Free Allegiance Forums

Hi guys,

I tested in the beta lobby with some other guys last night and we intentionally created some weird imbalances to see what the system would allow and what not. We were horrified to learn that newb stacking is a very real danger with the current implementation. I launched several 1 vs 1 games and then let the lobby join to see how the balance would work.
From what we saw it seems that the autobalance feature only takes total team Elo into account and not the number of players. This caused some hilarious team formations:
- Me and another player vs 5 other players (ELo was 27 vs 41).
- Me alone vs 4/5 newbs/voobs (Elo was 16 vs 25 i think)

essentially i think it is possible that as a level 16 vet I could face unlimited level 0 and at least 16 (probably 15 more) level 1 newbs in a 1 vs 100 game with ELO and your program thinking it was balanced /huh.gif" style="vertical-align:middle" emoid=":huh:" border="0" alt="huh.gif" /> I know I'm good but a 100 pilots is too much even for me /wink.gif" style="vertical-align:middle" emoid=";)" border="0" alt="wink.gif" />

unless i'm reading the current implementation incorrectly you do not take imbalance into account at all and this could cause some very strange teams.
eg. Vet16 joins team 1, then 16 newb1 join team2.

I think you might want to rethink your current implementation of the autobalance feature (btw because of limited numbers of testers i couldn't try a more than 3/4 player imbalance, so i don't know if you have a max imbalance built in, if you do then you might want to reduce the max imbalance limit for small games)

hope this helps /smile.gif" style="vertical-align:middle" emoid=":)" border="0" alt="smile.gif" />

Was the gamesettings set to 1,2 or off on imbalance? or did i totaly asked the wrong Q?

Zap

You can't face unlimited (0)s because they count as a (1) for balance purposes.

I actually think they were considering leaving imbalance at N/A as part of the autobalance. The rationale being that in a "Squad vs. the World" type game, it won't matter that the other team has more players of one team has significant talent, it may still be even.

That said, I think it won't work out to be very even. We are basing ELO currently on imbalance of 1 games, but N/A balance is a completely different style of gameplay and I don't think ELO accurately represents the difference. For example, I think 7v14, even if team ELO is even, will be horribly unfair for the 7 players because of the inherent power of "numbers" that ELO won't take into account. Such as having 2-3 bomb runs going, it won't matter how much skill the 7 have if they don't have enough players.

Maybe I'm mistaken, but I thought the current idea was to do ELO with imbalance N/A.

As far as I know, and as far as I've seen tested, when the Balance is set to AUTO for ELO purposes the imbal is since to NA.

guitarism wrote:QUOTE (guitarism @ Dec 7 2006, 09:01 AM) As far as I know, and as far as I've seen tested, when the Balance is set to AUTO for ELO purposes the imbal is since to NA.

How about:

If imbal is set to NA, then add the number of players to the elo number

IE:

Team 1 ELOs:
7 + 3
10 + 3
13 + 3
total : 29 (20 normal ELO + 3 for each player because there are 3 players)

Team 2 ELOs:
0 + 4
2 + 4
5 + 4
4 + 4
Total: 30 (12 normal ELO + 4 for each player because there are 4 players)

Grim, perhaps 'horrified' was a bit to strong of a reaction for something that is expected (or should be) /tongue.gif" style="vertical-align:middle" emoid=":P" border="0" alt="tongue.gif" />

Things get all 'wonky' in small games regardless of elo. Right now (or it was like this) games less than 5 per side don't count towards stats. Now that does not mean we should try and balance games smaller than 5 per side, but I would take any such attempt with a LARGE grain of salt.

The current balance code will allow for games of 5 vs 20 (in rare cases), like the squad versus world scenario (or more likely squad versus newbs).

A 'sliding scale' or weighting factor was proposed to deal with the situation you mention. In a 2 versus 4 game, adding a 5th to the side with 4 will likely add more capability to that team than an unmodified elo would reflect. But it quickly became apparent, trying to figure out such a modifier was unneccessarily complex with little return on the effort.

So don't sweat the small games and don't expect the balance button (or auto balance when joining a running game) to be of much use.

Dog

I'd like to repeat my suggestion of calculating the TeamRankSum with a minimum of (5) per player.

This would make all Newb(0)s be counted as a (5) when balancing teams, and will help prevent really weird team combinations like 10 vs 3 being seen as "fair" by the system.

--TE

Tigereye wrote:QUOTE (Tigereye @ Dec 7 2006, 01:43 PM) I'd like to repeat my suggestion of calculating the TeamRankSum with a minimum of (5) per player.

This would make all Newb(0)s be counted as a (5) when balancing teams, and will help prevent really weird team combinations like 10 vs 3 being seen as "fair" by the system.

--TE

Seconded. Good number, should roughly approximate the 'worth' of a newbie to a vet. At least, much better than using numbers < 5.

On my suggestion, 0s vs Vets

team 1 (5 players)
10 + 5
14 + 5
11 + 5
16 + 5
9 + 5
Total 85


team 2 (9 players)
0 + 10
0 + 10
0 + 10
0 + 10
0 + 10
0 + 10
0 + 10
0 + 10
0 + 10
Total: 90 (0s still count as 1, so 9 normal ELO + 9^2)