Loading...
  OR  Zero-K Name:    Password:   

Balancing teams based on player traits

17 posts, 877 views
Post comment
Filter:    Player:  
sort
24 months ago
Problem statement:
Players tend to have traits like being more offensive, economic, or defensive.
(Some players will know exactly in which category I belong, or also where others belong.)
The balancer only takes Elo rating into account, which leads to very balanced outcomes, but there can be cases in which apparently the Elos of both teams are equal, but the game is still imbalanced, fast, and not that fun.
I suspect in many cases when this happens, a lot of economic players are in one team, while the other is raiding well and fast.

Solution concept:
A rating of player traits could support the balancer to equalize the field more, additionally to Elo.
If you know games like DotA2, play style ratings are what I think of.
Each player gets
  • offensive / raiding,
  • defensive / turtle shell,
  • and economic (value)
ratings.
These ratings could be normalized from 0 - 100%.
Mabye offensive rating is 50% on doing 1/n damage in a game of n players, and of course capped on 0 damage and 100%.
Someone better with math can come with better systems here, I suppose.
Now the balancer tries to make teams with balanced traits.

Remarks:
I know this system is very complex and may never happen, I just wanted to make a post that maybe no one will read.
The system would need a silent start where player data is collected until the system can be used.

TL;DR: Additional ratings for off, def, and eco alongside with elo to assist the balancer.
+4 / -1
24 months ago
I thought sometimes at the same concept, but for me the biggest issue is that players do not always play in a certain style constantly - I assume mostly because it would be boring.

I would guess that what makes a player to have a high rank is the willingness to change the "traits" according to the situation. Nobody building economy? Build economy even if you would like to troll comm. Half team making economy? Be offensive even if you feel like making economy. And so on.

We could make a ranking of "who is best economy player" or "who is best defensive player", but you can do that with metrics after the games, I doubt you could "know" how someone will play the next game.

On the idea of "traits", what I would like in game would be some "warning" regarding various situations I see as very unbalanced such as: lots of metal income nobody building singu; 4 people building singu-s in parallel; no planes; no factory of certain type (in large games is nice to have diversity) etc.
+1 / -1
24 months ago
perhaps there should be a role in place for ppl to pick ,

Front Line
Eco
Air
Support
+1 / -2

24 months ago
Maybe if the game happens to notice how people play, perhaps in basic ways like offense vs eco, it could make choices that mix those attributes while also respecting the generic rating.
+0 / -1
You could consider that defense counters offense, offense counters eco and eco counters defense. Such a multidimensional rating model can be used to solve the non-transitivity problem of the elo and whr system (exists only in the German wikipedia version).

Make an actual algorithm and test it on the data to see if it performs better than pure WHR. If it does, you're good. I don't expect big improvements, though.
+1 / -1
24 months ago
There's only two traits: "makes units" and "doesn't make units"
+3 / -1

24 months ago
two traits, lob nor not lob.
+3 / -0

24 months ago
To build or not to build metal extractors, that's the question.
+3 / -0

24 months ago
quote:
There's only two traits: "makes units" and "doesn't make units"

Seems to me that (in lobsterpot) the random distribution of non-optimal play styles should make them, on average, a non-issue. Better than a non-issue, I think the randomness makes it interesting.

I suppose this comment largely is written as a performance to observers, since neither Tiny nor I are likely to change our opinions.
+0 / -0
24 months ago
quote:
perhaps there should be a role in place for ppl to pick ,

Front Line
Eco
Air
Support

People that queue their plop tells you what they are about.

Sometimes some high elos would change plop after seeing initial disposition, but lower elo players not committing to fact early makes it hard to plan around.
+0 / -0

24 months ago
THE PURPLES ARE TELLING US TO MAKE UNITS, TAKE MEXS AND TO NOT SUCK....

i AGAREE !

#PAAARAAADIIINS !
+2 / -0

24 months ago
If there was a magical system that could give individual scores to players on a scale of intrest(raiding, late-game decision making, overdrive contribution, map control, defensive strongpoints) it would sure be interesting. I would like to know if "High skill" correlates strongly to a higher total score across all categories or just some key catagories, or how common/possible is it that someone scores very high in one but almost zero in another.

Ofcourse, there's never going to be an autobalancer that does this in a satisfying way. The best way would be custom lobbies where the lobby custom selects the teams based on individual players like a pickup game of basketball. I remember back in the day of zkl and springiee there would be sometimes where lobbies "Override" the autobalance by !balance then shifting teams around and using !forcestart, but those days are gone. Also, I can imagine the vitriol when players say "Don't queue economy, queue DPS!" like some hellscape Overwatch matchmaker. Or when your DPS player decides to play bertha.
+0 / -0
A strong skill I possess is the ability to annoy my enemies and make my teammates happy. Does that qualify as a high skill positive trait?
+0 / -0
24 months ago
Making your team want to play (versus want to resign) is definitely something valuable considering how easy people start resign votes...
+0 / -0

24 months ago
quote:
ROrankForever A strong skill I possess is the ability to annoy my enemies and make my teammates happy


are you sure it is this way or the opposite? ;)
+0 / -0
Since nobody has explicitly mentioned this and to make clear what you're asking for: the major change requested here is to rate players on something other than winning. Once we break with that rule, anything is possible.

While forcing players into playing for something other than winning to maximize their rating might sound ridiculous, it could in fact be the only reliable way of balancing lobsterpot. Since elo and WHR cannot reliably predict these games without using data from smaller teams.
+0 / -0
24 months ago
quote:
Since elo and WHR cannot reliably predict these games without using data from smaller teams.
I would compare how reliably it can predict when compared to random games not to some "ideal balance" that nobody can define (also I guess size matters, 16v16 probably benefits less from WHR than let's say 5v5).

My feeling is that pot is not that badly balanced. Variations are due to so many factors (people choosing specific strategy, communication, luck, map, etc.) that I doubt we could improve much without restricting freedom.
+0 / -0