Idea for Matchmaking Service

I think that Riot Games might be using the same algorithm for low-ELO as they use for high-ELO. It seems reasonable to me that in order to address both the issues of matchmaking for each competitive player-base each of those playerbases need to be thought about separately and distinctively from one another. I don't think in this case a single matchmaking service is the best route to doing this. I favor at least the idea that players at higher-ELO should be matched more simply and liberally against each other.

Currently in order for the service to allow players of low ELO to compete in teams of higher ELO it ranges over a set of players around the average of that skillset (likely taking into account the maximum and minimum ranges of each set of players for both teams), the current matchmaking system seems to try to bring the average skill of those players as close to each other as possible in the quickest possible time.

As put by a member of the Riot games staff high ELO players have to make a trade off for time and quality. Low ELO players don't seem to make this trade off because there isn't an abundance of players.

Now I think one common assumption made of how to build a good matchmaking service is this: all matchmaking services made good for one set of players must be good for all the players (in our case low and high ELO's).

In the case of League of Legends it may be alternatively best to have fundamentally two matchmaking search algorithms. There should be a cut off point, say Masters, where the algorithm for the matchmaking service which applied to low-ELO players no longer applies. I don't mean the system should become more strict. I mean fundamentally remove all low-ELO players from the equation have a separate service which scales only upwards from there. In addition this matchmaking service ought to be more loose (as confined to the high-ELO region).

I have one solution for this. First let me recap what was discussed by RiotGames in the video...

**Solution 1. Either broaden the matchmaking range or reduce the percentage chance of getting the position you want choosing (top, mid .etc.) for high-ELO players to significantly reduce slicing and thus que times. **

See video for discussion on this.

Solution 2. Swap the respective matchmaking range into sets (not ranges)

This is a bit speculative since I haven't worked with the current algorithm. This entails slicing up each ELO range such that any set of players above a certain MMR level can only play with each other. I'll try to explain why this might be faster than using MMR ranges..

Philosophically there is going to be a point where getting better at the game is insanely tedious and you can define two sets of players such that the differences between the two is minimal. To illustrate, players at top challenger in SoloQ aren't all that much better than players in mid-Challenger in SoloQ or even some masters players, and if they are we can widen the range until that's the case and then just semantically change things around such that the current Masters Rank becomes a Challenger V MMR range or something. Solving that problem could be done by broadening the range of ranks until you hit a point where you can get a single tier of matchmaking ranks and define that these players could play with each other.

Eventually this would mean matchmaking for high-ELO players would be separate such that you can reliably place together a larger environment of players (so more quickly and more efficiently as long as the population is good enough).

Of course doing so could alienate the matchmaking experience for the next band up: if the band down has a larger environment of players there will be a smaller set of the matchmaking band above it. To illustrate, in our current system Challenger players are the highest possible band, and by removing all masters players you then have a significantly long que time for Challenger players compared to the old system.

However since we have expanded the set of ranks after a time period (i.e. 10 minutes), the matchmaking environment for challenger players expands to the environment below it. This isn't quite the same as going procedural to the next range of MMR ranks (say from a range 2000-2100 to range 1999-2100). This is moving reliably between bands of MMR ranks (so larger sets of players, i.e. range 2000-2100 to ranges 1900-2000 and 2000-2100). We only do this once we have expanded the number of ranks to sufficiently ascertain there are less players in each band such that all players of each band could be matched fairly against each other.

TLDR; Querying for sets of players is faster than querying for players within MMR number ranges and this is the means by which the matchmaking system ought to be implemented to suit the needs of only high ELO players. We should allow players to play with each other more liberally in lower ques (to suit the needs and wants of the low ELO and more casual market) but have a fundamentally different matchmaking service for the high ELO players (to suit the needs and wants of the competitive market).

What do you think?