Jump to content

- - - - -

How Does Elo Work With Groups?


13 replies to this topic

#1 Aceramic

    Member

  • PipPipPipPipPip
  • The Scythe
  • The Scythe
  • 110 posts

Posted 05 August 2013 - 04:04 AM

This is probably the wrong place for this, but I wasn't really sure where else to put it...

Anyways, my question is... How does ELO work with the matchmaker when you're in a group? Say, a group of 4. For the sake of discussion, let's say two of the four are very skilled high-ELO players, and the other two are very low-ELO, perhaps even new accounts.

What happens? Does the matchmaker average the ELO score of all four and try to place the group into a match with similar scores to the average? Or does it simply pick the highest of the four and go from there? Or... ??? I'm assuming nobody actually knows for certain, but any hints or insight would be nice.

I ask because of the outcome of a recent match I was in, of which I have a screenshot of the rather... Surprising after-match scoreboard. I don't want to post it at the moment, for fear of it turning into a "name and shame", but I also didn't want to send it in as a bug/exploit report, in case there's not actually anything wrong, and I'm just a noob. >.>

#2 Koniving

    Welcoming Committee

  • PipPipPipPipPipPipPipPipPipPipPipPipPip
  • The Guide
  • The Guide
  • 23,384 posts

Posted 05 August 2013 - 08:01 AM

It actually works better for groups than individuals. How it works is it uses the group of, say 4, in a base line for an average. It will then try to match that group of 4 against another group of 4. If it cannot, it will match that group of 4 against a group of 3 and 2 of somewhat lower skill.

Sometimes I've seen two groups of 4 matched against groups of 3, 3, and 2. You can tell when this happens because the 8-player system only splits groups up in 3, 3, and 2 for actual 3 man premades with 2 pugs or 2 random players.

Smaller groups tend to be pit against completely random pugs or other smaller groups, but as a group of 2 you'll never find yourself against 4 groups of 2 on the other side. It never happens.

When you group up with someone higher, it becomes an average of the scores and your opponents will be tougher.

Myself I like to group up with two or three newer players, as it brings my ELO average down a bit to where I see more trial builds. Our opponents are still tougher than the newer players might face alone, but they're not as tough as I would normally face when by myself. This lets me bring my stock builds and have more than just a chance.

Scores are usually not name and shame as long as you don't point out anyone specifically, but yes sometimes there are landslide victories. This can happen as a result of poor matching, or poor cooperation within the group.

Once it becomes 8 on 6, it'll very likely become a landslide. It's very important to protect each other and cooperate. Builds also need to be diverse, because of everyone brings a missile boat, who will protect them when the enemy gets close? If everyone brings a brawler, how will you get close if you can't divert attention away from you? If everyone is a sniper, what happens when everyone becomes too hot? And if they're all lights, what happens when the other side camps at the base?

But again to clearly answer, it'll bleed the numbers into an average. This is why sometimes one player will do exceedingly well despite not having a cheese build; he's from a much higher ELO and got 'bled in' by an average to offset his or her team's lower average compared to the enemy, or because he or she was part of a group of much lower players.

#3 BoPop

    Member

  • PipPipPipPipPipPipPip
  • Little Helper
  • Little Helper
  • 543 posts

Posted 05 August 2013 - 08:33 AM

you're assuming ELO works at all :)

honestly I think it just snatches the first 30 people looking for a match in whichever game mode, slaps the 2,3 or 4 man on one side, any other groups (after finding their averages) on the other side and then divides the leftovers based on their WLR/KDR and maybe "time played"

it then sloughs off the rest and voila, 16 people in a game.

this is why you can find a match very fast solo pugging in "Any" mode and also why it takes longer when you're in a skilled 4man trying to find assault only, cuz it's waiting for at least a closely skilled 4, 3 or 2-2mans to balance it out.

then I believe there is a timer, if by the end of that timer it can't match skills even remotely it a) fails to find a match or b ) just tosses in some poor brand spanking new players which results in what I call a "steamroll" (8-0 or 8-1 in about 5 minutes.)

but what do I know, i'm no mathematician or algorithmagician.

