+ Reply to Thread
Page 4 of 6 FirstFirst 1 2 3 4 5 6 LastLast
Results 46 to 60 of 77
Like Tree23Likes

  Click here to go to the first Rift Team post in this thread.   Thread: ELO, and other PvP MMOs

  1. #46
    Ascendant Zehne's Avatar
    Join Date
    Jul 2012
    Posts
    1,443

    Default

    Played alot of druid.
    Played alot of druid.
    Played alot of druid.
    Lost a bunch of WF's
    Then, when this screen was introduced:
    ELO, and other PvP MMOs-005-011.jpg
    ELO, and other PvP MMOs-006-011.jpg
    ELO, and other PvP MMOs-007-011.jpg
    ELO, and other PvP MMOs-008-011.jpg
    ELO, and other PvP MMOs-008-012.jpg
    ELO, and other PvP MMOs-008-013.jpg
    ELO, and other PvP MMOs-033-033.jpg
    Worked my way solo queing to around 2500 ELO at the 100-100(ish) mark. (no SS's archived).
    In this 2300ish zone, was quite clearly earning/losing around 5-9 points a match.

    At about the 100-100 zone, I started to Pre-made a bunch.
    Then at around 2650(ish) and a 55ish% W:L I stopped hard-core Pre-mading.
    Then I Pre-made some times, mostly duo que, and do some solo queing (I'd say around 20%/60%/20% respectively).

    Which is when this starts to happen:
    ELO, and other PvP MMOs-222-168.jpg
    ELO, and other PvP MMOs-223-168.jpg
    ELO, and other PvP MMOs-225-168.jpg
    ELO, and other PvP MMOs-226-168.jpg
    ELO, and other PvP MMOs-230-168.jpg
    And That's not the end, I've been continuing that meta. Don't have them handy, but atm I'm now at a 58% W:L, and around 2900 ELO.

    So maybe a 20 point shift is a stretch(but I have seen close to it). But the Average amount of points earned/lost per game did in fact go up for me from 2300ish to 2800ish ELO.

    So Why, when my history gets bigger and bigger, do my variations get bigger and bigger.

    When both a MUCH lower ELO person and a higher(I'm assuming tech is 3k+) person only shift by 1-2 points when I'm seeing 8ish per win?

    And i'm not talking about my ELO, I think mine is somewhat working.

    But I can't help but point at Ezuen who has undoubtedly played many more matches, and should have at an absolute minimum 2k ELO.

    But he doesn't, because his ELO only changes by 1-2 a match. And thus the climb from 121 has been long. I'd guess it'd take him nearly a year or more of playing to get his ELO to a score even remotely close to his actual play level.

    All because the scores of some players do not change fast enough(which i'm assuming is because of a 'history').

  2. #47
    Ascendant ecru's Avatar
    Join Date
    Jul 2013
    Posts
    3,741

    Default

    Quote Originally Posted by Ocho View Post
    Nope, no weighting by matches played. This is the impact of our K value system. Once your rating has been significantly established the k value decreases. It can increase again if your points raise substantially, but otherwise it keeps volatility in scores lower. There are lots of math reasons for this, but the system is designed for predicting wins accurately and it mathematically works out. It's less satisfying as a rating displayed for players, but that isn't what the system was designed for or its intent.



    No one else can see your ELO rating, so you aren't being judged on it. AFKing out however does substantially impact your team, it's absolutely the most detrimental activity you can engage in to ensure your team's loss. People remember that kind of stuff, and I would not be surprised if they harbor resentments over it.
    i'm still not seeing a good reason why someone who hasn't played in over a year should be stuck with a 4k+ ELO. knowing the matches i get, this sets up teams to be slaughtered over and over again.

    i get matched up like this pretty often. these people aren't killing anything, and i'm keeping them alive for a pretty long time considering the other team's dps, so i don't really see why someone with a higher ELO and 2/3rds of my stats is going to be expected to carry these people. it's ridiculous, it isn't fun, and it isn't rewarding in the slightest. it's a constant exercise in frustration and i can't even begin to imagine how much worse it would be while being severely undergeared.
    Last edited by ecru; 07-02-2015 at 12:55 PM.
    starplatinum@greybriar
    bluebarbie@greybriar
    twitch
    youtubes

  3. #48
    Ascendant SprawlnBrawl's Avatar
    Join Date
    May 2013
    Posts
    4,349

    Default

    Quote Originally Posted by Zehne View Post
    Played alot of druid.
    Played alot of druid.
    Played alot of druid.
    Lost a bunch of WF's
    Then, when this screen was introduced:
    Attachment 24927
    Attachment 24928
    Attachment 24929
    Attachment 24930
    Attachment 24931
    Attachment 24932
    Attachment 24933
    Worked my way solo queing to around 2500 ELO at the 100-100(ish) mark. (no SS's archived).
    In this 2300ish zone, was quite clearly earning/losing around 5-9 points a match.

