Make rank relative to ingame performance. Every time a game ends we are presented with a lovely assembly of statistics describing metal production, dealt damage etc.. that could be used to accurately describe the players actual performance instead of a ranking system that only counts hours spent in the game.
inb4 do it yourself.
chrank and the constant failure of balance
Moderator: Moderators
Re: chrank and the constant failure of balance
ZK does it! ZK has elo rating! Available on ZK autohosts!
(or something like that)
But yeah, could be nice to have some evolution of the lobby protocol.
(or something like that)
But yeah, could be nice to have some evolution of the lobby protocol.
Re: chrank and the constant failure of balance
How do you plan to deal with smurfs/elo whores?
Re: chrank and the constant failure of balance
Actually, I've been going to implement some form of this for PrincessHosts.
But before I start coding it in, there are quite many issues that need resolving.
1) What about players who, when the hosts are empty, come in and purposefully lose a few dozen times to a fake account? (Thinking on putting a time limit of 10 minutes a game or it wont be taken into account and allowing same team composition based on IP affect the balances only once every hour or two)
2) Should this also take into account time ingame in some form? (I'm thinking that time in-game accounts for 50% of the rank and the Glicko rating another 50%)
3) What about clanstacking/otherwise good players purposefully stacking, which still is the single biggest factor in imbalanced games? I suppose skill rating would eventually fix that, though.
I really think it's too abuseable idea to have the lobby server itself maintain somekind of a game skill relative ranking. There are so many issues that can go wrong. What you really need is a combination of several methods, with the main work done on host, rather than lobbyserver.
But before I start coding it in, there are quite many issues that need resolving.
1) What about players who, when the hosts are empty, come in and purposefully lose a few dozen times to a fake account? (Thinking on putting a time limit of 10 minutes a game or it wont be taken into account and allowing same team composition based on IP affect the balances only once every hour or two)
2) Should this also take into account time ingame in some form? (I'm thinking that time in-game accounts for 50% of the rank and the Glicko rating another 50%)
3) What about clanstacking/otherwise good players purposefully stacking, which still is the single biggest factor in imbalanced games? I suppose skill rating would eventually fix that, though.
I really think it's too abuseable idea to have the lobby server itself maintain somekind of a game skill relative ranking. There are so many issues that can go wrong. What you really need is a combination of several methods, with the main work done on host, rather than lobbyserver.
Re: chrank and the constant failure of balance
Don't just count victory and defeat, also take into account against who.
http://en.wikipedia.org/wiki/Elo_rating_system wrote:The number of points won or lost in a contest depends on the difference in the ratings of the players, so a player will gain more points by beating a higher-rated player than by beating a lower-rated player.
Re: chrank and the constant failure of balance
Yeah, ofc it'd have to be some solid system doing the rating. I don't think Elo's it. Glicko 2 all the way, since it better factors in gaming activity.zwzsg wrote:Don't just count victory and defeat, also take into account against who.
http://en.wikipedia.org/wiki/Elo_rating_system wrote:The number of points won or lost in a contest depends on the difference in the ratings of the players, so a player will gain more points by beating a higher-rated player than by beating a lower-rated player.
However, it'd be kind of rare that it really mattered how skillful opponent your team won. Why? Because if automatic balancing is done anyway, both teams would be close to equal in rating in any case.
Re: chrank and the constant failure of balance
Firing squad/banhammer.Beherith wrote:How do you plan to deal with smurfs/elo whores?
Zero K is cancer and should not be brought into serious discussions regarding the One True Mod.zwzsg wrote:ZK does it! ZK has elo rating! Available on ZK autohosts!
(or something like that)
But yeah, could be nice to have some evolution of the lobby protocol.
(BA)
Re: chrank and the constant failure of balance
Ban dodging -> banned.
