Jump to content

How Do We Find Out Our Elo


62 replies to this topic

#41 TheGreatNoNo

    Member

  • PipPipPipPipPipPip
  • Bridesmaid
  • Bridesmaid
  • 448 posts

Posted 20 February 2013 - 03:59 PM

I`m just hoping that I am not bottom of the lake. That is all I want to know from ELO, that I`m not the pond scum at the bottom.

Edited by Lost One, 20 February 2013 - 03:59 PM.


#42 ObsidianSpectre

    Member

  • PipPipPipPipPipPip
  • 289 posts

Posted 20 February 2013 - 03:59 PM

View PostRakashan, on 20 February 2013 - 03:45 PM, said:

For those interested in knowing how they rank, keep in mind. Once ELO is fairly well established you can judge your degree of improvement by tracking your W/L ratio. If you are around 0.5 then you are holding even. If you are above that then you are winning more than half the matches the MM thinks are balanced for you (i.e. improving). This is only applicable once you are well seeded and then only over a large number of matches.

You're assuming that the matchmaker is perfect, and every match you get in will be a perfect match in it's estimation. This is not going to be the case. Because matches will be imperfectly balanced (the player population not being infinite & the match maker not being perfect), I expect above-average players to have W/L ratios slightly above 0.5, just because when the matches aren't perfect, it's more likely to be imbalanced in the that player's favor because more than half the population is less skilled than they are. The reverse is true with less-than-average players, with their W/L ratios likely slightly below 0.5.

So why not track your improvement through variance in W/L? Because it can require a massive number of matches to actually see the difference, when the matchmaker already knows it. It also requires you to accurately track your W/L ratio over time, the variation from 0.5 should be pretty small if the MM is reasonably good, and you'll never be able to rule out statistical anomalies. Did it go up by 0.01 because you improved, or because you had a couple more matches imbalanced in your favor yesterday than statistically expected? It can take a long time to reduce the probability of statistical flukes to a reasonable level (you can't ever rule them out), and at the end of the day all you're really doing is deriving your relative Elo score from your W/L ratio in a very difficult manner.

Edited by ObsidianSpectre, 20 February 2013 - 04:01 PM.


#43 Vodrin Thales

    Member

  • PipPipPipPipPipPipPip
  • Bad Company
  • 869 posts
  • LocationFlorida

Posted 20 February 2013 - 04:02 PM

View PostOne Medic Army, on 20 February 2013 - 02:12 PM, said:

On one hand letting players see their Elo lets them manipulate it and can lead to cheesing.
On the other hand hiding it prevents people from seeing that they're improving (if they're improving).


They track won/loss stats so I don't know why ELO would be more informative. If you continue to improve you ratio of wins you are getting better, as ELO should ensure your competition continuously improves. Therefore if you are winning more frequently over time you are definitely improving.

#44 Punkass

    Member

  • PipPipPipPipPipPip
  • 212 posts

Posted 20 February 2013 - 04:05 PM

ELO should remain hidden forever. That way I can go on thinking my epeen is huge without any empirical evidence to the contrary.

#45 One Medic Army

    Member

  • PipPipPipPipPipPipPipPipPipPip
  • 5,985 posts
  • LocationBay Area, California

Posted 20 February 2013 - 04:06 PM

View PostVodrin Thales, on 20 February 2013 - 04:02 PM, said:

They track won/loss stats so I don't know why ELO would be more informative. If you continue to improve you ratio of wins you are getting better, as ELO should ensure your competition continuously improves. Therefore if you are winning more frequently over time you are definitely improving.
I suppose having a 55% win rate after Elo is implemented shows you're probably getting better, but having an actual metric to show your relative absolute skill would also be nice.
Possibly something like tiers? Just show someone which tier they're in, or what percentile of the players (to the nearest 1%), would make it harder to track fractional adjustments and thus manipulate the Elo score.

#46 LT Kinslayer

    Member

  • PipPipPipPipPip
  • 137 posts

Posted 20 February 2013 - 05:03 PM

Make it privately visible, so I would know what kind of teammates and opponents I can expect.
Also, if nothing is visible I cannot tell if my performance has improved or declined. If I cannot see, how would I ever be able to test which changes to my playstyle improve my performance?

