2
Quote
At the start of a match, all player's on ONE team have their Elo ratings totaled and divided by 8 (max players). You may realized that this is simply the AVERAGE of a team's total Elo. Team 1's average and Team 2's average are then used to calculate the probability of win (as per the formulas above). If Team 1 beats Team 2, then the appropriate math as above is applied to each player using the probability score calculated by the team averages.
Quote
The maximum amount a player’s score can change in a single match is ±50.
Quote
If the Match Maker determines that you're going to win and you actually win, then your Elo score isn't going to change very much (if at all). The same applies to a prediction of loss and you actually lose, your score may drop but it will be slight.
So the probability of a win for the group is what modifies the amount of ELO change for each member of the group. In your scenario...the probability of a win for Team A is 93.024% and Team B's is 6.976%.
This probability is determined using the following formula: 1 / 1 + 10^ (High ELO - Low ELO / 400)
New ELO scores are then calculated with the following formula:
New Rating = Old Rating +/- Maximum change x (WinFlag - Probability of the win)
Team A (Group Rating = Average of individual ratings = 1575):
Win:
800 = 800 + 50 x (1 - .9302) = 803.49
1300 = 1300 + 50 x (1 - .9302) = 1303.49
2200 = 2200 + 50 x (1 - .9302) = 2203.49
2000 = 2000 + 50 x (1 - .9302) = 2003.49
Loss:
800 = 800 - 50 x (1 - .0698) = 753.49
1300 = 1300 - 50 x (1 - .0698) = 1253.49
2200 = 2200 - 50 x (1 - .0698) = 2153.49
2000 = 2000 - 50 x (1 - .0698) = 1953.49
Team B (Group Rating = Average of individual ratings = 1125):
WIn:
500 = 500 + 50 x (1 - .0698) = 546.51
1000 = 1000 + 50 x (1 - .0698) = 1046.51
2000 = 2000 + 50 x (1 - .0698) = 2046.51
1000 = 1000 + 50 x (1 - .0698) = 1046.51
Loss:
500 = 500 + 50 x (1 - .9302) = 496.51
1000 = 1000 + 50 x (1 - .9302) = 996.51
2000 = 2000 + 50 x (1 - .9302) = 1996.51
1000 = 1000 + 50 x (1 - .9302) = 996.51
Personally, I see this system working quite well as designed. A team with a higher average ELO is predicted to win, and wins, it gets a small increase while the losing team takes a small decrease. If the team with a lower average ELO wins, it gets a large increase in ELO, and the team with higher ELO that loses, gets a large decrease in ELO.
Edited by Jordan Youngblood, 10 February 2013 - 09:44 AM.
