Elo Team Adjancency Bug Please fix for more accurant team rankings

Game Version:

  • **Build:**latest
  • Platform: Both
  • Operating System: Windows 10 or Windows 7 or Mac or Linux
  • Gamertag: replaceme


Play multiple team games and don’t do anything but collect the points after your team wins. Notice that after a sufficient sample size of not participating that you will be ranked as high as your teammates. Deeper Elo explanatation:
Elo is a system where players have ratings adjusted by a K factor(which acts like a scalar) between two ranked players to adjust players ranking. In 1-1 this works fairly accurately but in team games the current ranking system uses a estimated ranking of team. The problem with this is players can just park themselves and go up regardless if they win 50%(which normally elo would put them in the right rating category). Furthermore players with high volume of games can move up while being well below 50% win rate because they will benefit from stronger teammates winning the team game for them.

Reproduction Steps:

  1. Play successive team games with some active players . Don’t do anything but just sit in the game and don’t die . Wait for points to be awarded. Or run the ranking simulator isolated over a high volume of games and see the rating pattern.
  2. Notice that players who play high volume games don’t actually approach a mean which in elo ranking curve means something is wrong. The trend you should see is points keep going up.


See the top 200 of the team leaderboards and count how many players are below 50% in the top 200. 50% in elo means players are

Possible Fix

Understanding Elo Theory

To suggest a fix it’s first important to understand what elo means. Elo rank means in 1v1 that you are somewhere on the bell curve based on your winrate/ difficulty of opponent. A single win means you won a player of this statistical calibre and many wins say you have a collection of wins that support your a player of this calibre. A ranking with below 50% win rate means your current ranking suggests your above your mean.
Understand Team Elo Theory

Team Elo currently means that you won because on average your team is winning and so your moving up becauce your team is winning. This is because points are based on team averages.

A better team elo would be to order the players into respective buckets such as:
Team A Team B
1 1200 4 1200
2 1100 5 1100
3 1000 6 1000

Player 1 points are calculated against player 4
Player 2 against player 5
Player 3 against player 6

In other words players are sorted by ranking and applied ranking this way REGARDLESS if they are the highest contributor in the game. Wait how could you do this brainy? does that even make sense? Well remember Elo is just a statistical estimator of skill but if you play a large amount of games you will move up in the team estimator bucket that is player 5 might move up to the player 4 bucket and get more points reward. This would prevent adjacency bug because players will lose more points as they go up (which is what should be expected). Furthermore players can’t then rely on high volume of games but rather their ranking placement IN the game not just a team average which can inflate their reward.

Possible additional issues.
Another possible issue with Elo adjacency is what todo with a early defeated player on the winning team. Reasoning suggests they should get a loss. But currently their team still wins and if i’m not mistaken they will be awarded points. This is a possible 2nd issue which i suggest that at most award the losing player a loss and make remaining team members on winning team get additional points applied to each team member based on K factor and then divide it by the remaining players. For instance if one player in the example above(say player 6) got defeated player 4 and 5 should get player 3’s points divided by 2( since they also get individually points from players 1 and 2 respectively).

Hence the suggested fix completely eliminates any adjacency whatsoever.

Alternate way to calculate Team Elo. It also occurs me suppose team averages are correct for elo(or you believe it is). Why then are points assigned to players differently? You would treat the team as hypothetical player but I feel this bug as well(because how would you calculate the strength of a team as one player?). Hence i think 1st solution i suggested is better.

(What i mean is one player in team averages can get 100 points for win and second 5 points).