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

Calculating team skill

11 posts, 690 views
Post comment
Filter:    Player:  
sort
Previously I have experimented with DErankBrackman team strength and simple average, coming to the conclusion that averaging yields better results.

There are still a lot of complaints about imbalanced team games, thus I have tried two other strategies.
  • Maximum likelihood estimator: Weighted average where the weight is higher the less uncertainty a player has. This means the algorithm will try to put a similar number of inactive/newbie and active players on each team.
  • Minimum standard deviation: Average biased by standard deviation. This means the algorithm will try to put a similar amount of pros and nubs on each team.

Both of these turn out to perform comparably or worse than simply looking at the average skill on each team. This only takes into account the game's outcome, whereas the game itself is what really matters. Would you think one of these two might increase game quality, even if the games might end up less balanced?

To clarify the meaning of balance: The balancer currently assures that every player has a near perfect 50% win rate when playing team games. Whether those games are simply alternating stomps or actually interesting match ups isn't considered.
+2 / -0

20 months ago
Calculating that is a hard task. I find that current algorithm usually does it pretty well, with big errors sometimes.
+1 / -0

20 months ago
Teamstrength should make teams with similar skill distribution due to the massive numerical differences. For example, since under teamstrength a 2000 player is worth ten 1600s, the most viable way to offset him is to put another 2000 in the enemy team.
+0 / -0
20 months ago
Is it possible to implement things like the value-killed etc into this calculation?

Because that would allow to maybe deduce how much a person did to win a game.
+0 / -0
I like the calculation/algorythm they use in XVM:

https://koreanrandom.com/forum/topic/2598-%D1%84%D0%BE%D1%80%D0%BC%D1%83%D0%BB%D0%B0-%D1%80%D0%B0%D1%81%D1%87%D0%B5%D1%82%D0%B0-%D1%88%D0%B0%D0%BD%D1%81%D0%BE%D0%B2-%D0%BD%D0%B0-%D0%BF%D0%BE%D0%B1%D0%B5%D0%B4%D1%83-chance-to-win-formula/

Efficiency, WN7, WN8
http://wiki.wnefficiency.net/pages/WN8
+0 / -0
20 months ago
I played World of tanks, where Wn8 is often used to check a players skill lvl.

And i think that it is very hard to compare it the needs for Zerok with those of the Wargaming Products.


But maybe not impossible. It all depends if Dmg etc can be used to calculate the skill of a player.





DeinFreund, i think that the minimum standart deviation is the way to go. Because then a good player just needs to face his direct opponement.
+1 / -0

20 months ago
quote:
Is it possible to implement things like the value-killed etc into this calculation?

Not without solving ZK.
+1 / -0
20 months ago
what do you mean by solving zeroK?
+0 / -0


20 months ago
Solving Zero-K means knowing exactly, in maximum detail, what it takes to be good at Zero-K. With this knowledge we may be able to usefully compare people to this ideal. We do not have this knowledge so we are left with treating games like black boxes and only updating on the outcome.
+0 / -0
ATrankBigMac That is a perfect example of why you shouldn't rely on in-game statistics for a rating system. I've played the game as "unicorn" and "tomato" while keeping the same win rate. Such a system might be helpful if you want to determine a player's play style, but not to balance teams.

I wouldn't want to officially encourage behaviour as you get it with wn-rating in zero-k. The awards are already going in this direction. There's no need to encourage stat-padding any more than that.

It seems like standard deviation might be a good thing to try.
+1 / -0
Try standard deviation and check that the whr sum of each team are not too much distant (+- 10% ?)
Also it may be good to monitor the average win rate of players for a given algorithm to check if it's not too much imbalanced.
+0 / -0