#4 zraven7

    Member

  • PipPipPipPipPipPipPipPip
  • 1,207 posts
  • LocationDuluth, Georgia

Posted 05 August 2013 - 08:34 AM

Ok, it goes like this:

Step 1) Elo for each individual is listed.

Step 2) Elo is totaled.

Step 3) Elo Total is written on a slip of paper.

Step 4) All pieces of paper are thrown into a bowl and covered with banana slices.

Step 5) The bowl is given to a room full of monkeys, who then eat the bananas and slips of paper.

Step 6) The monkeys poop. The poo is searched through, and slips of paper that end up in the same pile have their players square off against each other.

#5 codynyc

    Member

  • PipPipPipPipPipPip
  • The People's Hero
  • The People
  • 321 posts
  • Google+: Link
  • Locationda Bronx

Posted 05 August 2013 - 08:47 AM

Actually this past weekend I have notice games are getting matched better. 5 -8 games and 7-8 games. Elo seems to he working better then it has been. Here's hoping

#6 BoPop

    Member

  • PipPipPipPipPipPipPip
  • Little Helper
  • Little Helper
  • 543 posts

Posted 05 August 2013 - 09:00 AM

View Postcodynyc, on 05 August 2013 - 08:47 AM, said:

Actually this past weekend I have notice games are getting matched better. 5 -8 games and 7-8 games. Elo seems to he working better then it has been. Here's hoping


yea me too, the fights have been feeling a smidge more balanced but last night we were getting people on our team that didn't know how to hotkey their weapons :\ and we were getting rolled ;(

#7 Sarsaparilla Kid

    Member

  • PipPipPipPipPipPipPip
  • The Benefactor
  • The Benefactor
  • 664 posts
  • LocationGold Country

Posted 05 August 2013 - 09:08 AM

No, still not perfect...had a steamroll last night, 8-0. Our side had 5 lights, which was at least 2 more than they had, a Victor as our only assault, and two heavies. Been seeing a lot of noobs in the game lately, too, as a few I spectated piloting a trial Atlas were just wandering around on their own, undamaged, and when the rest of the enemy came at them, they fired all their weapons at once and overheated, fired their LRMs at 100m, etc., and went down quick.

Edited by Sarsaparilla Kid, 05 August 2013 - 09:12 AM.


#8 mailin

    Member

  • PipPipPipPipPipPipPipPipPip
  • Big Daddy
  • Big Daddy
  • 2,033 posts
  • LocationWisconsin, USA

Posted 05 August 2013 - 09:31 AM

View PostSarsaparilla Kid, on 05 August 2013 - 09:08 AM, said:

No, still not perfect...had a steamroll last night, 8-0. Our side had 5 lights, which was at least 2 more than they had, a Victor as our only assault, and two heavies. Been seeing a lot of noobs in the game lately, too, as a few I spectated piloting a trial Atlas were just wandering around on their own, undamaged, and when the rest of the enemy came at them, they fired all their weapons at once and overheated, fired their LRMs at 100m, etc., and went down quick.


I've been seeing a lot of this too. Mainly when I pug but occasionally when I drop with a 4 man. I've been seeing new players in trial mech on both sides too, which I find rather odd.

#9 Redshift2k5

    Welcoming Committee

  • PipPipPipPipPipPipPipPipPipPipPip
  • Stone Cold
  • Stone Cold
  • 11,975 posts
  • LocationNewfoundland

Posted 05 August 2013 - 11:09 AM

View PostSarsaparilla Kid, on 05 August 2013 - 09:08 AM, said:

No, still not perfect...had a steamroll last night, 8-0. Our side had 5 lights, which was at least 2 more than they had, a Victor as our only assault, and two heavies. Been seeing a lot of noobs in the game lately, too, as a few I spectated piloting a trial Atlas were just wandering around on their own, undamaged, and when the rest of the enemy came at them, they fired all their weapons at once and overheated, fired their LRMs at 100m, etc., and went down quick.


