Feature request on UT3 : a smart autoteambalance

[MonsOlympus]

Hey heres an idea put the saved games in documents and settings for that user, so when another user logs in and plays the game they already have their own profile. Lmao vista XD

[Jrubzjeknf]

How does this unbalance more than now ? Can you enlighten us on this, please. Would be really great.

You base skill on previous played matches. If a person is playing that does not play according to that said skill, then the algorithm will balance that player out against another player that does play according to that skill. Therefore, your balancing algorithm has potential to actually unbalance teams.

To be less correct: you have different types of skill. In CTF, you have attackers, defenders and midfielders. A person's position depends on the map, which influences any measured skill. Some people can be great players, but suck on a certain map. Others generally suck, but are great at a certain map.

Bottomline: humans can not be predicted or fully measured. Since skill is a part of that, you can't measure that either.

Only if they change inbetween 2 matches during one playsession.

Balancing per playsession is even more useless, because most people play a few matches, and then leave again. You can't get correct results out of a few matches.

[MonsOlympus]

That sounds alot what I said...

I dont think its so much that stats arnt an accurate guide to skill but which stats give the best results (closest to true skill), or even the lack of stats for given skills when autobalancing.

You just need the right algorithm as you put it but I would say set of skills to guage for said match. Since people vary at different skills then it wont be just 1 which determines auto balancing. One skill is actually average skill over time and if it is higher you would assume someone is a better player on average. If theyre great one time but **** the next then you could assume lack of wisdom for lack of a better word and perhaps that they need experience to get better not that they are a "bad" player as such just they have less average skill.

I know 2k4 can keep track of stats per map though so hey you never know it could know that your better at one map.

For position that is a tricky one since there is no set position the best thing you can do is provide a set of stats which will cover all major positions and guage your offensive/defensive skill accordingly.

Its not totally impossible that a auto skill level adjust would actually work, it wouldnt be a simple system either though

[B0T0X]

I wouldn't mind having a feature like this.

...Teams will go from utterly stacked in my favor to utterly stacked against me. It got to the point where almost everyone left and it was me vs. two other guys playing CTF. If the game was smart it could've added an appropriate amount of bots to smooth things over (assumings bots will actually have 1/2 a brain in UT3...)

Anyway... I'm for it.

[fuegerstef]

@Hedge:

It is still better than how it is now:

RED: Lauke, Gitzzz, CombatCarl, Killu

BLUE: Player, Player, Poop, MegaSuperKiller

[Kronos]

Lol thoughs odds seem highly unlikely in a public server but I get your meaning.
At anyrate I'd like to ask a question about this feature on CTF. Suppose you have a flag runner who mostly dies alot and doesnt kill very often because he's constantly transing, speeding himself as fast as possible to the otherside of the map. How would it measure this player out? My guess is it would think he sucks and would ultimately throw him on a team of better players, causing that team to be heavily stacked with a good runner and perhaps real good defenders.

[MonsOlympus]

Well you could do deaths while holding the flag vs captures, that could show how skillful someone is at capturing the flag. So more deaths and less captures would mean less skillful, then again someone might not even have any flag grabs to their name let alone captures so it would make you think they would be more defensive or counter offensive players.

So while there isnt any predefined positions or roles in UT I think its possible for stats to show what roles people play most often and their avg skill at each.

[Luseferous]

Heres an idea. If the teams are judged unfair* why not just swap the top player from the winning team with the bottom player from the losing team if numbers are similar. If not just transfer the top player to the losing side.

Not perfect by any means but it should work reasonably well.

*Using some formula based on SPH/FPH/EFF for players who have played for more than 10% of the match so far. Some weighting based on the second & third players figures from top & bottom of this score table would also help.

[fuegerstef]

Heres an idea. If the teams are judged unfair* why not just swap the top player from the winning team with the bottom player from the losing team if numbers are similar. If not just transfer the top player to the losing side.

Yes, everybody trying to be only the second player will work out well. "I must do some suicides quickliy to lose points or I will be on the losing side."

TBH: I wouldn't like to put all efforts into winning just to be swapped to the losing side.

I say, balance it before the match and then let everybody try as hard as they can without being punished for being good.

[Luseferous]

Yes, everybody trying to be only the second player will work out well. "I must do some suicides quickliy to lose points or I will be on the losing side.

Then it would work, If the top 3 players are trying to out cr^p each other then the teams will be less unbalanced

Seriously though its usually one of the top couple of players who will swap teams to balance the game anyway.

[superwarrio]

to say it myself, i truly think featur like this would be awesome, but implementing it anywhere else than dm, would be very very very difficult, and inaccurate, and judging from previus posts maybe but unlikely in ctf.

but then it would be way better than the current system

so a bit of newthinking probably from epics side would probably be great cause i can't think of anything for warfare right now

[X-Cannon]

Actually there is a feature like this already implemented on purepressure.com

sweet friggin set up they have. The only caveat is that since it's fairly new there haven't been alot of games played other than TAM. The more games played the better the skill level balancing.

Right now over at warstriven (formerly onscentral) we are trying to figure out a good night to get everybody together and play in pugs. I figure when UT3 hits and there is a massive influx of players, more games will be played and the skill level for the site will work that much better.

[_Lynx]

Shambler's working on a "balance by skill" addin to his TitanTeamFix balancer.

We currently have it llinked directly to a MySQL database where it's reading and writing - so I expect it'll be up and running soon.

oh, wow. Can't wait to see that - interaction with MySQL from within US looks quite promising.

[DeathByDribbling]

The big question is how accurate the stats are. While I do agree that ut2004 stats are far from perfect the top scorers in any game type generally have the highest stats. To just distribute by those settings would be a reasonable basis for teams. I'm sure it wouldn't be perfect but it's bound to significantly improve the number of decient games and reduce the number of boring walkovers.

Obviously better stats recording would work even better - that's another job the ut3 developers have to do, particularly for the non DM gametypes - e.g. ons+ for ut2004 ons does a better job of scoring ons.

[Fixxer.Linux.T32]

I've read many of the comment since my first post.
All of those are interesting, but, remember, simply trying to sort the players using their "standard in-game" rankings is far better than simply "randomizing" teams.

Of course, you can't do anything for players switching their seats in front of the computer.
It's planned however, in UT5, to drive rocketted-webcams (TM), that will launch conventional missiles over the players switching in non-authorized way. UT6 do implement depleted-UA minigun turrets in the webcams.
Of course, you could also dream about perfect stats. But, from what I can see from all the games I play is that good players have good stats and bad players have bad stats.
Play against a team of players which score all a party at over 150 pph, within a newbie team, and I can predict that you won't even be able to see the first node that the server will already have switched the next map, because you will have lose in less time that it takes to tell !

To say that this idea is certainly not perfect, but could help better balancing the teams.
And we can imagine also a server switch that would allow to disable this "smart team balancing" for those who are sick about this idea !