The current method of team balancing works very very badly when there is a high contrast between player elo - eg, an elo 2000 and elo 1250 in the same game. This is especially exagerated in small games, to the point where 3v3 can become a good player and two hopeless nubs against 3 normal players, which will go to the 3 normal players every single time. I would like to suggest a new balancing method: The highest and lowest elo players are placed together on team A the next highest and lowest elo players are placed together on Team B the next highest and lowest elo players are placed together on Team A etc Until both teams are full. This would make teams considerably fairer than the current method, which is godde + 5 crippled noobs against 6 normal players, every time.
+0 / -0
|
I admit that it can be frustrating to get a teamful of nubs when you are high elo, but the current system is very flexible, and thus it allows for clans to play together in most games. In this rather strict way of balancing that you suggest, the same wouldn't be possible.
+0 / -0
|
The only way that this can happen is when Godde really IS worth those 400 more points than any other player in Zero-K. If he wasn't, he'd keep losing elo rather than gaining it. Unless he got it all in 1v1, where the effects of bad allies can't really weigh down your elo as much as it should (But he plays heaps of team games, so no, it's not that). It's frustrating to play with truly genuinely awful allies, because winning or losing often relies on whether or not they drop (Which isn't taken into account by elo). But your system would seriously throw around peoples elo, as even though Godde CAN carry these crazy teams with awful allies (His elo would go down if he couldn't), he'd get some average players too. No, the solution is to split rooms by elo.
+0 / -0
|
quote:
because winning or losing often relies on whether or not they drop (Which isn't taken into account by elo) |
Yes it is. All things being equal the droppers will have a higher elo.
+0 / -0
|
the current system makes small games almost impossible to balance with mixed company.. it is broken because elo is not a linear measurement - a 1700 player is a very different creature to a 1500 player.
+0 / -0
|
what you need is a system that tracks play style, instead of skill the main problem is when you have 1 eco player, 1 air player, 1 assult player, vs 3 assult players the balanced skill distrobution on team 1 will lose to the all out attack powere of team 2 right from the start beter balance would be each team having 2 assult players and 1 specialist the most basic way to do this it to track what lab you start with whenever you win a game, and your peak(relative) eco that game, and balance based on those stats so each team has an avarage skill, play style, and eco ability that is balanced its probably posible to tweak the maths of the ELO system to do this with suficient data
+0 / -0
|
This system is even more unfair than current one. It puts 1 and 3, 6 and 8 against 2 and 4, 5 and 7. Team 1 gets the best of the better players and worse of the worse players. The whole point is that elo difference between better players is more important than elo difference between low elo players. I think a system works where players gets splitted on the average elo of the group. Say average is 1600 elo in a group. All 1600+ players gets balanced in 2 teams. and all 1600- players gets balanced in 2 teams. Then the better 1600+ team gets combined with the worst 1600- team. and worse 1600+ team gets combined with better 1600+ team. So it compensates strong players with more less strong players. Rather than compensate strong players with better low elo players for the other team.
+0 / -0
|
The current system with elo balance is the best we can make of it imo. Like Saktoth said, in IKs example. The top player would not be able to maintain a high elo if he couldn't win against those odds. Apparently it IS possible and the system works as intended. The high elo cant keep up with the new situation and will drop in elo, and will get better allies next time. I agree with IK that elo isn't a linear scale. If elo is an absolute measurement of skill it assumes that a players has the same skill level always (which isn't true). The effectiveness of a player in a game depends on 100 different factors. It is almost impossible to track all of them. Ideally you want to make a separate ranking for each combination of -map -player count -starting factory -elo distribution -state of mind (fatigue, alcohol levels, if you just broke up with your gf etc) (this has more influence on games than you might think) Of course this is impossible... The only thing i'd change in the current ranking system is the seperation of 1v1 and team elo otherwise we have a pretty nice balance system.
