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

Game balancing

9 posts, 820 views
Post comment
Filter:    Player:  
sort
12 years ago
This isn't something i normally care about, but after a recent discussion with antelope about smurfing in which game balancing was heavily touched upon, i couldn't not post this.

http://zero-k.info/Battles/Detail/76169

I think this game perfectly illustrates my point, there are 3 sub 1300 elo players on my team, but none on the opositions team!

I don't understand how the balancing algorithim could EVER have come up with this, it's completely IMBALANCED.



Team 2 - Elo total 11726

Saylu 1285
HiKitty 1820
Shaddark 1594
Optimus_Ky 1286
Billy_Boy 1549
Ahira 1459
Princereaper 1243
Pantifax 1490

Team 1 - ELO total 12722

Dogma 1538
Kristis 1305
sfireman 2081
Sheep 1698
Gretchen 1614
Noobaccount 1487
MooningForces 1540
Fluffy Bunny 1459

Leaving a 996 total elo difference, meaning on average each player on team 2 was just under 120 elo better than team 1.
+0 / -0
Skasi
12 years ago
cbalance
+0 / -0
12 years ago
This one is easy.

Optimus and princereaper are in a clan together. They both are level 25+ with a sub 1300 ELO. Due to popular request clan balancing is now in effect in the main juggled rooms, so they play together. The rest kinda follows from that.


+0 / -0
12 years ago
I thought of that too, it's not just those 2 mentioned that are in the same clan in this game.

Despite the cbalance, the game could still have been balanced better, Saylu could have been swapped for dogma or gretchen.

We probably still would have lost, but the balance algorithim would still have been more accurate. in fact there are probably several different ways it could have been balanced mathematically and it would have been better.
+0 / -0
(Note: I wrote this before seeing Pantifax's post)

In that case though I would think that balance would sort itself somewhere else, by giving that team better other players. Theoretically bigger team sizes would lead to smaller elo differences. I guess there could be the clan-balance interactions interfering elsewhere also.

It is possible the algorithm doesn't work optimally all the time also. There is a reason this is "the easiest hard problem", and I'm sure clan balancing doesn't help that. (http://en.wikipedia.org/wiki/Partition_problem)
+0 / -0
12 years ago
We will just have Licho sift through all 87178291200 (=14!) possibilities of allocating 14 players to 2 teams to come up with the optimum solution. We may have to wait a day until balancing is done though.
+0 / -0
12 years ago
Hmm, I think you overcounted slightly, there are 14!/(7!7!) = 3432 ways of allocating 14 people to 2 teams (choosing 7 from 14).

Still, the point still stands, balance is tricky.
+0 / -0
12 years ago
Yeah, I did exaggerate a little (only by 8 orders of magnitude though, so that is not that much ...). Also, there are 16 players, not 14 :-)

I am thinking the current allocation algorithm doesnt quite work when there is a single clan with ELO that isn't about average.

Even I can come up with a better distribution uses the stupidest of algorithms:
1 first, put people of the same clan together and sum their ELOs. (there is only one clan here, so that is easy). Distribute clans over teams using their summed ELOs just as your would do for players (see below).
2 then take the highest ranking ELO player remaining and put that in a team with one player less
3 take the lowest ranking ELO player remaining and put that in a team with one player less
4 repeat step 2 until all players exhausted.

Arguably totally unfair as it groups the low and high ELO players together and even then I come up with:
12344 for the team with Optimus and princereaper (and HiKitty, sheep and yours truly)
12104 for a team consisting of fireman and all the remaining lowest ELO rank players (I am sure he will be amused).

Yet, not nearly as skewed as was balanced.

+0 / -0
12 years ago
Personally, i'm more than happy to beat the odds, i am only making this point after my discussion with antelope about smurfing imbalancing games.

We could have balanced the game better between ourselves, like players used to ;)
+0 / -0