Re: chrank and the constant failure of balance
i don't think you should worry about abusability.
it is just too much boring work, and you would not even know if it would be worth it, or the devs (you in this case) built in a prevention algorithm for the way the player tries to trick the system.
also.. even if someone successfully tricks the system, by spending two hours of doing boring stuff, then he can play one game of 30min as smurf, and devestate everyone, and then he will be banned, and would have to do the same thing again...
you don;t have to be 100% secure, just enough to prevent against reasonably assumable attacks. i would assume very few would even try it.
it is just too much boring work, and you would not even know if it would be worth it, or the devs (you in this case) built in a prevention algorithm for the way the player tries to trick the system.
also.. even if someone successfully tricks the system, by spending two hours of doing boring stuff, then he can play one game of 30min as smurf, and devestate everyone, and then he will be banned, and would have to do the same thing again...
you don;t have to be 100% secure, just enough to prevent against reasonably assumable attacks. i would assume very few would even try it.
Re: chrank and the constant failure of balance
Heh, yes, this is true.hoijui wrote:i don't think you should worry about abusability.
it is just too much boring work, and you would not even know if it would be worth it, or the devs (you in this case) built in a prevention algorithm for the way the player tries to trick the system.
also.. even if someone successfully tricks the system, by spending two hours of doing boring stuff, then he can play one game of 30min as smurf, and devestate everyone, and then he will be banned, and would have to do the same thing again...
you don;t have to be 100% secure, just enough to prevent against reasonably assumable attacks. i would assume very few would even try it.
With the hosts I put, I always had the goal of having a large amount of active admins who can do chranking and kick/ban players whose language or behavior goes really over the top. I suppose it'd be good enough to just have a Glicko rating system and admins overseeing it with some light standard protections like not counting games that lasted under 10 minutes..
Re: chrank and the constant failure of balance
What's the common opinion, are big games in ZK better balanced than in BA?
Re: chrank and the constant failure of balance
How do you define balanced? If an algorithm is supposed to make sure winning chances of each team is 50%, then if it works well, the win rate will be truly random. How do you validate that?
-
Silentwings
- Posts: 3720
- Joined: 25 Oct 2008, 00:23
Re: chrank and the constant failure of balance
There are statistical tests for 'true' randomness (i.e. maximal entropy - in short, what you look for is fairness and a lack of correlation between successive trials). If anyone is willing to share data sets about ZK balancing/results that would be very nice.
But what I really wanted to mention is that I spoke to bibim a while ago I remember he mentioned that he had this as a long term project but that it was still a way off complete. So I have another idea, which is that hoards of people tell bibim how much they'd appreicate to see it + find out if they can help
In the short term, finding some good admins for your autohost seems like the best bet.
But what I really wanted to mention is that I spoke to bibim a while ago I remember he mentioned that he had this as a long term project but that it was still a way off complete. So I have another idea, which is that hoards of people tell bibim how much they'd appreicate to see it + find out if they can help
In the short term, finding some good admins for your autohost seems like the best bet.
Re: chrank and the constant failure of balance
Good admins is a requirement ofc, but doesn't help with the problem that 80% of the players are rank 5, having very different skillz.
I wonder how hard it is to hack open spads rank>7, using rank<10 as "original rank" and ranks in autohosts smurf-db >=10 as ranks 0-50 (or whatever you like)
I wonder how hard it is to hack open spads rank>7, using rank<10 as "original rank" and ranks in autohosts smurf-db >=10 as ranks 0-50 (or whatever you like)
Re: chrank and the constant failure of balance
Hmm, I think a good metric would be, how often do people feel frustrated at the teams already upon game start? As long as everybody feels like they have a fighting chance, things are fine.Beherith wrote:How do you define balanced? If an algorithm is supposed to make sure winning chances of each team is 50%, then if it works well, the win rate will be truly random. How do you validate that?
Of course there's no way to measure this numerically, but people who regularly play big matches in several games might have some smell about whether ELO helps.
Re: chrank and the constant failure of balance
What. The. Fuck.good metric = how often do people feel frustrated
How do you even measure level of frustration? Number of frowny faces in chat? Capture their webcam and do emotion detection?
Edit:
The only good metric is that the win rate of every single player converges to 0.5
Re: chrank and the constant failure of balance
I feel a rank system that runs from 0 to 7 is already too limited by the amount of ranks it has. The difference in skill of someone who admins think is rank 6 can be so big to someone they think is rank 7, that the one with rank 7 wins vast majority of games on consistent manner for his team.
Of course, it's virtually impossible to have a perfect balance. I do feel however, that somekind of a rating system with admin support would be better to rely on for an approximately balanced game than rely on one based on time spent in-game and admin support.
