I've been interested in the matchmaking system ever since Zileas started talking about what Riot was doing back in Beta. Couple of thoughts on the discussion so far,
1) AFKs in Champion Select Lobby
This came up in the discussion already, but in regards to dodging AFK’s vs. randoming, I thought the system was intended to always dodge someone unless there’s a bug? If there is some sort of exception, doesn’t seem like it should be there.
Given what’s required for this to happen, (manually queue up the game and be AFK 1-2 minutes later.) Unless you’re at high ELO, the level of derp that causes champ select AFKs probably deserves a higher punishment.
In terms of the Dodge to -ELO point that's been floating around. I'd be worried what changing that system would do. The reason it was put into place in the beginning was because the level of queue dodging was so high and it disrupted matchmaker mechanics. Basically reducing the incentive not to dodge because a few players lower their ELO for fun doesn't seem like the best idea. A better idea might be to permenantly remove Gold and Platinum badges for players who do this
2) Duo-Queue Elo Disparities in Ranked
I completely understand where you’re coming from not wanting to restrict friends playing with friends, but this doesn’t quite mesh with focusing ranked on “bring your best, every single game.” Going back to your basketball analogy, a middle schooler might be great friends with Derrick Rose, but that doesn’t mean they’ll let him play in the NBA.
3) Skilled Ranked Players in Normal Modes
Normalizing sounds like a win
4) Premade Matching
Especially given the data you’re mentioning that talks about premades having either benefits or disparities based on their ELO shift, it seems like there might need to be a multi-part piece to this. Both matching pre-mades to pre-mades and changing team average ELO bonuses for premades depending on the ELO gap between those players.
5) Transitioning from Normal to Ranked Mode
Previously when we talked about Normal Elo affecting starting Ranked Elo, a lot of players voiced complaints that they play the different modes for different reasons and different intentions. So players use Normal Mode to test champions, try out fun builds, or mess around and they did not want this to affect their starting Ranked Elo.
I think this is the right idea just the wrong way of looking at how to implement this system. You don't need to permanently link Normal and Ranked ELO here. If you think about a player getting to level 30 through normal games, their ELO is more a picture of their skill level and evolution as a player rather than individual games where they tried something new. EVEN if they screwed around in some percentage of their games, their normal ELO whether it’s 400 or 2,000 is still going to be a way better indicator of where they should be initially seeded in ranked then just lumping everyone into a 1200 group.
For any player already in ranked, ELO’s are already established, new season ELOs are based off of soft resets, so any games they played in Normal wouldn’t affect their initial ELO seed.
Given how separate Ranked play is from Normal play, it would probably also be worthwhile to have a newbie island repeated at this level. It’s not a great analogy, but harkening back to DCUO, the game is basically level 1-30 are training, and then the disparity between new level 30’s and high level 30’s is night and day. A system that keeps these “new” players together and lets them fight it out would help to better adjust initial skill levels without throwing them in with 300+ game winners, especially if you’re not using normal game ELO seeding.
6) Free to Play Champions in Ranked Mode
Related to #5, I think eliminating these champions would be a great addition to ranked play. It encourages players not to try unknown, ELO skewing champions in “bring your best” games and also guarantees players will have at least played enough to earn the number of champions required to participate in Ranked.
7) Random Champions in Ranked Mode
Same as 6, education and options are two different things. It’s great to be able to educate someone on “this is what you should do for ranked.” But sports have rules and these are easy ways to visually and literally enforce more skilled game play. There’s no point in saying “bring your A game every time - don’t random, and oh by the way, we’re still going to let you random.”
8) Provisional Matches in Ranked
Addressed in #5, let new players play on Ranked Newbie Island. Normal ELO seeding would also help address this, starting closer to where you should be and moving around will put you closer to the right place no matter how your initial games worked out.
I suppose an alternative here could be to start everyone at 1200 ELO but give players different modified K-factors for wins and losses based on their Normal ELO.
9) Duo Queue Prevalence in Ranked
If the data says so, it might be worth a shot, I’d be more worried about what that would do to match finding times.
10) Level Disparities
Given the data we’ve seen all over the place, every one of these complaints has seemed more like “OMG I lost 1 game, time to shift blame” rather than any substantive issue with matchmaker. The system already gives slight ELO adjustments based on level, it seems like that works just fine. THIS is more of an education issue. We still see a lot of people who don’t understand MM in terms of win/loss ratio, level and ELO.
11) Team Margin of Victory
Dominion is easy, ending points :P
For Summoner’s Rift, I think the biggest single identifying factor is time. While certainly a team could “play” with an enemy just to farm out an “epic win” there’s a couple things that make this ok:
- Epic wins tend to come from the fact that “winning” teams don’t finish fast enough and give their opponents too much time to farm and turn the game around for a comeback.
- Even if this happens, both teams get an “epic game” bonus making it more worthwhile.