Originally posted by Major Crisis
View Post
All ratings are as they appear, except for 0/(none), which gets changed to 40. So Hellrazor and Cyclone are rated as 10. This might be a bit low, but probably not by much. Rating them properly is extremely important for team balance, though.
In S1-2 the randoms were just added at the end (via a snake assignment), so it wasn't as big a deal as it is in S3, because they were spread fairly evenly. One time you might get a better unrated player and the next time one that's worse.
But this season, their being rated accurately is far more important, because the bot attempts to balance the overall ratings of the two teams as closely as it can.
This means a player who is rated at 80 is effectively being judged to be "as good as" two players rated at 40. If a thoughtful captain were given the choice between the 80 at 10 deaths, or a pair of 40s (each with their own 10 deaths -- basically an extra player), it should be an almost impossibly difficult decision, under the current algorithm. If the smart pick would always be the 80 player, then those players aren't playing at a 40 level and should probably have their rating dropped.
More clearly: Look at the expectations of a player performing at 80, and halve them for the unrated players/players playing at 40. If the 80 is expected to go on average 10-10 in d/j, the 40s should average 5-10. If the 80 is expected to average 12-10, the 40s should average 6-10. (This is somewhat of an oversimplification, as technically 2 players could do some further teaming, but for the most part it works.)
Properly rating unrated players is one of the single biggest improvements we can make in S3. They're much more destructive unrated than they were before, so let's try to get them sorted ASAP.
(We can also look at changing the unrated rating to 35 or even 30, if need be.)
Comment