Extremely obvious example:
My regular Elo is 1900, but I get drunk with my friends two nights in a row and then play a couple of hours after I get home. Usually I would think I did pretty good, as drunk people do. But what if my Elo would show that I am at 1700 afterwards? Not only would I know I actually did bad, I would also be able to take more risks afterwards (read: fun builds, Leeerrrooys, etc) while trying to get back to my proper rating, or not.

From another angle:
In another thread, I saw one of our forum warriors, a ?full-time? Pugger, flabbergasted when people (including me) suggested that a high KDR does not necessarily mean that a player is actually really good. This particular player claimed that he was doing all he could, AND had a 3.7 KDR, but still he lost more than he won, therefore his teammates were failing and there was nothing he could realistically do to improve his odds.

Obviously this guy knows how to get to an advantageous position, how to aim, and how to finish.
He was probably also right with regards to his teammates failing.

HOWEVER:
I also know someone that is a full-time, no voice com, PUG player that also has a 3.7 KDR, with +-2000 matches under his belt after open. AND he has a 1.22 W/L rating (winning 55% of his matches) He says he could still improve on this performance. After a 2000 match sample size, this disparity cannot be easily discarded. It seems there are other things at play besides simply being able to kill opponents for getting better win rates and thus Elo.

Privately visible Elo would be a very nice tool for people to see if their gameplay is improving in various fields, without getting unnecessarily fixated on KDR. Especially considering the fact that the Elo system will strive for a 1:1 W/L. Besides simply trying to get a higher KDR and hoping this really improves one's skill, one could experiment in many ways. There would be a clearly visible indicator (Elo) of worsening/improving performance.
Proper timing and position for assisting, pushing, committing, hiding, escaping and distracting are all just as important for securing a win as being able to aim at a damaged component. Each player does make a difference, how much of a difference depends on the player and the situation.

Elo tries to keep 1/1 W/L, so: If I'm not at the very top or bottom: How the hell do I know if I'm improving or getting worse except for my highly unreliable K/D? I say: Give me privately visible Elo, and additional other stats as well.

To those worried about e-peeners because people shouldn't care about stats:
Why care about e-peeners if you don't care about stats anyway? They are only showing off something
you don't care about, right?

#47 semalferuzA

    Member

  • PipPipPipPipPip
  • 125 posts

Posted 20 February 2013 - 05:06 PM

View PostKraven Kor, on 20 February 2013 - 02:14 PM, said:

I am fairly certain Match Score is the key stat ELO is tied to. They seem to always be a number between 0 and 100.

So keep track of your match scores and keep a running average. Thus far, I'm still failing even if they grade on a bell curve ;)


Well I had 109 match score earlier that I noticed so it's not just 0 and 100.

#48 Bilbo

    Member

  • PipPipPipPipPipPipPipPipPipPip
  • The Nimble
  • The Nimble
  • 7,864 posts
  • LocationSaline, Michigan

Posted 20 February 2013 - 05:10 PM

View PostKraven Kor, on 20 February 2013 - 02:14 PM, said:

I am fairly certain Match Score is the key stat ELO is tied to. They seem to always be a number between 0 and 100.

So keep track of your match scores and keep a running average. Thus far, I'm still failing even if they grade on a bell curve ;)

No it isn't. Your Elo is adjusted up or down depending on whether your team was supposed to win or lose the match. Wins and losses........your actual performance in the match has no bearing.

#49 Lukoi Banacek

    Member

  • PipPipPipPipPipPipPipPipPip
  • WC 2018 Top 12 Qualifier
  • WC 2018 Top 12 Qualifier
  • 4,353 posts

Posted 20 February 2013 - 05:15 PM

Bryan Eckerman screwed up and told a player on these forums where he stood in a certain percentile in assault mechs in reply to a question about the matchmaker being ultra slow on day one of Elo implementation. I think knowing your percentile in a chassis/weight class wouldn't be bad information or unreasonable to share.

Gives a partial indication of your skill and therefore where you can best contribute to a team for example for competitive or CW play, isn't easily reverse engineerable and still satisfies the urge some people have simply to know how good they are or aren't.