    At about the 100-100 zone, I started to Pre-made a bunch.
    Then at around 2650(ish) and a 55ish% W:L I stopped hard-core Pre-mading.
    Then I Pre-made some times, mostly duo que, and do some solo queing (I'd say around 20%/60%/20% respectively).

    Which is when this starts to happen:
    Attachment 24934
    Attachment 24935
    Attachment 24936
    Attachment 24937
    Attachment 24938
    And That's not the end, I've been continuing that meta. Don't have them handy, but atm I'm now at a 58% W:L, and around 2900 ELO.

    So maybe a 20 point shift is a stretch(but I have seen close to it). But the Average amount of points earned/lost per game did in fact go up for me from 2300ish to 2800ish ELO.

    So Why, when my history gets bigger and bigger, do my variations get bigger and bigger.

    When both a MUCH lower ELO person and a higher(I'm assuming tech is 3k+) person only shift by 1-2 points when I'm seeing 8ish per win?

    And i'm not talking about my ELO, I think mine is somewhat working.

    But I can't help but point at Ezuen who has undoubtedly played many more matches, and should have at an absolute minimum 2k ELO.

    But he doesn't, because his ELO only changes by 1-2 a match. And thus the climb from 121 has been long. I'd guess it'd take him nearly a year or more of playing to get his ELO to a score even remotely close to his actual play level.

    All because the scores of some players do not change fast enough(which i'm assuming is because of a 'history').
    No you are just proving ochos k value description

  4. #49
    Ascendant SprawlnBrawl's Avatar
    Join Date
    May 2013
    Posts
    4,349

    Default

    Quote Originally Posted by ecru View Post
    i'm still not seeing a good reason why someone who hasn't played in over a year should be stuck with a 4k+ ELO. knowing the matches i get, this sets up teams to be slaughtered over and over again.

    i get matched up like this pretty often. these people aren't killing anything, and i'm keeping them alive for a pretty long time considering the other team's dps, so i don't really see why someone with a higher ELO and 2/3rds of my stats is going to be expected to carry these people. it's ridiculous, it isn't fun, and it isn't rewarding in the slightest. it's a constant exercise in frustration and i can't even begin to imagine how much worse it would be while being severely undergeared.
    This I agree with there should be decay of rating and k value reset. Zimmyz is at like 4.5k elo but hasn't fresh 65 that hasn't PvP'd in months. I haven't seen him on in a while so not sure if he even tried wfs yet

  5. #50
    Ascendant Zehne's Avatar
    Join Date
    Jul 2012
    Posts
    1,443

    Default

    Quote Originally Posted by SprawlnBrawl View Post
    Read Ocho post on k value.
    Quote Originally Posted by Ocho View Post
    Nope, no weighting by matches played. This is the impact of our K value system. Once your rating has been significantly established the k value decreases. It can increase again if your points raise substantially
    How? May I ask?

    Does one 'substantially' change their score (and thus the K value) if their score is 3500 and only changes 1-2 points a match?

    A player can change how they play performance wise quicker than the ELO system can account for.

    Thus a player can have a score of 300 and be playing at a 3k level.
    And a player can have a score of 3k and be playing at a 300 level.

    And thus, further inaccuracies are introduced into the calculation of ELO because such anomalies exist

    If you won't reset ELO, reset the K value. At least then everyone will fluctuate a bit more until they settle into their 'zone' once again.

  6. #51
    Plane Touched
    Join Date
    May 2014
    Posts
    250

    Default

    The usefulness of an ELO rating diminishes rapidly when you have larger and larger teams, because win or loss is so dependent on your teammates. This isn't just Rift, but in general. ELO is best for 1v1 scenarios, or if you have a team ELO on for example a professional gaming team like in DotA or something, where it is a consistent lineup and you aren't just jamming together a bunch of individual ELO ratings.

    Basically, don't worry about it. People fuss over it because they feel judged and a formula is telling you that you're bad, but you have to realize it's next to worthless, it's not like a chess ELO or something that you actually have an impact over.
    Rokola@Faeblight
    Ranaa@Faeblight

  7. #52
    Ascendant Zehne's Avatar
    Join Date
    Jul 2012
    Posts
    1,443

    Default

    Quote Originally Posted by Princess Celly View Post
    The usefulness of an ELO rating diminishes rapidly when you have larger and larger teams
    I think everyone agrees on this to some point.

    Quote Originally Posted by Princess Celly View Post
    People fuss over it because they feel judged and a formula is telling you that you're bad
    Pretty sure I don't feel bad, and my score isn't telling me i'm "bad".
    Pretty sure Ezuen doesn't believe his score is indicative of his playing ability.

