Jump to content

Concerning Pgis Elo Matchmaking Approach


31 replies to this topic

#21 AlexEss

    Member

  • PipPipPipPipPipPipPipPipPip
  • FP Veteran - Beta 1
  • FP Veteran - Beta 1
  • 2,491 posts
  • Locationthe ol north

Posted 31 December 2012 - 10:17 AM

i still think the second post have a point tho... We do not know exactly how the ELO will be balanced but there is a good chance that it will in it's final form use more then W/L as a data point as there is a ton of data on us as players that we never see.

#22 IceSerpent

    Member

  • PipPipPipPipPipPipPipPipPip
  • 3,044 posts

Posted 31 December 2012 - 10:17 AM

View PostJPsi, on 31 December 2012 - 08:48 AM, said:

There's definitely argument to include in a BV system, but if its absolutely necessary is unclear.


It is absolutely necessary, as it's the only way to account for different loadouts. A player fielding a "fully pimped ride, complete with chrome spinners" should not be considered a match for a player that uses a stock version of the same mech (or some sort of "for the lulz" build for that mater) - the difference is large enough to completely mess up the balance.

#23 p4r4g0n

    Member

  • PipPipPipPipPipPipPipPip
  • Knight Errant
  • 1,511 posts
  • LocationMalaysia

Posted 31 December 2012 - 10:19 AM

View PostLarkinOmega, on 31 December 2012 - 09:38 AM, said:

This analogy is so funny that I cannot resist replying. 5% of the time, sure, but the other 95% of the time you'd be sitting on your rear in 1 inch of water!


Its an oft quoted analogy to demonstrate the occasional deceptiveness of averages. You must surely have come across it before?

#24 Bloodred

    Member

  • Pip
  • 19 posts

Posted 31 December 2012 - 11:00 AM

What I'm worried about most regarding this system is the fact that it does not take into account the 'mech you're piloting (and the equipment you're using). What if I am a great Atlas pilot but I also frequently play a Jenner and almost never do any good in that? Judging by what has been described my ELO would most likely end up somewhere in the middle between the higher "Atlas ELO" and the lower "Jenner ELO", this would essentially mean that after enough games no matter which 'mech I drop in I'll either be too good for the match or too bad for it, this would ruin the balance of the games I'm in since my ELO is misleading despite being correct (according to the system).

#25 Red squirrel

    Member

  • PipPipPipPipPipPipPipPip
  • Legendary Founder
  • Legendary Founder
  • 1,626 posts

Posted 31 December 2012 - 11:03 AM

Also I think it might make more sense to use the median than the mean elo to determine the win/loss probability of a team.

View PostKousagi, on 31 December 2012 - 08:55 AM, said:

Plus we all know, anything EA touchs turns to crap. RIP Bioware.


ROTFL
http://en.wikipedia.org/wiki/Midas

Edited by Red squirrel, 31 December 2012 - 11:05 AM.


#26 Pugastrius

    Member

  • PipPipPipPipPip
  • Bad Company
  • 196 posts
  • LocationOn Top of Your Dead Mech

Posted 31 December 2012 - 11:50 AM

The fact is, they completely screwed up... Why they didn't just steal the math that StarCraft II uses is beyond me.

The SC2 model pits each player against all of the other players and their performance is based on total score. This pairs perfectly with MWO... in that you are already ranked at the end of the game.

So using the screenshot provided earlier:
Posted Image

1) The top of the list will have his personal ELO go up and all 15 other players go down.
2) 2nd Player: Goes down once (losing to the top rank) and goes up against the 14 other players.
3) The Highlighted Player: Goes down 4 times (to the people above him) and goes up 11 times.
So on and so forth...

This has serious advantages to the "average" method being employed by PGI:
1) It allows for 15 ELO changes per match, making it much more accurate much quicker.
2) Allows you to be pitted against your team mates
3) Allows your contribution to cause your ELO to go up even if your team gets crushed

Anyhow, their implementation is going to be very very very slow to work (if it even does)

Quote

Also think it might make more sense to use the median than the mean ELO to determine the win/loss probability of a team.


When employed properly (where teams don't really change between matches, WoW arena matches for example) the Math for ELO in team games doesn't work using Medians.

Edited by Pugastrius, 31 December 2012 - 11:54 AM.


#27 Red squirrel

    Member

  • PipPipPipPipPipPipPipPip
  • Legendary Founder
  • Legendary Founder
  • 1,626 posts

Posted 31 December 2012 - 12:05 PM

View PostPugastrius, on 31 December 2012 - 11:50 AM, said:

The fact is, they completely screwed up... Why they didn't just steal the math that StarCraft II uses is beyond me.

The SC2 model pits each player against all of the other players and their performance is based on total score. This pairs perfectly with MWO... in that you are already ranked at the end of the game.

So using the screenshot provided earlier:
Posted Image

1) The top of the list will have his personal ELO go up and all 15 other players go down.
2) 2nd Player: Goes down once (losing to the top rank) and goes up against the 14 other players.
3) The Highlighted Player: Goes down 4 times (to the people above him) and goes up 11 times.
So on and so forth...

