micr0c0sm

Theorem: Whichever team has more unregistered players loses.

Data: 24/27 games I have played in have satisfied this theory.

Observations: This is regardless of individual unregistered player skill, such as the time 'Maps and Atlases' and two other unregistered players had been on the same team where the other team had one unregistered player. Maps' team still lost even though he clearly exhibited what I would call skills.

Analysis: Unregistered players are not used to the level of teamwork, pace of play, and type of play on TG's server. Regardless of individual skill or NS experience when they play on Tactical Gamer they are forced not only to keep up with our level of play and teamwork, but also our rules and play styles which may have become natural to us but are by no means second nature to the newer player.

Conclusion: Two possible changes to the balance plugin, each with advantages and drawbacks.

1) Unregistered players are not recorded in the balance db, and always count as .1. The advantage is it's real easy to implement and code. The disadvantage is that they still have a 'value' :-p
2) Registered players that have played at least 10 games are balanced, then the remaining registered players are amx_randomed, then unregistered players are amx_randomed. The advantage is it is for the most part more fair. Disadvantage is 'harder' to code.

Either is no where close to perfect, and frankly, I wouldn't want perfectly balanced teams every game, but I think it would be significantly better than it is now, and good enough for most of the games. No matter how you slice it no plugin will perfectly be able to balance teams all the time, you can only hope for it to be fair (contrary to what Fox News says fair & balanced are not unseperable).

Agamemnon

The first option, setting them to 0.1, would have fairly disastrous results whenever there are an odd number.

The second option, if done backwards, is quite easy to do, and has no problems in terms of conflicts with the algorithm. All we would need to do is consider every ? to be an unknown player, instead of just those with fewer than 10 games. (Unimportant explanation: By "done backwards" I mean that it assigns new players first, then balances known players. This is done so the algorithm knows how many known players it is assigning to each team.)

That said, I would be a little uneasy about the incentives created by doing this, given that there is currently no real difference between ?'s and non-?'s who aren't opted in to gb.

Also, do be aware that this would be throwing out any hope of learning the skill levels of various ? players. One really good ? will keep the teams unbalanced indefinitely if this change is made. The upside is that this would mostly do away with ? stacks, where you wind up with significantly more ?'s on one side than the other (the only potential complicating factor is non-?'s with fewer than 10 games, who are also be counted as new). We could also up the threshold for a ? to be considered known to 50 games, but keep it at 10 for registered players.

rs_al

I have a feeling that if teams with more ? do tend to lose, it's the regulars on the team doing that are doing most of the losing. It doesn't take much to see that ? players, regardless of skill, playtime on TG, or any other factor, are treated like children and kicked (or threatened) for silly reasons. A team that refuses to communicate with several of its members (or communicates "at" them instead) is probably going to do worse than a team that communicates with all of its members.

So long as the ? icons stay on the scoreboard and regs are playing, the icons will contribute to the self-fulfilling prophecy micro theorizes.

Stupid

Really? That's pretty sad.

__________________
Former TGNS admin until WoW blinded me with flashy lights.

Flarfignuggen

That about covers it right there.

micr0c0sm

al, when do you play? I agree with the communicate 'at' them part but not with the 'being kicked for silly reasons' part.

And the difference between a reg'd with less that 10 games and an unreg'd is there ag, thats why reg'd get amx_randomed first and then *another* amx_random is done for unregistered players. This puts reg'd players squarely above unreg'd players. Can you also explain your 'fairly disastrous results' ?

Maybe I do not have a firm grasp on the inner workings of the balance plugin, so here is how i see amx_fair:

Database: Registered player list with 10+ games and therefore a valid value (which is just thier w/l).

Scenarios:
1: all in rr, amx_fair called
Sort valid value list, assign by alternating.
amx_random registered.
amx_random unregistered people.

2: some people joined teams, amx_fair called
calculate value of each team by summing valid values.
sort rr list of valid values.
place top valued player on team with lower value.
recalculate value until previous weaker team is stronger.
*SWITCH THE LAST ASSIGNED PLAYER*
proceed as scenario 1.

micr0c0sm

Although It would be nice to figure out the *reasons* that a team with more ?s do worse that a team without the simple fact remains that balance is not working. My goal is to fix balance right now with that limited data, and another thread for how to treat ?s can be made but honestly if you don't know how to then I think you oughtta re-read some of the intro docs to this community.