    But I am pretty sure that if ELO is used in ANY fashion to 'balance' matchmaking, ELO should in fact be accurate to a point.

    But ELO scores that are 'set in stone'(by a K value) in a game that quite clearly changes, with players that change, will not ever be accurate unless by coincidence.

  8. #53
    Ascendant Kronos v's Avatar
    Join Date
    Feb 2011
    Posts
    7,182

    Default

    Quote Originally Posted by Ocho View Post
    No one else can see your ELO rating, so you aren't being judged on it. AFKing out however does substantially impact your team, it's absolutely the most detrimental activity you can engage in to ensure your team's loss. People remember that kind of stuff, and I would not be surprised if they harbor resentments over it.
    damn right its resented.
    what is also resented is our inability to actually do anything about an obvious AFKer - because the "mark as AFK" function does nothing.
    the resentment only gets worse when the very next match, there is that AFKer again, continuing his "AFK to a free win" activity.

    we need an effective means of discouraging AFKing in the first place (such as stacking disgrace buff - first afk: 15 mins, 2nd: 30mins, 3rd: 1hr, 4th: 24hrs), of promptly removing AFKers (give us a "mark player as AFK" that doesn't require me to sit in my spawn and harass my team for 5 minutes to get the afker marked) , and of never having to see a known AFK on our teams again (extend to warfronts the LFG feature of not including players on my ignore list on my team).

    Quote Originally Posted by Ocho View Post
    Or to put it more simply, it doesn't matter if you're topping the charts if you're killing the wrong people.

  9.   This is the last Rift Team post in this thread.   #54
    Rift Team Ocho's Avatar
    Join Date
    Mar 2014
    Posts
    1,933

    Default

    I want to take a second to talk about the accuracy of the ELO system. Many people are arguing that item x makes the ELO system inaccurate, or item y makes it inaccurate, or item z. The fact of the matter this is easily testable on our end. If ELO is inaccurate, it will be a bad predictor of winning or losing.

    Many folks argue that premades exert a strong influence over winning or losing. So premades is a simple one to test against. We can look at a large number of matches over time, decide which teams had premades, and which teams have more premades than their opponents, and see if that is a more accurate predictor than ELO.

    We did this. ELO was more likely to predict winner than premades. Turns out the better a group of players is, the better they can work together fairly well even if they aren't in a premade. We also looked at number of people who used healing roles in teams, and a whole bunch of other potential predictors (some that have been posted about in the past, and some that haven't). ELO was the most accurate predictor out of anything we tested.

    The only thing that comes close to being as accurate as ELO was how many people in each team actually participated. I.E. took some sort of active (even if not particularly useful) role in the warfront. But that's not something we can build teams off directly. We can't say person x is actually going to play this round or not. ELO does somewhat account for this though because anyone who tries to leech wins is gonna tank their ELO rating really hard because we have learned they are frequently a detriment to their team.

    So accuracy arguments aren't very strong, because we can look at ELO and test its accuracy. It's pretty darn accurate. If you don't like ELO matching, or how we use it, that's fine... but you can't dislike its accuracy. Unless you want more imbalanced matches more frequently. Then I guess its fine to dislike its accuracy. But that's not a great position to be arguing from.

  10. #55
    Sword of Telara
    Join Date
    Sep 2014
    Posts
    888

    Default

    In an improvement thread a while ago I pointed out a few things based on your original post about how the ELO/matchmaker works which would cause pretty big imbalances.

    The NUMBER of premades is a useless factor. If I queue with a couple of friends who aren't that great, can I get paired with a serious killfarming 1200hit+ premade?

    I don't feel like repeating myself further, so...

    Quote Originally Posted by Foolio View Post
    1. It equates the NUMBER of healers/DPS. It it therefore implied that the ability of the healers/DPS is not taken into account. So we can have t1 geared healers vs potato healers. Hopefully the result is obvious. Also, you can have stupid unhelpful healer combinations like 3 chloros or 3 wardens.
    2. It doesn't care what the spread of ELOs is. It's a case of accuracy vs precision: in arguably every single warfront there are situations where people split off from the team. A team with one good player and the rest bad is at a disadvantage because that player can only be in one place at a time. A team with less spread ELOs won't have this problem. The total team ELOs might be similar but being in one place only is going to put the first team at a disadvantage.
    3. It equates the number of premades. If a good player queues with a few more casual friends, can they be matched against a team of t1 people farming their weekly? It sounds like it to me. And does the size of the premades even matter?
    4. Quitters/afkers, as was pointed out.
    5. Premades skew everything horribly. Your ELO rises with how much you join decent premades, which throws everything out if you then queue with different people or solo.