This has serious advantages to the "average" method being employed by PGI:
1) It allows for 15 ELO changes per match, making it much more accurate much quicker.
2) Allows you to be pitted against your team mates
3) Allows your contribution to cause your ELO to go up even if your team gets crushed

Anyhow, their implementation is going to be very very very slow to work (if it even does)



When employed properly (where teams don't really change between matches, WoW arena matches for example) the Math for ELO in team games doesn't work using Medians.


First they need to fix the XP system since some roles greatly benefit from the current system.
Also getting the killshot does not make you the better player. On the other hand dealing 10dmg to each mech (8assists) does not make you very useful either

#28 Lyteros

    Member

  • PipPipPipPipPipPip
  • 456 posts
  • LocationGermany

Posted 31 December 2012 - 12:18 PM

View PostRed squirrel, on 31 December 2012 - 12:05 PM, said:


First they need to fix the XP system since some roles greatly benefit from the current system.
Also getting the killshot does not make you the better player. On the other hand dealing 10dmg to each mech (8assists) does not make you very useful either


8 assists alone will not get you on top. The more information you incorporate, the less effect from the possibility to abuse (or actual abuse of) single variables.
The thing is: the better the end of game calculation and reward system is working, the better / easier it is to simply add it into ELO calculation. On the other hand it's quite hard to find the "right values" to take into account, without ending up getting into freaking large calculations.

The more your personal performance influences ELO instead of the team performance, the more accurate your ELO gets and this improves the game experience for everyone.

Edited by Lyteros, 31 December 2012 - 12:22 PM.


#29 Davers

    Member

  • PipPipPipPipPipPipPipPipPipPip
  • Legendary Founder
  • Legendary Founder
  • 9,886 posts
  • Facebook: Link
  • LocationCanada

Posted 31 December 2012 - 01:03 PM

Also Pul said in the article: "These include a more defined player skill rating and a Mech weight class balancing system." So it's not just Win/Loss ratio being taken into account.

#30 Tasorin

    Member

  • PipPipPipPipPipPipPip
  • Legendary Founder
  • Legendary Founder
  • 825 posts
  • LocationCartman 3050 HQ

Posted 31 December 2012 - 01:27 PM

Based on PGI's history of launching new game mechanics, the chances are that the ELO system is going to be really bad at first and get better over time as PGI invests more hours of development and implementation cycles in the ELO system.

What we have been asking for is a lobby and the ability to "set up matches" with some variant of tonnage or battle value system in order to have a level playing field in 8 mans for those of us who want organized competitive play.

PGI has about six months to figure our CW and whatever matchmaking system they settle on while fixing by rebuilding the net code from the ground up and dealing with economy and mek frame and weapon balancing.

No short order for 9 out of 10 development houses.

Good luck PGI, I personally would like to see you succed and give us a Galaxie to rule.

#31 Red squirrel

    Member

  • PipPipPipPipPipPipPipPip
  • Legendary Founder
  • Legendary Founder
  • 1,626 posts

Posted 31 December 2012 - 03:11 PM

View PostLyteros, on 31 December 2012 - 12:18 PM, said:


8 assists alone will not get you on top. The more information you incorporate, the less effect from the possibility to abuse (or actual abuse of) single variables.
The thing is: the better the end of game calculation and reward system is working, the better / easier it is to simply add it into ELO calculation. On the other hand it's quite hard to find the "right values" to take into account, without ending up getting into freaking large calculations.

The more your personal performance influences ELO instead of the team performance, the more accurate your ELO gets and this improves the game experience for everyone.


Generally I like your idea. But I see one flaw:
People are damn good at finding the loophole in how to get better stats. We already see all those people in PUG games that go for the killshot whatever or Alt-F4 before dying. Or remember all those suiciders?
In the end the XP mechanism needs to encourage teamplay but prevent exploition.
I am just not sure if this is possible since in many cases it is difficult to tell whether someone died for his teammates or because he did something stupid.

#32 Theevenger

    Member

  • PipPipPipPipPip
  • 194 posts
  • Facebook: Link
  • LocationCalifornia

Posted 14 January 2013 - 10:07 PM

View PostLyteros, on 31 December 2012 - 07:24 AM, said:

This post is based on this Dev statement:
http://mwomercs.com/...ost__p__1626065


The Problem I see:
From what I read here, the new matchmaking will entirely be based on win / loss statistics.

And those statistics will be created by the WHOLE team, not taking personal achivements of each pilot into account. So your personal influence on what happens with your ELO is 1/16 (since you're one single pilot of the team and the whole other team wants your head)

Since your teams (when PUG) are randomly drawn, the team based ELO rating is just as much as a gamble as the current system.

Spoiler


I think the system will not solve the mess we have with the matchmaking. The proposed ELO misses a lot of available and easy to use information, which can vastly increase the effectivity of the ELO calculation and the match making.

Some simple examples:
Spoiler




Proposed Changes / additions:
Spoiler




Please keep in mind: This is about improvement and better game experience for everyone, not whining / rabble or personal advantages. Discuss.


What about having a separate ELO rating for each of your Mechs, or at least per chassis? At the very least, one per weight class.





1 user(s) are reading this topic

0 members, 1 guests, 0 anonymous users