x
ETF2L is looking for new Staff to recruit! Are you interested in supporting the league? Then click here for more details on what we can offer and how to apply! 

Forum

What would make for even teams?

Created 29th August 2012 @ 23:55

Add A Reply Pages: « Previous 1 2 3 ... 5 Next »

¤____________¤

Quoted from Monkeh

Can we add up peoples divisions and make some kind of average?

1 div 3 medic and 1 div 5, 2 Div 4 sollies, 1 div 6 and 1 div 3, 1 div 2 demo and 1 div 4, 2 div 5 scouts, 1 div 3 and 1 div 6.

Blu team=Medic (5) demo (2) Sollies (4 + 3) Scouts (5 + 6) = 25 Divisions
Red team=Medic (3) demo (4) Sollies (4 + 6) Scouts (3 + 5) = 25 Divisions

‘Divisions’ is the new measure of a team’s skill btw.

Something like that might work… spread the skill man, spread the skill.

But you might want to allocate weightings to different classes

ReveraL

My skill ranking solution is based on Nirock Israeli gather bot.
A simple way to put 2 equal teams playing each other is ranking each player by their skill level. Now we all know that in every division there are skill differences between players so we can’t rank 2 players from the same division with the same skill level, for example:
demoman1 = numlocked -> prem
demoman2 = flinn -> prem
even tho they are both very good demomans and both are playing in prem numlocked is still have the upper hand. So to solve this problem we need what is called a “Dynamic ranking”.
Each player got a rank from 100-1800 that automatically change after every pickup.
Winning team get +X points for each player (different points for each player based on the player k/a/d).
Losing team get -X points for each player, again different points for each individual.

Now the first rank to each player has to be done by hand, an admin has to go through the entire player list and rank everyone.

as for the players division, we will have:
* medic1, demoman2, scout1, scout4, soldier1, soldier4
* medic2, demoman1, scout2, scout3, soldier2, soldier3

When we have that kind of ranking we have more diversity and equal pickups.
Lets make 2 teams of prem+d1 players that based on this ranking system for example:

numlocked – 1800
hys – 1400
F2 – 1700
Kritzonite – 1300
shocky – 1800
wltrs – 1700
tack – 1400
gubbins – 1300
Mike – 1700
Forsaken – 1300
daleth – 1400
Tek – 1500

(*rankings made by me for this example*)

Team1 – F2, hys, shocky, gubbins, Mike, Forsaken (team rank: 9200(all players ranks together)/6 = 1533)
Team2 – Kritzonite, numlocked, wltrs, tack, Tek, daleth (team rank: 9100/6 = 1516)

Now, the ranks are “starting ranks” that’s means that after every game they will change to make them more realistic.

That’s all, Thanks to Nirock for his awesome bot :D

AnimaL

dunno how pu2 does ballance but i assume its smth like reveral said and there is issue with that… often bot ballances couple of bad players with adding more prem players with them in same team and end up playing against stable d1 that doesnt have any prem players and they get rolled by those on class prem players


Last edited by AnimaL,

Squirry

nNT ≠

Umm one mroe suggest, in the moment to add classes, pick like a virtual division. I mean that.

!add scout div3
!add medic div5
!add demo div3

And get like a random selection, is hard to explain it really, but if people use it well, we could get like a balanced pickup really, and so much combinations.
Reveral’s thing is just like what I said, or kidn of, instead points, divisions. Indeed we can use divs to assign points.

For example

div6 <1000 points
div5 <1200 points
div4 <1400 points
div3 <1600 points
div2 <1800 points


Last edited by Squirry,

Trath

Quoted from ¤____________¤

I think you are mistaken here. It’s been a long time since I did stats but I think you essentially have two groups of 4 and want to pick 2 from each group, in such a way that the order in which they are picked does not matter (making it a combination not a permutation – which is what you are thinking along the lines of). The mathematical function for this is 4C2 (at least that is how it appears on a calculator :P)

4C2 = 6 (ie 6 ways of picking 2 items from a total of 4), but you need to divide this by 2 as the pair left over is the same as some of the possibly pairs picked.

Hence 3 ways of splitting the 4 scouts (1 and 2, 1 and 3, 1 and 4 and the corresponding counter pairs).
Then once the scouts/sollys are in groups of two you have 2 different ways of pairing up the pairs. Making there 6 possibilities overall.

Now, the above might be wrong, haven’t done stats for like 4 years now, but I think it’s right.

No, because I’m not pairing only the scouts up, I’m pairing Scouts and Soldiers together, at the same time, to make for teams rank average as close as possible.
I think we might both be wrong.

Also, I was thinking of giving weight to each class as well, but that is the least of my concerns as I don’t think it will be needed, and if it will – it would be super easy to add/implement.

Quoted from Monkeh

Can we add up peoples divisions and make some kind of average?

one div 3 medic and one div 5, two Div 4 sollies, one div 6 and one div 3, one div 2 demo and one div 4, two div 5 scouts, one div 3 and one div 6.

