Casual players like me

The matchmaking system should be similar to what FIFA Seasons has (I know my gaming taste is all over the place)

In that matchmaking system you have people of certain rating and their next match against someone is based on the most recent match.
If you win, you play against someone else who also won but is within that rating range.
If you lose you play against a loser of same rank.

I think thats a pretty good idea.

In terms of matchmaking, the most widely used metric for rating a player's performance is the Elo system. I'm not going to describe it in detail here, but it's used fairly extensively in chess, as well as a few other online games. The gist of it is that players are rated numerically based on how often they win or loose. If a highly rated player beats a lower rated player, their Elo will only raise by a small quantity, and the loosing player will only go down by a small quantity, to account for the expected difference in skill. However, if the Elo raitings for two players are closer together, or a lower rated player beats a higher rated one, then the changes can be more drastic.

0K applies this to online games fairly well, including a system which artificially drops the value for new players, so as to help match them against similarly skilled individuals. 0K fails on two main particulars:

1. The number of people playing 0K is (in my experience) not large enough to guarantee a good match whenever you want to play a game. This may have changed since the last time a played a game. Given the amount of interest that PA is receiving, I don't think this will be a problem.

2. Elo used to include the results of team play, as well as 1v1. Although Elo provided a useful metric for balancing teams (if occasionally a little eccentric), people used to become annoyed if their Elo dropped due to the actions of a team-mate who was phaffing about or griefing. There are ways around this, such as having a different rating for team games, or having only 1v1 be ranked.

There are a few other variants as well, such as TrueSkill, but I'm sure some appropriate system can be found. The key is artificially dropping the starting rating of new players, to ensure that they meet players who have a similar lack of skill, rather than putting them straight into the middle of the table (as would be the case normally).

That's not how trueskill works.

If you are really interested, I've wrote a lot of "literature" about it in for my FAF implementation.
It's been almost two years and it works almost flawlessly (the problem currently is that we have more noobs than experienced players, so it's hard to match top players against someone else).

http://www.faforever.com/mediawiki/inde ... ing_system

While there is some room for improvements (ie. the deviation should increase over time while not playing, simulated by adding some on every game for now), it's probably the best system you can have.

Not sure if I'm saying that in the wiki, but elo is a approximation of a what trueskill (and others probability systems) is doing, invented at a time where computer did not exists and where the system has to be easily computable by "hands". It's obsolete now.

Comparing numbers of games you have played with number of played games in general could help here. Or better number of played games for all players divided by number of players.

Edit: i read the wiki about trueskill, it is interessting. But as it seems maps will be random on PA.