Jump to content

Why Is There Elo Separation For Leaderboard Tournaments?

Gameplay

100 replies to this topic

#21 Nightmare1

    Member

  • PipPipPipPipPipPipPipPipPipPip
  • Ace Of Spades
  • Ace Of Spades
  • 7,636 posts
  • Twitch: Link
  • LocationPeeking over your shoulder while eating your cookies.

Posted 25 October 2014 - 09:17 PM

I'd settle for a limit to the number of LRM boats in a match. The amount of spam is really frustrating. :(

#22 Ghogiel

    Member

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

Posted 25 October 2014 - 09:24 PM

View PostArtgathan, on 25 October 2014 - 09:16 PM, said:



There's only 3 Elo brackets. I forget the exact numbers, but it's essentially 'low', 'medium' and 'high'.

You mean Light, medium, heavy and assault ? Because there are no actual brackets..

#23 Artgathan

    Member

  • PipPipPipPipPipPipPipPip
  • Knight Errant
  • Knight Errant
  • 1,764 posts

Posted 25 October 2014 - 09:25 PM

View PostGhogiel, on 25 October 2014 - 09:24 PM, said:

You mean Light, medium, heavy and assault ? Because there are no actual brackets..


Well in theory there's three. You're supposed to be assigned one for each weight class (for instance, be 'low' in lights, 'high' in mediums, and 'medium' for heavies and assaults) and then matched based on that. In theory.

#24 Ghogiel

    Member

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

Posted 25 October 2014 - 09:27 PM

View PostArtgathan, on 25 October 2014 - 09:25 PM, said:


Well in theory there's three. You're supposed to be assigned one for each weight class (for instance, be 'low' in lights, 'high' in mediums, and 'medium' for heavies and assaults) and then matched based on that. In theory.

In what theory?

#25 Mcgral18

    Member

  • PipPipPipPipPipPipPipPipPipPipPipPip
  • CS 2019 Top 8 Qualifier
  • CS 2019 Top 8 Qualifier
  • 17,987 posts
  • LocationSnow

Posted 25 October 2014 - 09:28 PM

View PostGhogiel, on 25 October 2014 - 09:27 PM, said:

In what theory?


There was mention of Elo buckets at one point.

Not sure if that ever came about.

#26 Quicksilver Aberration

    Member

  • PipPipPipPipPipPipPipPipPipPipPip
  • The Nightmare
  • The Nightmare
  • 11,496 posts
  • LocationKansas City, MO

Posted 25 October 2014 - 09:32 PM

View PostAresye, on 25 October 2014 - 09:06 PM, said:

I guess I should stop using the term, "top player." Perhaps, "most dedicated player" would fit better.

I'm honestly happy that there are new faces on the leaderboard, and that some players that wouldn't normally be there get to experience that, but it's sort of a bittersweet moment, because while some players are shocked at seeing themselves on the leaderboard, there's a large number of competitive players that (by their very nature) entered into this competition for the sole purpose to win it, but end up having to put in 4x the effort to do so.

It's like a reverse underdog story.

Without hard stats, you have no other basis for this other than that's how it feels, which is poor grounds to put limitations on. Honestly I would love to see PGI release some stats on this just to sate my curiosity about this and whether it is an oft occurring phenomenon.

I could say that most players could be given 4x the amount of time and still won't be able to top scores players like Proton, heimdelight, or Odwalla could put up in a weekend. Even then, why does it even matter if the top players are putting in 4x times the effort (of which I highly doubt) if they are still winning in the end anyway? Simply because you feel it isn't smooth sailing for the top players to top the charts, or this unfounded (imo) belief that if bad players play enough they can get good enough matches to top the charts (I can understand 1-2 good matches, but 10?)?

#27 Ghogiel

    Member

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

Posted 25 October 2014 - 09:32 PM

View PostMcgral18, on 25 October 2014 - 09:28 PM, said:


There was mention of Elo buckets at one point.

