Jump to content

Whatever Happened To Weight Matching?


29 replies to this topic

#21 Hayashi

    Snowflake

  • PipPipPipPipPipPipPipPipPip
  • Bridesmaid
  • Bridesmaid
  • 3,395 posts
  • Location輝針城

Posted 07 April 2013 - 07:26 PM

ELO is supposed to put your win/loss ratio at 1. Balance is supposed to make games close matches.

At the moment, it's still alternating curbstomps in pubs, and the WLR is high enough above 1 to show that it isn't even working to alternate stomprate at 1:1 either.

Posted Image

As you can see the ECM chassis have a significantly higher WLR than the others showing a pretty clear case of the kind of advantage ECM gives your team that even ELO cannot erase.

Strangely it seems the 'lousiest Hunchback' is significantly better than almost all lights though. If the other team was given a non-ECM spider for my Hunchback, this would be enough to push the game in my team's favour already.

I don't see why we can't have both - the ELO mechanism used as a second sorting mechanism after first applying the mech class filter. I mean, we could very well just decide on the players for a match by classed pairs (8 pairs), assign the players to respective teams based on their ELO so that the average ELO on each team is identical, and only add players in pairs so that if one side has 7, the other will also drop with 7. If a 4-man group is dropping, then they'll all be on the same side; if they're all very good players, then have 4 good players ELO wise on the opposing team, and seed the rest to keep overall ELO the same (which means that people attempting to bias winrates by dropping with skilled friends will end up getting the worst players around as their 4 teammates. The bad players can then get to learn from the better ones while simultaneously being enough of a liability in that game such that we don't get a pubstomp situation. This is easier to implement - ALL of the required stats and information needed to do this kind of balancing already exist.

Alternatively, we can look into multiplying people's ELO stats with the BV of their loadout, and matching pairs of people with similar weighted BV, and deciding which guy goes into which team by keeping the total weighted BV as close as possible. I would consider this more as an ideal solution, but it'll entail extra work in assigning BV for every piece of equipment - following Sarna values wouldn't be sufficient as some things are differently implemented here (for instance ECM should have a sky-high BV addition since it has so many more functions here than lore ever gave it). This ideal solution has the advantage that it would allow 8-mans to drop against PuGs and still have fair fights in terms of the overall skill and equipment mix, so friends can play together instead of having to split nonstop. To seal the discrepancy on the side of VoIP, addition of properly integrated VoIP in the client itself (such that players by default can hear any teammates who have microphones), and a better chatwheel system in the style of Left 4 Dead would allow for functional quick chat even by players without microphones.

This current mechanism of attempting to match one player's ELO to another player's ELO on the other team, disregarding team averages, and not caring about mech classes is just bad practice. And adding players one by one, instead of in pairs, creates this annoying 7-8 situation.

I mean, if there were two of me, and one of each of us is added to each team but one piloted a CPLT-K2 while the other pilots a COM-3A, the result of the game will be more or less obvious. ELO is going to be very imprecise unless it is somehow weighted by the efficiency of the build, either by filtering them in series, or by mathematically combining them into a single value.

Edited by Hayashi, 07 April 2013 - 07:53 PM.


#22 Traigus

    Member

  • PipPipPipPipPipPip
  • 303 posts

Posted 07 April 2013 - 07:56 PM

I don't think they are going to bother to fix it.

Whatever they are doing with the new lobby system probably makes and change DOA and a waste of programmer time.

OR

It is doing what it is supposed to but the "good" end result is still waiting on a feature or 2 to be added on.


it is so random it is kind of like a slot machine

I had 2 matches 200 tons down (BEFORE disconnects) today... I was on a conquest team that lost.. that was roughly 200 tons heavy. The lightest mech was me in a pretty slow Cat (f). They had 6 (yes 6) 3Ls a X-5 and a Stalker (of all things).


..and 2 8 v 6 (again before disconnects) I think I had one on each side.

