Random teams sux!

1
I miss good old capting with pistol duels or teampicker, teams were more even and thanks to this game was more fun.
Now even when we separate me from Randir, they are far away from being equal. ALSO skillboost is not the best solution to this, i could have skillboost 5 and still sux on cargo.
We should create categories and rules for creating teams.

In my opinion it could look like that:
Category 1: Sith, Duo, Csr, Onasi
Category 2: Black, Grab, Alpha
Category 3: Lord, Randir Piston.

Example, we play 3s with Lord, Randir, Black , Grab, Sith and Onasi
Teams should be: Lord, Grab, Sith vs Randir, Onasi, Black
OR Lord, Black, Onasi vs Randir, Grab, Sith
OR Randir, Black, Sith vs Lord, Grab, Onasi

In 2s with Duo, Lord, Black, Alpha the only one option should be Lord, Duo vs Black, Alpha

OFC sometimes it will be unable to make it even: Csr, Black, Alpha, Grab, but at least it will not be overstacked.

If someone got better ideas, post them here.

Re: Random teams sux!

2
Captain duels aren't perfect, either. The guy who won the duel will stack the teams. You generally have to do a 1-2-1 system (or 1-2-2-1 in 4s pugs) to make it fair, but even then, there are still situations in which 1-2-1/1-2-2-1 will give the loser of the duel an advantage.

I could make a tier-based team generation system if there is enough interest. It's been proposed before. My idea is that you would define tiers by rcon (there are basically always admins on, so it shouldn't be a problem), and maybe just do callvote teams 3 3 or something. It would assign each tier a certain # of points, and try to make teams with similar scores. If you try to run the team generator and someone doesn't have a tier assigned, it will alert you so that an admin can assign them to a tier.

By the way, remember that it's possible to specify multiple separations in the vote command. In the example of 6 people that you wrote, you could get even teams by calling the vote:
callvote randomteams 3 3 lord randir black grab sith onasi

You can also just propose swaps in the chat (or call a q vote) before starting.

Re: Random teams sux!

3
Duo wrote: Sat Apr 27, 2019 10:04 pm Captain duels aren't perfect, either. The guy who won the duel will stack the teams. You generally have to do a 1-2-1 system (or 1-2-2-1 in 4s pugs) to make it fair, but even then, there are still situations in which 1-2-1/1-2-2-1 will give the loser of the duel an advantage.

I could make a tier-based team generation system if there is enough interest. It's been proposed before. My idea is that you would define tiers by rcon (there are basically always admins on, so it shouldn't be a problem), and maybe just do callvote teams 3 3 or something. It would assign each tier a certain # of points, and try to make teams with similar scores. If you try to run the team generator and someone doesn't have a tier assigned, it will alert you so that an admin can assign them to a tier.

By the way, remember that it's possible to specify multiple separations in the vote command. In the example of 6 people that you wrote, you could get even teams by calling the vote:
callvote randomteams 3 3 lord randir black grab sith onasi

You can also just propose swaps in the chat (or call a q vote) before starting.
People are lazy and multiple separations aren't often used, sometimes 1 is too much, i remember 2s where i played with Randir, it happened like 2 times and skillboosting us both didn't help at all and we lost ofc. When i said before start these teams are joke i was just ignored and we played it. Yesterday it worked when was me Grab Randir vs Duo Sith Onasi, and we changed it, but it is not happening often. Some people just like to play in stacked team and win easily.
This tier-based team generation system could be the thing im looking for,

Re: Random teams sux!

4
My last memory of a captain duel in siege is probably from 2015: ceasar and black running in circles around boxes on eat shower (yes that is an actual mapname in case anyone forgot), for about 5 minutes. (Map contains a surprisingly large amount of boxes.) I'm not too sad about pistol duel being gone. In CTF we just use cointoss to determine the winner, that way walls stay alive.

Re: Random teams sux!

