Rating Servers

[Scarhand]

I don't like Patriot's suggestions much. They aren't easy to balance, and don't really fix the real issue. The problem is rating distribution, as testforecho showed. However, he didn't factor in how easy it would be in each server to get high rating.
I think an important and unused factor is who is on each team. If you join a team where all the high rated players are on it, you will have an easier time, but your rating will not be calculated any differently than if you were on the other team, where it would be very difficult for you to get a high ppm. It shouldn't use the total match averages, but the team averages. I think this would make it more accurate, and less likely to cheat with team-stacking.

[testforecho]

I don't like Patriot's suggestions much. They aren't easy to balance, and don't really fix the real issue. The problem is rating distribution, as testforecho showed. However, he didn't factor in how easy it would be in each server to get high rating.
I think an important and unused factor is who is on each team. If you join a team where all the high rated players are on it, you will have an easier time, but your rating will not be calculated any differently than if you were on the other team, where it would be very difficult for you to get a high ppm. It shouldn't use the total match averages, but the team averages. I think this would make it more accurate, and less likely to cheat with team-stacking.

I think the new K/D system will be based on team averages, unless I totally misunderstood.
But the main problem with many rating systems lies in the fact that players don't really face each other, it's like in chess elo score, where if player1 beats player2 and player2 beats player3 then player1 will have a higher elo score than player3 even if they never played together. Makes sense, but if from that elo scores difference you want to calculate the winning probability, then it's all theory until they face each other. With ET it's much worse, because of renaming, the number of servers, and because some communities of players (etpro, for example) are sort of isolated from others.

The only case when scores can actually be compared is when 2 players play the same match (in the same server, that is) and they are in the same team. But comparing 2 ppl who play in different servers is a problem that can't be solved without headaches, and maybe it can't be solved at all.

That being said, the system is not totally flawed, since if I see someone with a higher rating and I guess that he plays on an easier server, then I can join that server and the same team where he is, and then I can see if I am really better or worse. But that leads to the aforementioned problem of people joining always the same servers, penalizing new servers in a game that is already dying, and leading to a game with few overcrowded and laggy servers where everyone plays. The solution would be to be able to compare players who don't play together in the same map, but how? With arbitrary balancing factors decided by TB admins? The fact that a server can be harder than another is a fact. But finding a balancing factor without being biased is difficult.