Just because a match becomes an 8-0 does not mean the teams were uneven and it does not mean the Elo was broken.

Once a team is down a couple of players, the remaining 8 vs 6 can very easily turn into a complete rout, even between even teams. You can see this happen for yourself when playing scrims within your own merc corp: Two teams of 8 playing multiple matches with each other in a session, and some of those matches will still be 8-0! Tactics and chance play a big part, and a single tactical error can cause a wipe among otherwise equal teams.

#10 Sarsaparilla Kid

    Member

  • PipPipPipPipPipPipPip
  • The Benefactor
  • The Benefactor
  • 664 posts
  • LocationGold Country

Posted 05 August 2013 - 12:28 PM

View PostRedshift2k5, on 05 August 2013 - 11:09 AM, said:


Just because a match becomes an 8-0 does not mean the teams were uneven and it does not mean the Elo was broken.

Once a team is down a couple of players, the remaining 8 vs 6 can very easily turn into a complete rout, even between even teams. You can see this happen for yourself when playing scrims within your own merc corp: Two teams of 8 playing multiple matches with each other in a session, and some of those matches will still be 8-0! Tactics and chance play a big part, and a single tactical error can cause a wipe among otherwise equal teams.


True, and as I mentioned, there have been some noob issues lately...

#11 Aceramic

    Member

  • PipPipPipPipPip
  • The Scythe
  • The Scythe
  • 110 posts

Posted 05 August 2013 - 01:17 PM

Just to clarify, overall, the match was actually pretty even. It didn't "feel" like we were getting completely wiped, we had a good sized brawl going between roughly half our team and about 2/3 of the opposing team, and we were doing fairly well until the rest of the opposing team showed up (we got three kills out of it - I'm not really sure where the rest of our team was, or what they were doing...) I didn't really think anything of the loss until I saw the scoreboard and noticed that the top two people on the opposing team combined for a mere 97 damage less than our entire team combined, and were responsible for 6 of the 8 deaths of our team.

#12 Wintersdark

    Member

  • PipPipPipPipPipPipPipPipPipPipPip
  • 13,375 posts
  • Google+: Link
  • Twitter: Link
  • LocationCalgary, AB

Posted 05 August 2013 - 06:08 PM

You really can't use a single match outcome as a barometer for how well Elo is working. There are countless reasons why it could be wonky (exceptionally good players getting into a weight class where they have a low score, for example) and while the system strives to match people of roughly equal average elo scores, this doesn't mean everyone in the match has the same score.

You need to keep in mind its picking from players who've clicked Launch inside a roughly 30 second window. Even with tens of thousands playing, that's a fairly small pool.

The further you are from average, too, the more likely you are to see more random matches. This is because Elo scores, by the nature of the system, conform to a bell curve. There's always going to be exponentially more average players than exceptionally good or bad ones. Hence "average".

This means you're only likely to get high team average elo scores during very busy times; otherwise there just aren't going to get enough exceptional elo players clicking launch inside the window.

#13 Wintersdark

    Member

  • PipPipPipPipPipPipPipPipPipPipPip
  • 13,375 posts
  • Google+: Link
  • Twitter: Link
  • LocationCalgary, AB

Posted 05 August 2013 - 06:12 PM

Also, as has been previously addressed, even equally matched teams will see battles that are grossly one sided, and sometimes just one or two players will have really awesome games.

I had one with five "one shot" kills in the height of the poptarting craze - not because I was cheating or amazingly good, but just because I got absurdly lucky. Sure felt awesome though.

#14 Rascula

    Member

  • PipPipPipPipPipPip
  • Talon
  • Talon
  • 387 posts
  • LocationWord of Blake Protectorate, Epsilon Eridani.

Posted 06 August 2013 - 01:22 PM

And now with 12 v 12 this elo issue is has become oh so more apparent.. the next few weeks should be interesting!





1 user(s) are reading this topic

0 members, 1 guests, 0 anonymous users