Then you make it private, so if yes they copy/paste it into their sigs from a screenshot and then they decide to get all high and mighty on forums folks can easily clown them for having their ego so tied into a videogame that their sheltered, ****** lives as Mom-n-Pop basement dwellers will be better served by the attendant humiliation that you KNOW they'll feel regardless of whatever silly internet bravado they choose to display.

Edited by Lukoi, 20 February 2013 - 05:16 PM.


#50 semalferuzA

    Member

  • PipPipPipPipPip
  • 125 posts

Posted 20 February 2013 - 05:22 PM

I think that their intent must be to eventually make the elo rating visible. In their post they talked about the amount the rating changes by, how it changes, and that teams points won or lost are weighted by the disparity in team ratings.

An example they used included one team being high rated and winning very little or losing a lot of points against another team that was low rated and losing few points or winning a lot of points.

If they never intended for rating to become visible this information would be of little use to us and the match maker would seek to get as close to a 50-50 match as possible, which could result in long queue times. If the matchmaker matches uneven teams in order to decrease queue times then the players should at least know that it was not an even match and as such they only lost 5 points instead of 30. As long as the matchmaker is weighting teams it is likely that even people of average skill will not be 50% until the population is much larger.

Having elo visible has benefits for both skilled and unskilled players. I don't particularly care if I can see every ones rating but I'd at least like to see mine and the average ratings per team on the scoreboard.

Edited by semalferuzA, 20 February 2013 - 05:24 PM.


#51 Helmer

    Member

  • PipPipPipPipPipPipPipPipPip
  • Stone Cold
  • 3,272 posts
  • LocationColumbus, Ga

Posted 20 February 2013 - 05:33 PM

View PostObsidianSpectre, on 20 February 2013 - 03:03 PM, said:

Other stats won't let you accurately judge your own skill levels except at the very lowest and highest levels of play.

<snip>

It is now, but they never said it will always be so. I don't think they've committed either way.


As with all Statistics they are subject to manipulation and , at times, interpretation. Just as you've demonstrated above , the ELO formula also may not allow you to accurately judge your own skill levels as it is Team Average A vs Team Average B. Yes, you can assume that you are a great medium pilot due to your High ELO, however, if you're running in a dedicated group that communicates well , you may very well be carried to a Higher ELO than what you yourself are actually capable of.
Whereas a high skilled PUG player, may just have horrible luck with drops, get 7 kills in every match, and yet lose everymatch and watch his ELO plummet.



You are correct, this is subject to change. There is nothing that has been published in the open forums to indicate a commitment either way. We will see what the future holds, however, i think the over abundance of statistics coming in the future may be enough for stat lovers.


Cheers.

#52 semalferuzA

    Member

  • PipPipPipPipPip
  • 125 posts

Posted 20 February 2013 - 05:39 PM

View PostHelmer, on 20 February 2013 - 05:33 PM, said:


As with all Statistics they are subject to manipulation and , at times, interpretation. Just as you've demonstrated above , the ELO formula also may not allow you to accurately judge your own skill levels as it is Team Average A vs Team Average B. Yes, you can assume that you are a great medium pilot due to your High ELO, however, if you're running in a dedicated group that communicates well , you may very well be carried to a Higher ELO than what you yourself are actually capable of.
Whereas a high skilled PUG player, may just have horrible luck with drops, get 7 kills in every match, and yet lose everymatch and watch his ELO plummet.


But at least a person would know if their current elo rating is accurate or not. Obviously if I only drop with friends my elo rating is only accurate when I drop with those friends. If I hold a particular elo rating for a period of time and then one day win 10+ in a row or lose 10+ in a row I didn't magically get a lot better/worse and neither did the community. Plus those games might have gone as they should have. I.e. winning 3-5 points or losing 3-5 points, not actually big rating changes. Either way I can still gauge my skill level. Right now it's kind of just a guess. I know I do well pretty much every single match in relation to my teammates but I really don't know where that puts me overall.

Edited by semalferuzA, 20 February 2013 - 05:41 PM.