Consistency is also important from the admins. And I do feel that too rarely people are chranked down; Unfortunately, doing so often hurts their egos and has them quit the host. (Yet, they go on rampages of whining how balance sucks..
)
For a metric.. I've heard awards suggested, damage dealt suggested, but I think the only reasonable way of adjusting rank is simply defeat or victory. It creates a system that is constantly fluid and changing but after enough many games, consistency should emerge. When you lose, you'll next time be more likely to end up to better team; And when you win, you're more likely to end up to worse team.. Sure, it might be bad for a while for a new player who just happens to hang with clan-stacking pros for a few games and next time winds up to be ranked high and loses, but he'd then be lowered in rank and on the long run, the average might be surprisingly good.
In any case, obviously some testing is required and loads of initial data, so before it would be implemented live, there should be a test period and old replays fed to the algorithm, as Beherith suggested in chat.
Of course, it's virtually impossible to have a perfect balance. I do feel however, that somekind of a rating system with admin support would be better to rely on for an approximately balanced game than rely on one based on time spent in-game and admin support.
Consistency is also important from the admins. And I do feel that too rarely people are chranked down; Unfortunately, doing so often hurts their egos and has them quit the host. (Yet, they go on rampages of whining how balance sucks..
)For a metric.. I've heard awards suggested, damage dealt suggested, but I think the only reasonable way of adjusting rank is simply defeat or victory. It creates a system that is constantly fluid and changing but after enough many games, consistency should emerge. When you lose, you'll next time be more likely to end up to better team; And when you win, you're more likely to end up to worse team.. Sure, it might be bad for a while for a new player who just happens to hang with clan-stacking pros for a few games and next time winds up to be ranked high and loses, but he'd then be lowered in rank and on the long run, the average might be surprisingly good.
In any case, obviously some testing is required and loads of initial data, so before it would be implemented live, there should be a test period and old replays fed to the algorithm, as Beherith suggested in chat.
Re: chrank and the constant failure of balance
i like the idea of an auto learning or measuring piece of software (maybe in spads for each host or globally via DB/interface) and set the chrank'ings automatically and maybe uses a higher usage of the lower rank's so that the "beginner" is weightened more "better" and so we dont have the problem with less resolution between 6 and 7
(more linear than logarythmic)
and maybe we dont need to have such rankings for spads internals
cause the thing knows every skill for every player then and just needs to autobalance with that information and not with low res rankings
then the ranks would be obsolete and just a sign of how "old" a player is - but not how good
i support the idea and i will help in any cases to bring this meachanism into an live thing
(more linear than logarythmic)
and maybe we dont need to have such rankings for spads internals
cause the thing knows every skill for every player then and just needs to autobalance with that information and not with low res rankings
then the ranks would be obsolete and just a sign of how "old" a player is - but not how good
i support the idea and i will help in any cases to bring this meachanism into an live thing
Re: chrank and the constant failure of balance
Well, going to just upload the info about a game when it ends to a Python script online and use a remote database for holding the ranking.muckl wrote:i like the idea of an auto learning or measuring piece of software (maybe in spads for each host or globally via DB/interface) and set the chrank'ings automatically and maybe uses a higher usage of the lower rank's so that the "beginner" is weightened more "better" and so we dont have the problem with less resolution between 6 and 7
(more linear than logarythmic)
This is so that multiple hosts can easily use same ranking list.
-
[LOeT]Zinn
- Posts: 33
- Joined: 19 Oct 2009, 01:27
Re: chrank and the constant failure of balance
I vote for some kind of ELO-like system and strong identities.
For instant relief I would like to have a single new command:
!oppose <player1> <player2> <timeperiod>
where player1 and player2 are nicknames and the optional timeperiod defaults to 1 hour and is written like !ban does it. Just like !ban there should be a list and a unoppose command and the balancing might tell about involved oppose-rules.
It should be voteable.
Bimbim, hear my prayer!
For instant relief I would like to have a single new command:
!oppose <player1> <player2> <timeperiod>
where player1 and player2 are nicknames and the optional timeperiod defaults to 1 hour and is written like !ban does it. Just like !ban there should be a list and a unoppose command and the balancing might tell about involved oppose-rules.
It should be voteable.
Bimbim, hear my prayer!