TL;DR: SEE THIS. If matching and skill estimation are decoupled from the playlist and players are allowed to search multiple gametypes, then truly custom searches become possible even with low population.
Friendly request: If you think this is a good idea, consider leaving a response (however short) to keep the thread alive. The longer it stays alive, the more likely it is that someone important may see it.
With the MP for Guardians billed as “next-generation arena”, perhaps 343i ought to consider making the matchmaking next generation to suit. Specifically, the playlist-style matchmaking system used by previous games and the MCC is antiquated and unnecessary. It has the following downsides:
The matching pool is restricted to individual playlists, reducing the number of matching opportunities.
To keep matching pools large enough, playlists must contain multiple gametypes that may not appeal to all players searching in the list.
Multiple gametypes within a list are the reason behind the suboptimal voting method in H4 and MCC, and contributes to pre-game quits if the desired map/gametype combination is not selected.
The above contributes to either uneven teams, restart of the search and matching process, or the need for JIP.
Gametypes that are not popular enough to sustain their own lists are either eliminated or made rotational, which limits the opportunities for those who wish to play them.
Matching is not done on a first-in, first-out basis, leading to a large number of players tied up in partial lobbies which both extends search times and reduces the quality of matching.
However, there is a way to eliminate or minimize all of the above. Let’s investigate.
Rather than restricting players to playlists, players should be allowed to customize their search in whatever way they want. Using the MCC as an example, why could I not search like this? For people who just want the ease that standard playlists provide, well, that is present. For those who wish a bit more control over what they play, that is present as well.
When performing matching, an Xbox Live server pools all players by gametype (with “CE One-Flag” and “CE Multi-Flag” being treated as separate gametypes, for example). Per the image above, players can be present in multiple pools. Then the following process is repeated:
The best possible match is calculated for the player that has been searching for the longest time across all gametypes based on whatever factors 343i wants to use for weighting (which will most likely vary based on search time).
If no good match can be found, the system moves to the next oldest player.
If the search time for the oldest player is sufficiently long, that player can begin “reserving” other players that meet match criteria (but not otherwise).
Once a good match is found, connect those players’ consoles and place them in a lobby.
Return to step 1.
The reason for matching as described is to prevent partial lobbies. Any time a partial lobby forms, players are removed from the matching pool for all other lobbies. This extends search times for everyone. If the first-in, first-out method is used (with reasonable exceptions), the only players that are stuck in lobbies are ones in full lobbies.
Since everyone has an opportunity to search only for the gametypes they want to play, lobbies can be locked immediately upon formation. There is no longer any need to allow people to quit a lobby. Moreover, there is no longer any need for people to vote on gametype. The gametype can be system selected upon lobby formation since only people that searched for that gametype will be included in the lobby. Once in the lobby, the only “vote” required can either be FOR a map choice or VETOING a proposed map (after which the system automatically selects an alternate map).
In terms of skill ranking / matching, this can be done by grouping. For example, the skill estimate for group “CTF” could include performance in gametypes CE 1-Flag, CE Multiflag, H2 1-Flag, etc. The “CTF” group would be used for the skill portion of the matching for each of those gametypes. The benefit of decoupling skill from the playlist is that playlists can be freely altered without affecting anyone’s skill estimate or rank.
Since most players will search multiple gametypes at once, the effective matching pool for all gametypes increases. Right now, 4v4 Slayer, FFA, SWAT, Snipers, and Big Team Slayer all have separate matching pools. A player cannot appear in more than one pool. However, many players (like me) would search all of those lists simultaneously if given the option to do so. That greatly increases the number of players for matching for all gametypes.
Take H4 as an example. When Legendary Slayer was implemented, I could only search in either Team Slayer or Legendary. Because Legendary was less popular, the matches were less even and, hence, less fun. So in H4 we had (example):
Team Slayer: 4,500 player pool
Legendary: 800 player pool
However, many of those Team Slayer players might have wanted to play Legendary, but were avoiding it because of the matching difficulty. Let’s just suppose that a mere 15% of the Team Slayer searchers fit into that category. Let’s also assume that 25% of the Legendary players would likewise have been happy with a Team Slayer match. With the method described here, the effective pools become:
Team Slayer: 4,900 player pool
Legendary: 1,475 player pool
Unlike the traditional playlist method - which ties both skill and matching pools to the playlist content - this method (a) matches by gametype, (b) calculates / matches skill by group, and (c) utilizes playlists only to provide pre-set search parameters.
Matching like this would yield the following benefits:
Players would always get the gametype they want.
The need for eliminating gametypes or rotating playlists disappears.
The effective matching pool increases for ALL playlists, leading to faster and more even matches.
Lobbies can be locked, reducing lobby quits and the need for restarting searches or JIP.
Even players searching for games with miniscule populations can eventually find a game without diluting the player pools from other playlists.
Altering, reorganizing, and regrouping games in playlists has no impact on search pools or skill / rank.
The reason I chose MCC for the menu example is that it is far more complex and contains far more gametypes than any standalone game. I believe the generic format presented is simple, intuitive, and easy to navigate*.* And if it can be made that way for the MCC, it can most certainly be made to work in a standalone game.
I hope 343i considers something like this. It’s about time to move the matchmaking process forward.