Matchmaking with uneven numbers

Having played a few games recently, I think the matchmaker is slightly off with uneven numbers with a limited number of players. As far as I can tell it tries to average out the skill level. While this in theory works as the team with less players has more resources individually, attention itself is a resource that you don't get more of. I think that generally the smaller team should get an advantage.

For example, if you have a blue, yellow and red in the lobby, it will tend to chuck the orange and blue against the yellow - the blue team should win this every time. Blue probably still wins if it's orange and yellow vs blue, but I think it would generally be closer.

I'm not certain how better players affect this. It may depend whether they are just generally solid, or whether they are the type that win games by themselves. I feel like purple players can probably actually take advantage of the extra resources properly.

I think the fix might just be to give the bigger team a slight elo handicap?

A nice analysis (2 years old) using numerical analysis can be found here: https://zero-k.info/Forum/Thread/36555

It computes also a range of elo difference for which things should be more honest.

As far as I remember the analysis linked above has a complete re-implementation of WHR as well (based on all the games played), so it can regenerate the ratings.

For example, right now the predict feature is telling me that myself and a red player have a 13.5% chance of winning against a yellow - I don't believe that figure is even remotely close to accurate.

The biggest counterargument to attention favoring the larger team, is the ease of which a lone player can coordinate his forces and efforts.

Take for example a Thunderbird strike followed up by raiders.
Without voicechat, it is much easier for a skilled player to coordinate a strike or abort depending on the situation, which can change in just a few seconds, but which might also have to be premediated up to several minutes of planning, positioning and unit compositions.
However, countering enemy bombers with Swifts requires almost constant attention to the movement of the enemy air forces and ground AA, and are best done by a dedicated air player.

For a low skilled player, just coordinating two different forces can be hard enough that it is better done by several players.

Controlling a single group of Scythes behind enemy frontlines requires almost constant attention to not run into enemies and accidentally uncloak.

Some other examples that favors of Ease of Coordination over abundant Attention, is to control a single Grizzly or Paladin with some backup, than it is for several people to counter them.

So Attention versus Ease of Coordination depends largely on what type of map and unit compositions that the players go for.

Some types of Coordination require quick decision making, like a Thunderbird strike, while other types of coordination, like switching all your production to counter an enemy threat, is slower, but convincing your allies to realize the threat and stopping their own production, can be a hard sell in a random teamgame, making even long term coordination processes slow and difficult.

Aqua I think step 1 of working out where the matchmaker isn't working needs to be done before step 2. Once you know how it's not working, the maths more naturally follows. - ie., I think come up with a bunch of scenarios where it doesn't work, analyse why, and then work from there.

Godde I agree, however, I think this advantage really only applies to blue players and up, and particularly to purple players. Purple will make good use of that and base strategies around it - ie., being able to coordinate two coms is quite powerful, but I don't think yellow and below have a chance of this.

The 2v1 strategy is clearest to me.

2 silver + Blue, it will stick silver vs blue and silver - it definitely should stick blue against two silver.

If there's more of a skill gap between players I think it's more complicated and I think you can't get fair teams in some situations.

Blue, silver, and red, it will stick blue and red against silver - I think this is fair in this situation.

IMO uneven teams 3v4 or smaller should be no-elo'd by default.

A lot of elo changes hands in odd ways in those games but they mostly come down to how well a specific player can multifac.