#53 xX_Nero_Xx

    Member

  • PipPipPipPipPipPipPip
  • FP Veteran - Beta 1
  • FP Veteran - Beta 1
  • 649 posts
  • LocationDallas,Texas

Posted 20 February 2013 - 05:42 PM

thank you for the info ,just was wandering why i get put in to matchs so i can play to my strengths where im week i could improve on is all .
if im put in a certain are cause im good or bad would have been nice to know

#54 Helmer

    Member

  • PipPipPipPipPipPipPipPipPip
  • Stone Cold
  • 3,272 posts
  • LocationColumbus, Ga

Posted 20 February 2013 - 05:49 PM

View PostsemalferuzA, on 20 February 2013 - 05:39 PM, said:


But at least a person would know if their current elo rating is accurate or not. Obviously if I only drop with friends my elo rating is only accurate when I drop with those friends. If I hold a particular elo rating for a period of time and then one day win 10+ in a row or lose 10+ in a row I didn't magically get a lot better/worse and neither did the community. Plus those games might have gone as they should have. I.e. winning 3-5 points or losing 3-5 points, not actually big rating changes. Either way I can still gauge my skill level. Right now it's kind of just a guess. I know I do well pretty much every single match in relation to my teammates but I really don't know where that puts me overall.



I agree. Lets what to see what the future brings.


Patience.



Cheers.

Edited by Helmer, 20 February 2013 - 05:50 PM.


#55 xX_Nero_Xx

    Member

  • PipPipPipPipPipPipPip
  • FP Veteran - Beta 1
  • FP Veteran - Beta 1
  • 649 posts
  • LocationDallas,Texas

Posted 20 February 2013 - 05:52 PM

thats kool and all but when you dont let ppl see things that affect them then its human nature to want to see it .
the more you hide it the more we will want to see it lol

#56 Mackman

    Member

  • PipPipPipPipPipPipPip
  • 746 posts
  • LocationCalifornia

Posted 20 February 2013 - 05:56 PM

View PostParticle Man, on 20 February 2013 - 03:24 PM, said:

Like is aid already. there are actual real stats on the way. there's no need for this stat that can be easily used to abuse the system and that might not mean anything to anyone anyway considering we have no real idea what is used to gauge it.


How can stats be more "real" than those that are used for matchmaking? It's meant to match people of equal skill (or, if you want to be pedantic, of "equal chance to win the match"), so what future stat could possibly be a better indicator of skill? And if they don't want to show Elo, then they should at least show a leaderboard derived from Elo.

I want to know where I stand as a player. I play games to have fun, and for me, having fun in a multiplayer setting means getting better. And not being able to tell where you are skillwise is not very fun.

#57 Demoned

    Member

  • PipPipPipPipPipPipPip
  • Legendary Founder
  • Legendary Founder
  • 670 posts
  • Locationi Died went to heaven, then died again now I'm in Equestria

Posted 20 February 2013 - 06:03 PM

i would like to know my ELO for 1 reason only
am i getting better at the game? that's it lol

and i must say it must be working at the end of a game my team (and myself)
have a very spread damage score we all do between 300-500 damage,
now i can solo drop with confidence.
and we always tend to be fighting till the last few mechs, so I'm getting put against some great teams as well.

#58 Tickdoff Tank

    Member

  • PipPipPipPipPipPipPipPipPip
  • Bad Company
  • Bad Company
  • 2,647 posts
  • LocationCharlotte NC

Posted 20 February 2013 - 06:05 PM

Judging by the people I am grouped up with I think my Elo score is in the negatives.

#59 xX_Nero_Xx

    Member

  • PipPipPipPipPipPipPip
  • FP Veteran - Beta 1
  • FP Veteran - Beta 1
  • 649 posts
  • LocationDallas,Texas

Posted 20 February 2013 - 06:13 PM

thats also the point do you want to know if you were suppose to win or lose the match

#60 Kassatsu

    Member

  • PipPipPipPipPipPipPipPip
  • The Devoted
  • The Devoted
  • 1,078 posts
  • LocationColorado

Posted 20 February 2013 - 06:30 PM

View PostOne Medic Army, on 20 February 2013 - 02:12 PM, said:

On one hand letting players see their Elo lets them manipulate it and can lead to cheesing.
On the other hand hiding it prevents people from seeing that they're improving (if they're improving).


Not really. It's pretty easy to lame them even without seeing an exact score.

Are the enemies giving you a challenge? y/n

If no, keep doing what you're doing. If yes, intentionally lose X matches. Repeat question and answer.





4 user(s) are reading this topic

0 members, 4 guests, 0 anonymous users