Thanks for the re-post Hoagie, having this news as a separate thread makes more sense.

Hello dear League players, we would like to get feedback on the new score system.

As you know, when we moved to a new site we started to use ELO system, instead of the basic "wins give +1, losses give -1" type scoring. The reason is, ELO is a more sophisticated system that takes player skill (in time, the rating / score of a player will become and indicator of player's skill or at least his probablity of winning) in to account when calculating the amount of rating gained or lost as a result of a game.

​We are in the process of improving our score system. Here are the things we should look into (please add more if you think of other things):

1) Overall average score: I believe this should be a weighted average (by number of games played per Theme) of your Theme scores, meaning that Themes you played more should have more effect on your overall average and the ones you have never played should have NO effect

2) K Factor: Is 128 optimal for us? is it giving too much variation? Should we have different K factors based on different type of games (duels, FFA, teamers etc.)

3) FFA games (or team games where there are more than 2 teams): Winner getting a big increase in score (total of all the scores lost by losers), is this fair? Does this make big jumps in score when there are more players (teams)?

4) Team games: When calculating scores, should each team member be compared individually to each opponent, or a team average score should be used?

PS: you can post here if you want or joins us on slack for discussion over there: https://civilizationleague.org/slack/

To complicate the ELO calculations, I think we have another "type" of game to think about:
Games with 2 winners (ex: Theme 11)

Normally we have duels and FFAs. Team games are I think a subset of these.
2v2 for example can be thought as a 1v1, once we decide on what to do on teams. If we decide that a team score as the average of the players' scores in the team should be used when doing ELO calcs, then it is clear that 2vs2 can be thought as the same as 1v1.

Similarly, 3v3v3 can be thought as a 3FFA, with the same logic.

If we decide to use a different method for team games, then we can not think these team games as duels or FFAs. So we would have duels, FFAs, team duels and team FFAs.

However, Theme 11 does not fit in any of these categories as it is an FFA but has 2 winners. The problem is with the conservation of ELO scores.
If it was an FFA, all losers lose points and the winner gains points equal to the sum of the lost points, where each loser's lost point is calculated by comparing his current score to the winner's current score.

In Theme 11, if we do this for both of the winners, the losers would lose 2 times of the score of a regular FFA. Maybe that is OK?

Example :5 FFA and 5player Theme 11 (2 winners), where all players have same ELO score (1500)

5FFA:
all of the 4 losers lose 64 points, winner gains 64 x 4 = 256 points (looks like a lot of points, but remember, the winner defeated 4 players, he had 20% chance to win, anyway this is another discussion)

Theme11:
all of the 3 losers lose 64 points each, winner 1 gains 64 x 3 = 192
but for winner 2 to gain the same 192 points, the 3 losers should lose another 64 points each, total loss for them from this game would be 128 points each.

The solution I can think for this situation is:
Take the average of winner 1's and winner 2's current scores, use it for the comparison with each of the losers one by one and each winner gets the HALF of the sum of the points lost by the losers.

In the above example:
The average of current score of Winner1 and Winner2 would be 1500,
Therefore each loser would lose 64 points each,
Total of lost points would be 192,
Each winner gets HALF, so 96 points each.

I am sure this will complicate the coding of the ELO engine, but I have no other idea...