| 1 |
@CrazyEddie, the balance and the rating system are not independent. In particular, the rating will "adapt" to the balance (and team compositions etc.), so method 1 is very unlikely to work out.
|
1 |
@CrazyEddie, the balance and the rating system are not independent. In particular, the rating will "adapt" to the balance (and team compositions etc.), so method 1 is very unlikely to work out.
|
| 2 |
\n
|
2 |
\n
|
| 3 |
In the end, elo is just a model. It cannot reasonably predict the outcome of games that are influenced by so many factors that are unrelated to player elo: Map choice, time of day, game balance, player moral... You name it. Any other model will suffer from the same. The model will more or less adapt to these things (i.e. a player that is playing more during his "bad" time of day will lose elo).
|
3 |
In the end, elo is just a model. It cannot reasonably predict the outcome of games that are influenced by so many factors that are unrelated to player elo: Map choice, time of day, game balance, player moral... You name it. Any other model will suffer from the same. The model will more or less adapt to these things (i.e. a player that is playing more during his "bad" time of day will lose elo).
|
| 4 |
\n
|
4 |
\n
|
| 5 |
Enter team compositions: The balancer is, in the end, just another one of those factors. The elo model doesn't know about it, and it likely can't model it properly, but it will be influenced by it one way or another. This is why using the current elo with another prediction algorithm (i.e. balancer) will not be fair.
|
5 |
Enter team compositions: The balancer is, in the end, just another one of those factors. The elo model doesn't know about it, and it likely can't model it properly, but it will be influenced by it one way or another. This is why using the current elo with another prediction algorithm (i.e. balancer) will not be fair.
|
| 6 |
\n
|
6 |
\n
|
|
|
7 |
[quote]There are also social issues. If this system gets used then everyone has a wrong value because the systems are self-reinforcing[/quote]
|
|
|
8 |
You could just have both systems running in parallel, where old elo and balance is used until the new system has established reasonable values with how it would've predicted the games and who actually ends up winning.
|
|
|
9 |
\n
|
| 7 |
\n
|
10 |
\n
|
| 8 |
PS: I'm waiting for IvoryKing to show up and tout his method. He does it in every single one of those threads (hope he's not breaking the tradition).
|
11 |
PS: I'm waiting for IvoryKing to show up and tout his method. He does it in every single one of those threads (hope he's not breaking the tradition).
|
math bork