5
The player base is small enough that in a single session we'll use every combination of players so it's a non-issue to me. Capt duels just add time on to the out, often up to 5 minutes. Also, we are at a point where the margins are pretty small so a couple of errors or one player just having a good game often decides the match.

Re: Random teams sux!

6
I would be open to reworking the current random teams to be "smarter." I'm not sure what our current capabilities are, but I would suggest creating something like elo. Except not. Since every round produces stats, I would suggest taking advantage of this and exporting those to a database, where they are tracked on a player-to-player basis. I would only export relevant ones, such as accuracy, kpm, damage dealt etc... Instead of elo, I would like to suggest a number ranging from 0 to 1 indicating player effectiveness. (where 1 = perfect and 0 = uninstall) Through some simple algorithm, the different (relevant) stats can be combined and normalized to range from 0 to 1. Then, all random teams would need to do is minimize the difference between the (average) effectiveness of each team.

If you really wanted to be extra, you could even implement logic that checks for correlation between players. I.e. the algorithm notices that Duo and Csr in the same team are more effective together than, say Duo and Black. (sort of like how in-phase sinusoidal waves add) Then, if Duo and Csr are a permutation, that permutation would be assigned a modifying scalar (say k) to indicate that particular permutation is stronger, and should be accounted for during randomization.

Re: Random teams sux!

8
I really don't mean to be the bearer of bad news, but such a system will never work. You can call it whatever: elo system, player effectiveness indicator, the logic is more or less the same behind them, and they don't work in team based games. I think we could all list a whole bunch of team based games where there is a ranking system, and then we could easily highlight tons of issues with it.

Just take a look at the stats ceasar mentioned in his post. Accuracy: totally irrelevant. The moment you start spraying to deal additional damage to say a far away target, your accuracy will tank. KPM: this would make sense in a tdm enviroment (maybe), but in siege there are certain classes people need to play, that are not very effective at getting kills: D Techs with shields for example. Damage dealt: once again, certain classes are better at it than others, it does not mean anything. D Jedi on urban won't outdamage a D Demo, and yet he is a great pick against certain comps.

Also maps differ vastly from each other, a unified elo system doesn't really make sense. You could separate elo (or however you want to call it) for different maps, so people would have a different number for cargo than hoth, but that's a lot of work, and I'm not convinced it is worth doing. Because people also have their favourite (and most effective) classes to play, so you would want to track stats for classes separately, except every game is a bit different, and there is no guarantee said player got to play his preferred class in the game.

And then you add personal performance into the mix as well, because let's face it, people perform differently at different times. This is absolutely impossible to account for. I suppose you could decrease their elo then, but that also brings a whole bunch of things into question, like what happens when your teammate refuses to switch from a class that is hard countered by an enemy class. You will lose elo for something out of your control. What siege absolutely does not need is people ragequitting matches because they are afraid of losing precious elo and ego points.

And then at last I'd like to highlight the fact that such a system has existed in CTF since 2014, and it is absolutely meaningless. It is a fully developed ELO system, and it can predict game outcomes based on team scores. Doesn't really work. It predicts the obvious stacks, but fails to be accurate when teams are really close in terms of skills.

I 100% agree with Black's post on this matter.

Re: Random teams sux!