#23 Roadbuster

    Member

  • PipPipPipPipPipPipPipPip
  • Legendary Founder
  • Legendary Founder
  • 1,437 posts
  • LocationAustria

Posted 08 April 2013 - 04:14 AM

Pre ELO matchmaking - match result 8:0 or cap win
Post ELO matchmaking - match result 8:0 or cap win

Working as intended...


EDIT:
I forgot to mention weight matching. I love playing conquest on Alpine or Tourmaline with 8 heavy-assault vs a team with 4+ light mechs.

Edited by Roadbuster, 08 April 2013 - 04:18 AM.


#24 Felipe Rigueira

    Rookie

  • 6 posts

Posted 08 April 2013 - 04:19 AM

Let's just put all the Atlas on one team and all the spiders on the other.

#25 Skunk Wolf

    Member

  • PipPipPipPipPipPip
  • FP Veteran - Beta 1
  • 286 posts

Posted 08 April 2013 - 04:55 AM

Anybody else notice that the Hero Mech number affects it too? It's getting silly. I look at the friendly roster, see that we have an Awesome, a Stalker and Fang. I brace for two ECM Atlases and Heavy Metal.

#26 Star Captain Obvious Kerensky

    Member

  • PipPipPipPipPipPipPip
  • Bad Company
  • Bad Company
  • 500 posts

Posted 08 April 2013 - 05:28 AM

I have had many more close matches under the Elo system, I'll give it that. But since the vast majority of matches I play in have a vast tonnage disparity, I cannot give total credit to the Elo system for making the matches close.

I like playing against more skilled opposition, but it has some serious drawbacks....

I still get the moron rambo players. Which means you're playing a man down on a regular basis against opposition that is going to be fairly skilled. Under the old system skills were random, so playing 6 on 8 wasn't nearly as difficult to pull off an underdog victory.

The system discourages veteran players from grouping with new players. Putting one newbie on our 3 man team of vets consists of the newbie being way-behind skill curve and not being able to survive long, no matter how much we try to protect him.

I see the same mech configs every match. Veteran players gravitate to the same play style and same mech configurations. It gets repetitive.

All we managed to do was trade tonnage balancing for skill balancing. Skill is supposedly better, but being down 300-400 tons isn't going to be a close game.

Edited by Eldragon, 08 April 2013 - 05:28 AM.


#27 JuiceKeeper

    Member

  • PipPipPipPipPip
  • The Territorial
  • The Territorial
  • 172 posts

Posted 08 April 2013 - 05:29 AM

i am playing with this idea for a bit. what if we would get matched by not just weight but also by CBills limit? I was thinking about this becouse like Hayashi said when u have same tonnage and game tries to match one raven against another one and one got ECM and other one doesnt. It harms team at current game state. Becouse its obvious that that second non ecm raven will dont have sufficient fire power or anything else special to counter that bonus of ECM mech.
But if game would matchmake based on Mech values and their sorta average values or something similar. It might help balance matches quite nicely becouse even tonage of those 2 ravens are same their value is completly different ECM mech at least lighs are much more expensive more in line with pricing of mediums then their normal friends.
So if there would be a difference in teams based on whole value of group. Game would just try to find players which are running expensivier builds and so on.
Like prize of 2xERPPC 1xgaus 3D will never be same like prize of 2xML 2xLL 1Gaus 3D.

#28 Ralgas

    Member

  • PipPipPipPipPipPipPipPip
  • Overlord
  • Overlord
  • 1,628 posts
  • LocationThe Wonderful world of OZ

Posted 08 April 2013 - 05:32 AM

Elo + weight matching = 15+ min ques in some brackets. That why it got wound back to what it is now

#29 Ghogiel

    Member

  • PipPipPipPipPipPipPipPipPipPip
  • CS 2021 Gold Champ
  • CS 2021 Gold Champ
  • 6,852 posts

Posted 08 April 2013 - 06:48 AM

View PostHayashi, on 07 April 2013 - 07:26 PM, said:

ELO is supposed to put your win/loss ratio at 1. Balance is supposed to make games close matches.