Not sure if that ever came about.

I have heard the term bucket used by PGI in reference to ELo many times. He said bracket though.

Edited by Ghogiel, 25 October 2014 - 09:32 PM.


#28 Budor

    Member

  • PipPipPipPipPipPipPipPipPip
  • 2,565 posts

Posted 25 October 2014 - 09:33 PM

I just took a look at the boards and have to admit that i have never seen some off the higher ranked players BUT i am really bad at recalling names so i want to believe that everything is legit.

To keep with this spirit i think everyone on the boards deserves this no matter which hive he farmed:
Posted Image

#29 Artgathan

    Member

  • PipPipPipPipPipPipPipPip
  • Knight Errant
  • Knight Errant
  • 1,764 posts

Posted 25 October 2014 - 09:37 PM

View PostGhogiel, on 25 October 2014 - 09:32 PM, said:

I have heard the term bucket used by PGI in reference to ELo many times. He said bracket though.


I might be confusing the terms (in which case, sorry). The way the matchmaker is supposed to work is that every player has an Elo score assigned on a per-weight-class basis. This Elo Score falls into one of three categories, Low, Medium or High.

I use the terms bucket/bracket to refer to the "Low/Medium/High" part of the Elo score. I assumed this was how the OP was using it as well, as I can't really think of another way to use it. The MM is "supposed" to seperate all the players based on Elo level, so a 'Medium' Elo match would see all the participating pilots be in the 'Medium' Elo Bucket/Bracket for their selected weight class.

#30 Ghogiel

    Member

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

Posted 25 October 2014 - 09:47 PM

View PostArtgathan, on 25 October 2014 - 09:37 PM, said:


I might be confusing the terms (in which case, sorry). The way the matchmaker is supposed to work is that every player has an Elo score assigned on a per-weight-class basis. This Elo Score falls into one of three categories, Low, Medium or High.

I use the terms bucket/bracket to refer to the "Low/Medium/High" part of the Elo score. I assumed this was how the OP was using it as well, as I can't really think of another way to use it. The MM is "supposed" to seperate all the players based on Elo level, so a 'Medium' Elo match would see all the participating pilots be in the 'Medium' Elo Bucket/Bracket for their selected weight class.


Pretty sure there aren't any low, medium, high Elo brackets sorting in the MM. afaik it is still just picking a target Elo for a match and trying to pick players closest to target, then over time increasing the threshold at which it can pick from the player pool, up into the +/- 1000 Elo range from target.

I think the OP is refering to a metaphorical bracket, and not a literal Elo segregation the MM is actualy using.

Edited by Ghogiel, 25 October 2014 - 09:51 PM.


#31 Xtrekker

    Member

  • PipPipPipPipPipPipPip
  • Legendary Founder
  • Legendary Founder
  • 865 posts
  • LocationOn your six

Posted 25 October 2014 - 09:50 PM

I like the idea of a tournament only Elo-free queue. Also, make it always available.

#32 Wintersdark

    Member

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

Posted 25 October 2014 - 10:07 PM

View PostGhogiel, on 25 October 2014 - 09:47 PM, said:


Pretty sure there aren't any low, medium, high Elo brackets sorting in the MM. afaik it is still just picking a target Elo for a match and trying to pick players closest to target, then over time increasing the threshold at which it can pick from the player pool, up into the +/- 1000 Elo range from target.

I think the OP is refering to a metaphorical bracket, and not a literal Elo segregation the MM is actualy using.

There was talk about breaking Elo into buckets, and simply matching players against others in their bucket, but to my knowledge that didn't happen.

Russ frequently discusses team vs. team elo variance, as does Karl in his thread, and they tend to range down to 50-100pts between teams - so very good.

Karl does discuss "in team" Elo variance at one point (after much digging), and it's typically 200 pts.