9
Accuracy: totally irrelevant
Maybe in ctf, but I would argue accuracy is much more relevant in siege, albeit, not the best indicator of effectiveness. However, this could easily be remedied by giving it a lower share of the overall effectiveness. Only golan and demp don't require some degree of accuracy. Even then, a properly aimed demp can be devastating compared to a poorly aimed one.
KPM: this would make sense in a tdm enviroment (maybe), but in siege there are certain classes people need to play, that are not very effective at getting kills: D Techs with shields for example.
I beg to differ. KPM is very important, especially for classes like jedi and scout. IIRC, back in the day, the requirement for you to play jedi or scout was 3kpm. If you weren't averaging that, you would be raged at by your team to "gtfo scout." If we played hoth 24/7 like we used to, I could see how that might skew the stat if someone elected to play Jan the entire match. But we don't. I would say (with the excption of grab, who feels tech is always an option, especially against scout and assault) kpm should come out in a wash. With our current map rotation (which we normally play all the way though) your kpm should average out over all maps, because there isn't one class that is ALWAYS required.
Also maps differ vastly from each other, a unified elo system doesn't really make sense. You could separate elo (or however you want to call it) for different maps, so people would have a different number for cargo than hoth, but that's a lot of work, and I'm not convinced it is worth doing.
Averaging over all maps is exactly the point. You want someone on your team who can play all classes on all maps. If you suck on cargo, I think that should be reflected in you effectiveness. This is the whole purpose of an average.
What siege absolutely does not need is people ragequitting matches because they are afraid of losing precious elo and ego points.
Nice euphemism.
And then at last I'd like to highlight the fact that such a system has existed in CTF since 2014, and it is absolutely meaningless.
I would further recommend that our system be linear. IIRC, ctf elo is log-based or something along those lines.

Re: Random teams sux!

10
ceasar wrote: Mon Apr 29, 2019 3:10 pm Maybe in ctf, but I would argue accuracy is much more relevant in siege, albeit, not the best indicator of effectiveness. However, this could easily be remedied by giving it a lower share of the overall effectiveness. Only golan and demp don't require some degree of accuracy. Even then, a properly aimed demp can be devastating compared to a poorly aimed one.

That still doesn't solve the issue. You want to be spraying on many occasions, because every little bit of extra damage helps. And now, you would be penalised for it. Doesn't matter if it has a lower share or not, because you are still tracking it. Also golan does require aim, everyone please stop randomly shooting golan balls everywhere it's such a terrible strat.

ceasar wrote: I beg to differ. KPM is very important, especially for classes like jedi and scout. IIRC, back in the day, the requirement for you to play jedi or scout was 3kpm. If you weren't averaging that, you would be raged at by your team to "gtfo scout." If we played hoth 24/7 like we used to, I could see how that might skew the stat if someone elected to play Jan the entire match. But we don't. I would say (with the excption of grab, who feels tech is always an option, especially against scout and assault) kpm should come out in a wash. With our current map rotation (which we normally play all the way though) your kpm should average out over all maps, because there isn't one class that is ALWAYS required.

You are confusing things here I feel. You claim that good players will naturally have higher KPM. I am not controverting that. I am claiming that having KPM as a stat to determine the calibre of a player is not a good idea. Because if someone wants to be rated higher they need to farm kills. Because it is a stat you track, simple as that. Never mind that the amount of kills can be totally irrelevant to the outcome of the game. (Being in the vicinity of an objective, and instead of shooting the objective you keep shooting at the defence, sounds familiar?) Or how about killstealing. How would you account for that. What if someone keeps playing scout on hoth second obj and gets 1 kill per spawn, but his teammate gets constantly rekt by the defending team. Such a wonderful team player right, he should be rewarded for having a decent KPM.

There are so many cases where going for the kill is not what you want to do. (Like when you try max'ing an enemy player for the next spawn, but he sk's at the perfect time. You didn't get the kill, but was attempting to make a smart play.) And so making it a stat that plays a significant role in determining one's skill level is counterproductive to the game.

Once again, these are issues all team based competitive games face when there are ranks involved.
ceasar wrote: Averaging over all maps is exactly the point. You want someone on your team who can play all classes on all maps. If you suck on cargo, I think that should be reflected in you effectiveness. This is the whole purpose of an average.
Then your system is incapable of suggesting fair teams. People play differently on different maps, which should always be accounted for when making teams. That's why you don't put me and grab together on hoth, but we will have no issues on cargo whatsoever. If you intend to make this system with the purpose of determining the best possible teams, then you can't just ignore this.

ceasar wrote:
onasi wrote: What siege absolutely does not need is people ragequitting matches because they are afraid of losing precious elo and ego points.
Nice euphemism.

Let's not pretend that's not precisely what would happen.
cron