Blu team=Medic (5) demo (2) Sollies (4 + 3) Scouts (5 + 6) = 25 Divisions
Red team=Medic (3) demo (4) Sollies (4 + 6) Scouts (3 + 5) = 25 Divisions

‘Divisions’ is the new measure of a team’s skill btw.

Something like that might work… spread the skill man, spread the skill.

Have you not read any of the comments? that is exactly the direction this is going, getting team rank average as close as possible.

ReveraL

Quoted from AnimaL

dunno how pu2 does ballance but i assume its smth like reveral said and there is issue with that… often bot ballances couple of bad players with adding more prem players with them in same team and end up playing against stable d1 that doesnt have any prem players and they get rolled by those on class prem players

maybe it’s working the on the same ranking system, but it doesn’t use the dynamic ranking. changing the ranking after each and every game will make more stable teams after few games!

Trath

Quoted from AnimaL

dunno how pu2 does ballance but i assume its smth like reveral said and there is issue with that… often bot ballances couple of bad players with adding more prem players with them in same team and end up playing against stable d1 that doesnt have any prem players and they get rolled by those on class prem players

What ReveraL said is exactly what I am going for / trying to copy. However he is wrong on the way he thinks the teams are divided, I’m pretty sure Nirock made it way better / complicated than that.

Nirock is a nice guy, I asked him for his code for dividing teams, but he won’t give it to me cause he wants me to push and do it myself.

Anyways WTF spent so much time reading and writing on this thread. If anybody that reads all my comments can help me with the programming side, please type your way here.

envy

Quoted from Squirry

Umm one mroe suggest, in the moment to add classes, pick like a virtual division.

!add scout divprem
!add scout divprem
!add scout divprem

Squirry

nNT ≠

Quoted from envy

[…]
!add scout divprem
!add scout divprem
!add scout divprem

Thanks for your collaboration.


Last edited by Squirry,

Monkeh

.:ne:.
.:ne:.

Quoted from Trath

Have you not read any of the comments? that is exactly the direction this is going, getting team rank average as close as possible.

Yes I read them all and no-one has suggested adding up players ‘divisions’ and aggregating them…hence why I suggested it.

Nicolas Cage

BM

Why not implement an ELO system for each player and class?

You may know it from the MGE stats, chess ranking or League of Legends. It’s self correcting (it will take a while to get accurate but it may work). For instance my ELO would be different for medic, demo, solly and scout. When I sign up as a medic I use that ELO. That way people that aren’t registered in a particular division or maybe played div 3 a few years ago but are not at that level anymore aren’t pigeonholed by a division.

Once you’ve got a bunch of ELO’s, pair up the teams randomly so as to keep the total ELO on each team as similar as possible.

The biggest downside I see is that it would require people to play regularly or the system would break down. Start everyone off at 1000 ELO and if nobody is truely active, you’ll never see ‘true’ ranks.

Another downside would be: how much does a ‘high ELO medic’ have on the game when compared with a ‘top ELO demo’?

A relatively neat feature could be to have a website with ELO ranking for each class – something that would push those players who want to see their name in lights to play.

Trath

Quoted from Monkeh

[…]
Yes I read them all and no-one has suggested adding up players ‘divisions’ and aggregating them…hence why I suggested it.

hmm your right, my bad – it wasn’t mentioned in this thread. Aggregating was mentioned, and so was ranking, but I guess you couldn’t understand what I was meaning if your not from the Israeli Pickups.

Its what ReveraL said, the ranking is with big numbers, that generally represent divs as in:

prem = 1800
div1 = 1600
div2 = 1400
div3 = 1200
div4 = 1000
div5 = 800
div6 = 600

since there are differences in divisions themselves (top and bottom), the numbers represent the top.

So for example, in scout rank:
Stefan, Cookye, Shintaz, Quad = 1800
Flippy, Bybbenn, Wltrs = 1700

Quoted from Nicolas Cage

Why not implement an ELO system for each player and class?

Quoted from Trath

Also, the plan is to have players rank change according to win/loss and Team1RankAverage divided by Team2RankAverage.
So the initial skill rating isn’t too big of a deal, because over more and more matches, it will make itself more right. (Expectancy to win) http://en.wikipedia.org/wiki/Elo_rating_system

nTraum

\V/ Gold
LAME

Still, this is no reason to not use ELO. It is proven to be good in chess, cs, lol & other (epsort) tournaments. It is easy to use, it’s easy to implement and easy to adapt to your needs. Inventing your own ranking system will result in nothing but worse balancing than you’d achieve with mature systems as ELO.

/thread

Edit: of course you’ll need to modify it to get it work in 6s (instead of the 1on1), but it’s as easy as any other linear mathematical optimzation. You’ll have the best balancing when the difference bewteen the ELO points of both teams (which equals the sum of the players of that team) is minimal.


Last edited by nTraum,

Squirry

nNT ≠

At the moment the last idea about ranking, or ELO, is perfect. I think if you get a talk with Nirock, he’d be happy to help us. And hopefully, in a near future, would be more fun, and more balanced pickups :)

atmo

How can you apply Elo, a 1v1 rating system, to a team-based game, especially where players generally have less than 10 matches?

Add A Reply Pages: « Previous 1 2 3 ... 5 Next »