At the moment, it's still alternating curbstomps in pubs, and the WLR is high enough above 1 to show that it isn't even working to alternate stomprate at 1:1 either.

Posted Image

As you can see the ECM chassis have a significantly higher WLR than the others showing a pretty clear case of the kind of advantage ECM gives your team that even ELO cannot erase.

Strangely it seems the 'lousiest Hunchback' is significantly better than almost all lights though. If the other team was given a non-ECM spider for my Hunchback, this would be enough to push the game in my team's favour already.

I don't see why we can't have both - the ELO mechanism used as a second sorting mechanism after first applying the mech class filter. I mean, we could very well just decide on the players for a match by classed pairs (8 pairs), assign the players to respective teams based on their ELO so that the average ELO on each team is identical, and only add players in pairs so that if one side has 7, the other will also drop with 7. If a 4-man group is dropping, then they'll all be on the same side; if they're all very good players, then have 4 good players ELO wise on the opposing team, and seed the rest to keep overall ELO the same (which means that people attempting to bias winrates by dropping with skilled friends will end up getting the worst players around as their 4 teammates. The bad players can then get to learn from the better ones while simultaneously being enough of a liability in that game such that we don't get a pubstomp situation. This is easier to implement - ALL of the required stats and information needed to do this kind of balancing already exist.

Alternatively, we can look into multiplying people's ELO stats with the BV of their loadout, and matching pairs of people with similar weighted BV, and deciding which guy goes into which team by keeping the total weighted BV as close as possible. I would consider this more as an ideal solution, but it'll entail extra work in assigning BV for every piece of equipment - following Sarna values wouldn't be sufficient as some things are differently implemented here (for instance ECM should have a sky-high BV addition since it has so many more functions here than lore ever gave it). This ideal solution has the advantage that it would allow 8-mans to drop against PuGs and still have fair fights in terms of the overall skill and equipment mix, so friends can play together instead of having to split nonstop. To seal the discrepancy on the side of VoIP, addition of properly integrated VoIP in the client itself (such that players by default can hear any teammates who have microphones), and a better chatwheel system in the style of Left 4 Dead would allow for functional quick chat even by players without microphones.

This current mechanism of attempting to match one player's ELO to another player's ELO on the other team, disregarding team averages, and not caring about mech classes is just bad practice. And adding players one by one, instead of in pairs, creates this annoying 7-8 situation.

I mean, if there were two of me, and one of each of us is added to each team but one piloted a CPLT-K2 while the other pilots a COM-3A, the result of the game will be more or less obvious. ELO is going to be very imprecise unless it is somehow weighted by the efficiency of the build, either by filtering them in series, or by mathematically combining them into a single value.

Your W/L ratio overall is not even 1.5.

Your ECM equiped mechs barely have better stats than non ECM mechs.

The raven 3L being better than it's other variants is a given, as far as lights go, the 2X and 4X are about the worst mechs in the game anyway, and no amount of ECM would even the playing field for them.

#30 Xenodraken

    Member

  • PipPipPipPipPip
  • The Blazing
  • The Blazing
  • 133 posts
  • LocationCanada - no not in an igloo O.o

Posted 18 April 2013 - 08:44 PM

I dont need math to tell me that the weights are uneven.. when i see a buncha heavies goin against assaults on the other side its pretty simple. Matching is not working....


Tonnage.. period.. throw everything else out. If people in assaults tend to wait a lil longer for matches.. well.. get into something a lil lighter to up your chances.


The way it is now.. everybody will soon be in assaults.....

Edited by Xenodraken, 18 April 2013 - 08:51 PM.






1 user(s) are reading this topic

0 members, 1 guests, 0 anonymous users