balancer test

News, server & forum rules
User avatar
laboRHEinz
Administrator
Posts: 1257
Joined: Fri 4. Sep 2009, 13:28
Description: Old Fart
Location: Hamburg

Re: balancer test

Post by laboRHEinz » Mon 29. Aug 2011, 22:55

In fact, the rematch functionality is disabled here. Wormbos balancer runs here because of its pph-based (=kinda skill based) player shuffling at match start and sophisticated & improved rebalancing features.

User avatar
Pegasus
Posts: 1230
Joined: Wed 4. Nov 2009, 23:37
Description: ONSWordFactory
Location: Greece

Re: balancer test

Post by Pegasus » Mon 29. Aug 2011, 23:02

Quick question here: since when does CEONSS do time-conditional rematches? Edit: N/m, heinz just struck that down.

Also,
Crusha K. Rool wrote:[...]let the players vote at the end of the round (with F1 for 'yes' and F2 for 'no') if the match should be restarted with balanced teams if it ended within the threshold time. That way you wouldn't force players to play a map again that was finished just a few seconds too early and thus needed to be replayed, although some players might want to start another one now.
That's a patently bad idea. Not only would you not be smoothing the jagged issue of rematches, you'd be creating yet another point of contention between the players, splitting them into 2 groups and very likely causing a part of the rematch dissenters to leave should the opposing option prevail. Match balancing should be fair, reasonable and as much outside the sphere of ingame players influence as possible. Similar to the Law, it's a solid and immutable, take-it-or-leave-it kinda deal that doesn't take sides or is affected by dynamic social circumstances.

Btw, for the sake of every misinformed crybaby who ever ragequit blaming the balancing scripts for specifically inflicting unfair misfortunes upon them - but also for the better information of the rest of us - can we get some detailed info in this thread about exactly what this latest version of EvenMatch does and what factors it takes into account while doing so? Also helpful if the person who'd elaborate on this could predict some of the aforementioned ridiculous assertions and explain what this script doesn't do to them (odd, I know, but go with it). Cheers.

PS: Welcome to the msg. board, crocface :).
Eyes in the skies.

User avatar
Crusha K. Rool
Posts: 118
Joined: Sun 28. Aug 2011, 23:14
Description: Coding Crocodile
Location: Germany

Re: balancer test

Post by Crusha K. Rool » Tue 30. Aug 2011, 02:10

Pegasus wrote:PS: Welcome to the msg. board, crocface :).
I came because I missed your posts. 8-)


And in order to promote some of my future creations here.

UltimateMappingTools are in RC status and undergoing intensive bug checking and another thing called UTMapRate is in work, which basically is supposed to let players give their rating on the current map and also specify what they dislike about a particular map (ratings are stored as INI on the server together with the player's unique ID, so a player has only one vote per map but can alter it later if he changes his mind).
I figured that would help admins to get some real feedback from every player about their map list rather than making assumptions about a map's popularity based on how many players tend to leave when it comes up or it's average PPH count.

And mappers get more concrete feedback as players who give a bad rating can also specify the most striking issue of the map as reason for their vote.
  • long walk ways
  • too much spam
  • stalemates
  • lack of weapons
  • ugly design
  • not many tactical possibilities
  • bad linksetup
  • comeback too hard
  • asymmetric balance
  • performance issues
  • bad choice of vehicles
Feel free to come up with more.

User avatar
Wormbo
Posts: 384
Joined: Sun 28. Aug 2011, 11:52
Description: Coding Dude

Re: balancer test

Post by Wormbo » Tue 30. Aug 2011, 08:34

Ok, features of Even Match. The mod basically has two modes of operation: It can shuffle teams to even out the skill levels and it can gradually rebalance teams to ensure similar sizes.

Shuffling does not happen during the match, only before the match starts or if the first round was too quick. As pointed out already neither is mandatory. Even Match remembers each player's points per hour (PPH) for up to 48 hours, but only for the last match that player played. Due to GUID sharing, I had to tie this data to IP+GUID, so if you're on a dynamic IP, your data is discarded as soon as you get a new IP. Shouldn't be much of a problem, though. Usually you join mid-game anyway and thus establish a PPH value before it is needed.

Rebalancing during the match takes team "progress" into account. Progress is calculated by counting the number of nodes in possession and their current state, as well as core health and vulnerability. If both teams are approximately equal this corresponds to a progress of 0.5, one team currently being better raises that team's progress and lowers the other's. Progress 1.0 basically means victory.
The admin can configure a progress value for the smaller team, above which no rebalancing takes place. In the latest version, this value is shifted more towards 1.0 if the size difference grows.
Rebalancing is performed in two stages. If a size difference is detected, the first stage is "soft rebalancing". During that time, new players are forced to join the smaller team and additionally respawning players of the larger team may be forced to switch teams. The algorithm will exclude the "valuable" players of the bigger team here, which by default are the top 75% players on the scoreboard. That percentage can be configured. If soft rebalancing is successful, no further action is taken. However, if after a configurable amount of time teams are still unbalanced, random players from a list of candidates on the bigger team are switched to the smaller until teams are even. Valuable players are not determined by their placement on the scoreboard anymore, but by the current state of the game. If a player is currently healing, defending or attacking a node or driving an important vehicle, then that player won't be switched.