Agamemnon

If you give an unkown player a value of 0.1, the balance plugin will do what it needs to to get a fair game (0.5). Essentially, you'll wind up with all the best players on the same side as the extra unknown player. As I don't think that the seven best players on the server + one ? against the bottom eight known players is a very good result, I would be opposed to this solution.

This is very different from the methods used by the current plugin. I won't try to explain the math here, but it is possible to incorporate this idea without throwing the current algorithm overboard, and all you have to do is flip the order in which you're doing things. First split unregistered, then split new registered players, then run the existing algorithm on the known players. The reason it needs to be done in that order is that the algorithm doesn't limit itself to just alternating, which would yield poor results. Instead, it can assign any number of players to one side, until it hits the number of players that should be on that team. Because of this, it needs to know before it begins how many known players it will place on that team, which means it needs to assign the unknowns first.

I would prefer to use the existing algorithm with the above changes, as a simple alternation will not yield good results, and splitting the unreg'd and then new reg'd players before running the existing algorithm accomplishes what you want to do without abandoning the more effective algorithm.

micr0c0sm

Cool, so basically what amx_fair would be is:

amx_random(* FROM unregistered)
amx_random(* FROM registered WHERE ((wins+losses)<10))
amx_balance(* FROM readyroom)


Can we try it?

micr0c0sm

bump for the best idea ever

Pokerface

It's on our plate. Trouble is, so are 150-gabillion hour work weeks for Wyzcrak.

__________________

NS Game Officer. TF2 Admin. BF2 Admin / Scripter. PM with issues.
Tempus: Pokerface is nailing it right on the head. Everyone who is arguing against him is simply arguing against reality.
<anmuzi> it is not permitted to have privacy or anonymity
<LazyEye> yeah when I play on TG the server digs though my trash
Arm yourself with knowledge: TG NS TF2 BF2

micr0c0sm

rgr tht, if you send me the plugin source maybe I could figure it out and test it!

Donat

balance command should be preceded by the ready room command, and people shouldn't switch or it defeats the whole purpose of the command imo lol,

__________________

Pokerface

But that isn't required at all, with the plugin as it's written. Unless all the best players immediately stack one side, the plugin will figure out how to spread people out. It takes into account the players presently on a team in determining who goes where.

Balance can be executed with two people in RR 20 minutes into a game with no ill effects.

__________________

NS Game Officer. TF2 Admin. BF2 Admin / Scripter. PM with issues.
Tempus: Pokerface is nailing it right on the head. Everyone who is arguing against him is simply arguing against reality.
<anmuzi> it is not permitted to have privacy or anonymity
<LazyEye> yeah when I play on TG the server digs though my trash
Arm yourself with knowledge: TG NS TF2 BF2

Eternaly_Lost

I could see problems with this. We have ? mark who play here semi often who are skilled, and then we have ? who just installed the game. Random worked the best when all players being random are about equal. TGNS has the problem in that our player vary in skill more then most servers. By recording wins, we can make predictions on how someone will do. With you idea, ? marks would be split, with a chance that 2 very good ? could end up on the same team.

Then, you do the same for uncommon regs.

Then you try to balance from a much smaller pool, with out paying attention to the players on the teams already (? count as .5 ). So to balance it, it would give equal skill players to both sides, resulting in a stack on one team, about every 1 of 2 games with just 3 ? people, assuming 2 good ?, not unreasonable I see it every so often (being about once every day or more ). 1 good player on a team, 50% chance that the next player is also good.

The end result, very bad rounds.

Balance was made to help over time.

The biggest flaw I could see would be time wise. (TGNS @ prime time is much harder on teams cause everyone is both skilled and a team player , vs TGNS @ most other times which is minus the skill. )

If we add a time element to balance it might do better.

For example think of Player A.
Durring the day player A is very good at wining, but at night he is only average at being the major factor at win.

Games during the day will make him look valuable, that he is durring the day, but once night rolls around he is not as good as his numbers say, and they drop until he stops. Next day, he throws off the balance again because he is listed as worse then he is.

__________________





Current game name : |TG| Lost the Phantom Thief