#33 MischiefSC

    Member

  • PipPipPipPipPipPipPipPipPipPipPipPip
  • The Benefactor
  • The Benefactor
  • 16,697 posts

Posted 25 October 2014 - 10:16 PM

I get the logic of removing Elo for contests like this but in the end....

competitive players don't have to 'work 4x as hard' as nubs to get onto the leaderboard. They just have to work about the same - which is harder than competitive players generally have to work to do well.

#34 Ghogiel

    Member

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

Posted 25 October 2014 - 10:19 PM

View PostWintersdark, on 25 October 2014 - 10:07 PM, said:

There was talk about breaking Elo into buckets, and simply matching players against others in their bucket, but to my knowledge that didn't happen.

Russ frequently discusses team vs. team elo variance, as does Karl in his thread, and they tend to range down to 50-100pts between teams - so very good.

Karl does discuss "in team" Elo variance at one point (after much digging), and it's typically 200 pts.

This is why we can't have nice things.

PGI talk about average Elo deviation on a team and say it's down to 200 points or whatever. When its putting 10 people close to target on the team but then throws 2 low or high players in there that are 800 points above or below that target.. then they average it out and say 'hey look it's averaged out it's within 100 points'. gee gee.

It's bascially moot everytime a high or low elo player makes it into any match not near their own Elo rating.

#35 MischiefSC

    Member

  • PipPipPipPipPipPipPipPipPipPipPipPip
  • The Benefactor
  • The Benefactor
  • 16,697 posts

Posted 25 October 2014 - 10:23 PM

View PostGhogiel, on 25 October 2014 - 10:19 PM, said:

This is why we can't have nice things.

PGI talk about average Elo deviation on a team and say it's down to 200 points or whatever. When its putting 10 people close to target on the team but then throws 2 low or high players in there that are 800 points above or below that target.. then they average it out and say 'hey look it's averaged out it's within 100 points'. gee gee.

It's bascially moot everytime a high or low elo player makes it into any match not near their own Elo rating.


Here's the thing though -

If Elo didn't work then 'high Elo players' would have the same Elo as everyone else.

Elo works, that's why there are high and low and moderate Elo players.

#36 Ghogiel

    Member

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

Posted 25 October 2014 - 10:26 PM

View PostMischiefSC, on 25 October 2014 - 10:16 PM, said:

I get the logic of removing Elo for contests like this but in the end....

competitive players don't have to 'work 4x as hard' as nubs to get onto the leaderboard. They just have to work about the same - which is harder than competitive players generally have to work to do well.

Don't be daft.

You can tank Elo to get into the middle/low brackets. New players are in the middle brackets. Middle bracket has statistically has the highest skill variance.

in short underhive ghetto = easier to get big scores.


None of that is possible or true with a high Elo or in higher Elo games.


View PostMischiefSC, on 25 October 2014 - 10:23 PM, said:


Here's the thing though -

If Elo didn't work then 'high Elo players' would have the same Elo as everyone else.

Elo works, that's why there are high and low and moderate Elo players.


well it doesn't work then. I CAN have a lower Elo if I choose.

Edited by Ghogiel, 25 October 2014 - 10:27 PM.


#37 MischiefSC

    Member

  • PipPipPipPipPipPipPipPipPipPipPipPip
  • The Benefactor
  • The Benefactor
  • 16,697 posts

Posted 25 October 2014 - 10:40 PM

View PostGhogiel, on 25 October 2014 - 10:26 PM, said:

Don't be daft.

You can tank Elo to get into the middle/low brackets. New players are in the middle brackets. Middle bracket has statistically has the highest skill variance.

in short underhive ghetto = easier to get big scores.


None of that is possible or true with a high Elo or in higher Elo games.




well it doesn't work then. I CAN have a lower Elo if I choose.