+0 / -0
|
Well @[GBC]Springclone that is not the problem. It is not about if a player has a elo which represents its skill. It is about balancing these players and get a balanced game. You can even maintain your elo level with only imbalanced games. One time it is stacked in your favor and win, the other time it is stacked in enemies favor and they win. You win 50 % of your games and maintain elo. But all games are unbalanced and not so fun as balanced games.
+0 / -0
|
So the problem is that the team elo-value isn't a measurement of their actual capabilities? Do you want to make a system that modifies the team elo with a factor based on elo distribution within the team?
+0 / -0
|
you could just use the equasion ((Elo-1500)/100)^3 it will cause goddes elo to tank badly... but it will balance games players <1500 will have exponentialy more -ve value players >1500 will have exponentialy more +ve value it will also prevent people geting silly elo
+0 / -0
|
Actually thats not exponentially, thats cubic. And it will just result in almost everyone being in between 1400-1600 where the curve is pretty much completely flat. Not that I have a better suggestion. I think ELO as a whole is just a broken concept in large games..
+0 / -0
|
my point is that elo is too strong atm - one good player will weigh multiple awful players into his team. quote:
Like Saktoth said, in IKs example. The top player would not be able to maintain a high elo if he couldn't win against those odds. Apparently it IS possible and the system works as intended.
The high elo cant keep up with the new situation and will drop in elo, and will get better allies next time. |
Everything about this line of thinking is wrong. I lose most team-games at the moment, because I am frequently the highest elo player in the room and therefore teamed up with all the lowest elo players. I regain the tiny losses from teamgames instantly by winning 1v1.
+0 / -0
|
then restrict elo gain/los to 1v1 games... its what elo was designed for in any case also in my system 1 really bad player will have the weight of normal palyers so mr elo 1000 will have numerous good teamates to save his ass while a team of 1500 will have no good or bad players just avarage
+0 / -0
|
quote:
I think ELO as a whole is just a broken concept in large games.. |
in 8v8 one player is 1 / 16 15 / 16 is the other players. So your elo in teamgames is based more on others than on your own skill. Ofc any system building on this is whack.
+0 / -0
|
http://en.wikipedia.org/wiki/Elo_rating_systemthat is the elo system its designed for 2 player games and rates relative skill thats why deriving an elo value from team games is a really bad idea unless you meta analyse hundreds of games the ((elo-1500)/100)^3 makes 1500 elo as 0 and any deviation as a positive or negative value, it should allow for the game to balance based on elo more acurately then just avarage of all players(you can tweak the exponent how you like, as long as it is not an even number it should be fine, 1.2-1.5 is probably a more resonable exponent to use)
+0 / -0
|
This was discussed before already, and also pretty much everyone agreed that elo has to be seperated between teamgames and 1v1 (even 2v2 and 3v3). http://zero-k.info/Forum/Thread/3097And I repeat: Elo might be made for 1v1 games as a ranking, but it is much more usefull in teamgames to balance them. ((Elo-1500)/100)^3 sounds a bit too random to me. I see what you are trying to do, but I dont see how it solve the problem. players with 2000 elo get 125 elo more. It is still the question how this extra 125 elo gets compensated. If the enemy team get 2 players with 1462 instead of 1400 you keep the same problem.
+0 / -0
|
true i pointed out that using ^1.2 would be better elo can be used for team games but it requires meta analysis of several games instead of just single game lets say you have players 1 2 3 4 5 6 and they play 5 games 124 v 356 126 v 345 135 v 246 145 v 236 256 v 134 in all cases team 1 won from this you can see that: 1 is the best player 2 is second best 5 and 6 are mediocre players 3 and 4 are noobs we can derive this from the fact that 1 won all his games exept the one he was teamed with 3 and 4, that 256 won agains 134, and that 256 were in winnig games and lost in games where they were teamed with 3 and 4 not shure how to make an algorithim to derive this data tho...
+0 / -0
|
That might be complicated :)
+0 / -0
|
well thats elo for team games tahts why its the best idea to restrict it to 1v1 games
+0 / -0
|
