Forum

Quoted from AB

though a lot of people might start trying harder instead of messing about if they know that if they throw and lose their rating will go down.

I believe so too.
A transparent system where you know what you have to do to advance in rank could not only make people play more seriously, but also help people improve if implemented in a good way.
Of course, it can be incredibly hard to make such a system that cannot be exploited badly and I appreciate all work Jon has done in the current ranking system.

One big problem, I guess, is how to handle different skill levels of players in one team. If one higher level player plays in a team with lower level players and the team looses badly, then the better player should not really get punished in rankings too much or at all because of that.
Also, if someone have an easy time in one team because the teammates are carrying then it should not really make the persons rank increase too much.

I don’t know the details of the current system, but I imagine that the above mentioned things are very hard for an algorithm to detect in a game where teamwork means so much.

Quoted from 4hp

Examples of good teamplay like bodyblocking (or airblasting) an uber to save your combo or causing a distraction for your spy to get a big pick are things that can’t really be quantified in an algorithm.

They could be if they were logged, or maybe if dem file parsing was more advanced. Stats surrounding uber manipulation are sorely absent.

Obscurity isn’t really a good solution to the problems of manipulative behaviour but without running a pickup system myself then it’s the only method I have available.

The strongest players (Mike) in 6v6 generally have a win % of around 70%, so if he’s one of the last players to join a queue you will get people leaving the queue immediately because it’s a coin flip whether they win or lose a ranking point.

It’s possible for the site to run a ranking punishment for leaving the queue, but then you have a very difficult balance to maintain between pissing people off for just being afk and protecting the system.

The scene is probably too small at the moment to support anonymous queueing where levels of ability actually matter, it’s still everyone in the same pickup anyway.

@GentlemanJon not a complaint about your system at all. That sort of limitations is unavoidable. I think a lot of people play lobbies “for the logs” regardless of how the player ranking works anyway, because winning a lobby doesn’t matter to most players. I think only the newer players who are learning to play comp in tf2center really care about winning their games.

And I’ve noticed sometimes it’s good to be on the losing team of a lobby if you’re playing against prem players or whatever. Your team might be getting roflstomped but if you hold your own well you often go up in the rankings, whereas if you’re in the team that’s stomping but not doing much you can go down. So yeah I don’t avoid playing against the Mikes of this world, idk what the others think about it though.

Quoted from 4hp

@GentlemanJon not a complaint about your system at all. That sort of limitations is unavoidable. I think a lot of people play lobbies “for the logs” regardless of how the player ranking works anyway, because winning a lobby doesn’t matter to most players. I think only the newer players who are learning to play comp in tf2center really care about winning their games.

And I’ve noticed sometimes it’s good to be on the losing team of a lobby if you’re playing against prem players or whatever. Your team might be getting roflstomped but if you hold your own well you often go up in the rankings, whereas if you’re in the team that’s stomping but not doing much you can go down. So yeah I don’t avoid playing against the Mikes of this world, idk what the others think about it though.

The idea is to reward individual play, but also to supply balancing data primarily. You can keep track of your own win/loss easily enough, by definition 50% must be average, and I’ve told you the upper bound is around 70% so the closer you are to 70% the better. It will be be in a bell curve distribution so getting from 50 to 55 would be significantly easier than getting from 65 to 70.

This works sufficiently well for outlying players who have great skill, but the more mediocre you are the more it’s dominated by who you have played rather than how you played. A Bayesian system like Trueskill will do a perfectly respectable job of coping with that, but when I’ve look at such a system for balance it was ineffective.

Quoted from S@GA

Do you have an idea how I can have stats for demo in tf2pickup when I can tell for sure that I never played a single pickup in my life? ;o

http://logs.tf/610209

Quoted from Gentleman Jon

[…]

http://logs.tf/610209

That shouldn’t be a pickup log though, maybe it got pickup up because of TF2Pickup.net in the name?
Our log names are like these http://logs.tf/1002413