igotnoname

Unremarkable User
A steamroll is one thing in Payload. But when it happens 2, 3, 4 times in a row, people don't like it. The auto-scramble sucks, and as someone that has dabble in data in the past, even I can't even begin to figure out a pattern. There's 32 people, the scores are there, but the formula sucks.

I'm not sure what is being weighed and considered. Historical stats are one thing, but that can't be 100% of it. Not everyone goes 100% all the time, so there has to be consideration for gameplay since you joined. Another thing is it should take into consideration support classes as well if possible. There are some medic, spy, pyro, etc mains. Yet, they sometimes end up on the same team. If someone through the historical data tells us that they're more likely to be a pyro on defense, but a demo on offense, that should be taken into consideration.

If you want to share any of the scrambling scripts with me, I can see if I can play around with them and kick them up your way. I skimmed the Skial rankings, and they're not as straight forward as I hoped they be. If that is used to scramble, that suggests a problem. I also hope auto-scramble is literally not random. I feel as if votemenu scrambles are better than the auto-scrambles.
 

Bottiger

Administrator
The scrambles are not random and votemenu scrambles are exactly the same as "auto-scrambles".

You can read the "details" here. The only thing that's missing is that we weigh the ratings by how much time is spent on each team and the average win time per map.

https://www.skial.com/threads/new-rating-and-balance-system.55615/

It is not likely it can be improved. Rating systems take exponentially longer to converge the farther it gets away from 1v1, and this problem is compounded by the fact that people can leave/join change teams at will. And the fact that you can switch between 9 different classes with varying levels of proficiency.

The reason why there are still major stomps is probably because of people leaving/joining, or restacking teams.