most players I’ve talked to, the system is hit or miss. You’re either paired with a team that does reasonably well usually against a team that’s not so good, or you get the reverse, stuck on a team that has some less than skilled players playing against people that know what they’re doing. In rare (sometimes even not so rare) circumstances, you’ll matched with someone who has no business playing in your games, someone like the level 3 I was matched with just a few days ago.

The Garen patch brought with it a bunch of changes to the matchmaking system, including some optimization for level balance alongside the current ELO balance. There were also some changes made so that 5-man premades would be placed against other full premades more often. While all of this is good, I have to ask, now? We’re just now getting these changes and also getting word of them? Why hasn’t this been a part of the forum discussions for months. No doubt one of the highest contributing factors in the number of forum posts lambasting the matchmaking system is the lack of a Riot response. There was very little indication that these things were being worked on, and the general sense was that it just wouldn’t be fixed.

Now Zileas has stepped forward requesting feedback on the new changes, and he uses language that I think points at some of the design attitude around matchmaking. Take a look at his last point: “4) Any other weirdness that is obviously very bad, not just subjective “this guy really sucked’ type stories.” While a player’s assessment of another’s skill is subjective, there is also some empirical data we can look to for determining whether matchmaking is doing its job. A while back I cited a player who had a significant number of losses in his last 10 games, nearly all of which included stats to support the theory that he’s not a very good player. I’d hardly call that amount of data subjective, and it took me a total of three minutes to discover without any analysis tools.

The bottom line here is this – if you want a more accurate ELO, you need to find four friends you believe to be of similar skill level and premade, premade, premade. Riot’s matchmaking system will never be able to account for individual skill in the midst of unskilled teammates unless it moves to some sort of performance-based system, which is unlikely at best. Until then, its probably best to just keep quiet and enjoy the five to ten percent of your games that turn out to be a decent match.