Highest skill variance is actually on the ends. The difference between 1200 and 1600 is pretty meager. The difference between 1800 and 2,000 is dramatic. gaining (or losing) points when you move up or down the curve is correspondingly more difficult) and 'tanking' your Elo isn't as easy as you make it sound. Tanking from 2k to 1600 for example would take losing something like 50 games in a row, maybe more. Why? You're only going to lose a bit of Elo on about 1/2 your games.

#38 Wintersdark

    Member

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

Posted 25 October 2014 - 10:41 PM

View PostGhogiel, on 25 October 2014 - 10:19 PM, said:

This is why we can't have nice things.

PGI talk about average Elo deviation on a team and say it's down to 200 points or whatever. When its putting 10 people close to target on the team but then throws 2 low or high players in there that are 800 points above or below that target.. then they average it out and say 'hey look it's averaged out it's within 100 points'. gee gee.

It's bascially moot everytime a high or low elo player makes it into any match not near their own Elo rating.


But that's impossible to avoid. By their very nature, very high (and low) elo players are very rare. Everyone likes to think they are Super Special, but the vast majority of players are average. Those actually very high Elo players will only rarely ever get into matches where everyone is at their level because that would require millions playing at the same time. The MM can only draw from who's clicked Play in a comparatively short period of time. So, you get what you get. It's not a failing of the system, it's a failing of the inputs.

The MM doesn't dump you into lower Elo matches because (as my sig says) it's a spiteful ******* or incompetent, it does so to get you a match period.

#39 Wintersdark

    Member

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

Posted 25 October 2014 - 10:52 PM

View PostMischiefSC, on 25 October 2014 - 10:40 PM, said:


Highest skill variance is actually on the ends. The difference between 1200 and 1600 is pretty meager. The difference between 1800 and 2,000 is dramatic. gaining (or losing) points when you move up or down the curve is correspondingly more difficult) and 'tanking' your Elo isn't as easy as you make it sound. Tanking from 2k to 1600 for example would take losing something like 50 games in a row, maybe more. Why? You're only going to lose a bit of Elo on about 1/2 your games.


Indeed.

Tanking your Elo is a hell of a lot harder than it looks. At the very highest end, you can pull yourself down fairly easily as the matches are tight - loss of a player is critical (if that player is, say, just not really contributing).

However, you still need to lose a match for your Elo to go down. It can go down at most 50 points per match, but then only if your opposing team is of far lower skill than your team. This is unlikely, as skill variance between teams is ~100 points, which translates to a roughly 40 point Elo loss per match lost to a enemy team ranked 100 points below yours.

But... Figure you're going to be matched against teams above yours too. Each of those matches isn't going to substantially impact your Elo score. So, throw out half your matches that way. Then, of the remainder, only the losses will result in Elo reduction, and even then it's not going to be incredibly fast.

Things get worse as you get down the ladder, though. As play loosens up, it's easier and easier for a team to do well even with a useless player suiciding. Player numbers at that Elo rating increase, so match Elo variance decreases, in turn resulting in slower movement on your own Elo rating.

So, yeah, you COULD smurf your Elo down, but the reality is it'll take a hell of a lot of matches, lots and lots of useless losses to get there.

#40 Ghogiel

    Member

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

Posted 25 October 2014 - 10:52 PM

View PostMischiefSC, on 25 October 2014 - 10:40 PM, said:


Highest skill variance is actually on the ends. The difference between 1200 and 1600 is pretty meager. The difference between 1800 and 2,000 is dramatic. gaining (or losing) points when you move up or down the curve is correspondingly more difficult) and 'tanking' your Elo isn't as easy as you make it sound. Tanking from 2k to 1600 for example would take losing something like 50 games in a row, maybe more. Why? You're only going to lose a bit of Elo on about 1/2 your games.

It's easier to tank from high Elo because the variance is at the top. And that's why when shoving a player 800 points higher than the 1600 rated player and averaging the variance and saying: guise 200 average varianc gg. Sounds good on paper but probably isn't really.





1 user(s) are reading this topic

0 members, 1 guests, 0 anonymous users