Other options for the balancer include randomly switching sides before the match starts, ignoring on-join team preference and preventing players from switching to the winning team, where "winning" is a progress above a certain threshold and not linked to the rebalancing progress limit. Rebalancing can be delayed at the start of the match so players can finish downloading maps and stuff.


One weak point I found yesterday is that bots are taken into account for team size during rebalance. That may not be a problem on a full server, but certainly isn't fun for the team with more bots when there are fewer players. Next version will ignore bots here as well. They are already ignored (or actually removed) for the quick round team shuffling.

borgqueenx
Posts: 47
Joined: Mon 21. Nov 2011, 21:47
Description: Ut2004 player

Re: balancer test

Post by borgqueenx » Tue 22. Nov 2011, 13:01

Ive made a new topic about the balancer please check it out:)

SpaceEngineer
Posts: 75
Joined: Sat 3. Mar 2012, 22:26
Description: what is it?

Re: balancer test

Post by SpaceEngineer » Sat 10. Mar 2012, 00:37

Sorry if I post in wrong thread, but is it possible to enable match restarting if round was too short? I.e. if one team win within 5 minutes after match begin, it seems something wrong with balance, and match should be restarted with rebalanced teams. It is so sad wait for favourite map for a hours, vote for it, and have just 5 minutes game on it.

Alternative solution - allow voting for just played map, so if many players want replay it, they may vote for it.

User avatar
laboRHEinz
Administrator
Posts: 1257
Joined: Fri 4. Sep 2009, 13:28
Description: Old Fart
Location: Hamburg

Re: balancer test

Post by laboRHEinz » Sat 10. Mar 2012, 12:40

Hi SpaceEngineer,
technically, it wouldn't be a problem. But there are also cons about it. For instance, if you don't like a map and have to play it again. And there are side-effects like "Hey, team, don't attack core, wait a minute, otherwise we'd have to play it again".

We'd need some more votes. Please start a public poll within the ceonss/server setup section.
Greetings!

SpaceEngineer
Posts: 75
Joined: Sat 3. Mar 2012, 22:26
Description: what is it?

Re: balancer test

Post by SpaceEngineer » Sat 10. Mar 2012, 22:59

laboRHEinz wrote:For instance, if you don't like a map and have to play it again.
Whatever, if game goes to overtime on map that I dislike I should wait till match end.
laboRHEinz wrote: And there are side-effects like "Hey, team, don't attack core, wait a minute, otherwise we'd have to play it again".
But this is good. I am sometimes go away from the core and just defend key nodes, if game seems to be too fast.
laboRHEinz wrote:We'd need some more votes. Please start a public poll within the ceonss/server setup section.
Sorry, I can't figure out how to start a poll. Can you do it instead of me? And then tell about that to all in game. I'm sure at least 10 players will support these idea.

User avatar
Pegasus
Posts: 1230
Joined: Wed 4. Nov 2009, 23:37
Description: ONSWordFactory
Location: Greece

Re: balancer test

Post by Pegasus » Sat 10. Mar 2012, 23:29

Considering that forcing everyone to sit through another round past the server's understood preset victory points just because the match happened to be shorter than some'd like would obviously be a courtesy to one group (those who like the map) at the expense of another (those who don't), I think a more elegant compromise would be to code a ServerActor that just leaves a map unlocked (place it at the top of the locked stack, push every other locked map back one spot) if the match's duration doesn't exceed some agreed upon value, say 7mins. That way, if most people agree that the teams just happened to be badly stacked but the map's still worth replaying, the majority can get its way by revoting it immediately. If there's fewer who think a rematch is in order compared to the rest, the majority can vote elsewise and won't have to suffer. Obviously this proposal would only address the problem on absolute terms ("Should a map be left to get locked if the match ends very quickly? Yes/No") and offer no recourse to any attempts to engineer a desired match length through ppl avoiding to destroy the core before the threshold point. That is a social problem and, as we well know, there's no system that can't be gamed eventually; tech is the best answer to technical problems, people are the answer to "people problems".

Just my 2 cents.
Eyes in the skies.

SpaceEngineer
Posts: 75
Joined: Sat 3. Mar 2012, 22:26
Description: what is it?

Re: balancer test

Post by SpaceEngineer » Sun 11. Mar 2012, 14:16

Thanks laboRHEinz. Btw, the problem was that I don't find a button "New poll" (like on my own site), just "New topic".

Also, Pegasus proposes another good idea. Maybe yo can add new item in that poll:
* Restart
* Allow vote for restart
* Don't change current system

The problem is that on the SEONSS recently played maps becomes locked, and match lasts up to one point. I was confused by this system then I first played here. Another suggestion is make game for two points, but with modified rules. If team vins before overtime, it gets 1 point and next round will start on the same map, with teams exchanged their bases. This will allow losing tream make revenge on asymmetric and bad balanced maps like Spriffingrad. If team vins after overtime, it gets 2 point and match ends. This will prevent to very long and boring 1.5-hours game that often occured with standard server rules if teams are good balanced (or map badly designed like BitchSlap).