    Anyways, I went on to suggest why ELO bracketed warfronts are the way to go, as well as plausible ways to dynamically determine suitable brackets to improve the matchmaking without too adversely affecting the queue time.

    Those above points still seem like valid criticisms of the current matchmaker in my books.

    Edit: The above stuff is not why ELO is inaccurate, it's more generally reasons why games CAN be unbalanced with the current system based on information you've provided. The ELO system does a lovely job of forcing 50% w/l ratio, but does "accurate" in your posts mean "closeness of games" or "overall win rate"? The former in my opinion could be much improved and would make warfronts generally more enjoyable...
    Last edited by Foolio; 07-02-2015 at 03:33 PM.
    Spellbook@Greybriar
    Rejuvenation@Greybriar

  11. #56
    Ascendant Zehne's Avatar
    Join Date
    Jul 2012
    Posts
    1,443

    Default

    Still the question remains.

    If a player changes how they play.
    Or an expansion comes out.
    Or a soul is completely re-worked.

    How is a 3500 ELO player going to 'significantly' increase their score(and allow the K value to be reset, and allow their ELO to in turn adjust accordingly).

    Unless the use of the term 'change' should of been used in lieu of 'increase'.

    What you basically told us is that after a certain point, the K value makes it so that a players score is essentially set in stone. And it would take moving heaven and earth to get the K value to reset and allow their scores to change.

    And if 'increase' was in deed intended, that means that players 'stuck' at 4k or higher ELO would need a MIRACLE to get the K value reset. (because, what is a 'significant increase' to a 4k ELO?)

    Even if this 'significant' was only 10% of a persons score(in what world is 10% 'significant' ?) that would mean a 4k Alt would need to gain/lose 400 points at 1-2 points a game just to allow their ELO to change by a more significant factor.

    That would mean a 4k alt would have to purposely lose 200 games in a row just to get the ball rolling.

    The way you described it, the K value system was a series of 1 way streets.

    First off, ELO can change significantly only until the system decides it's changed enough.
    Once that happens, ELO changes are significantly restricted.

    But the only way to reverse this is to 'significantly' change your rating?

    The K value restricts the K value from changing.

    I don't see any reason why the K value should not be reset at certain intervals.

    If a player is rated at 3k, and they play at a 3k level. Then allowing their score to fluctuate for a week or two +/- a hundred points or so( 50-50 W:L avg). Will not have a significant impact on the game.

    But allowing a 4k alt to go into HUNDREDS of WF's with a score that is in no-way indicative of their current 'ability to win' before their score even begins to change will have an impact on the calculation of ELO for everyone that is in those matches.

  12. #57
    Ascendant ecru's Avatar
    Join Date
    Jul 2013
    Posts
    3,741

    Default

    Quote Originally Posted by Ocho View Post
    I want to take a second to talk about the accuracy of the ELO system. Many people are arguing that item x makes the ELO system inaccurate, or item y makes it inaccurate, or item z. The fact of the matter this is easily testable on our end. If ELO is inaccurate, it will be a bad predictor of winning or losing.

    Many folks argue that premades exert a strong influence over winning or losing. So premades is a simple one to test against. We can look at a large number of matches over time, decide which teams had premades, and which teams have more premades than their opponents, and see if that is a more accurate predictor than ELO.

    We did this. ELO was more likely to predict winner than premades. Turns out the better a group of players is, the better they can work together fairly well even if they aren't in a premade. We also looked at number of people who used healing roles in teams, and a whole bunch of other potential predictors (some that have been posted about in the past, and some that haven't). ELO was the most accurate predictor out of anything we tested.

    The only thing that comes close to being as accurate as ELO was how many people in each team actually participated. I.E. took some sort of active (even if not particularly useful) role in the warfront. But that's not something we can build teams off directly. We can't say person x is actually going to play this round or not. ELO does somewhat account for this though because anyone who tries to leech wins is gonna tank their ELO rating really hard because we have learned they are frequently a detriment to their team.

    So accuracy arguments aren't very strong, because we can look at ELO and test its accuracy. It's pretty darn accurate. If you don't like ELO matching, or how we use it, that's fine... but you can't dislike its accuracy. Unless you want more imbalanced matches more frequently. Then I guess its fine to dislike its accuracy. But that's not a great position to be arguing from.
    is accuracy fun? does it make the overall pvp experience more or less enjoyable?
    starplatinum@greybriar
    bluebarbie@greybriar
    twitch
    youtubes

  13. #58
    Ascendant Kronos v's Avatar
    Join Date
    Feb 2011
    Posts
    7,182

    Default

    I don't have a problem with the ELO system as such, more with the way it mixes players of different ELOs onto the same teams.

    I'd rather see players grouped into ELO bands and to play with and against players within the same band.

    I believe the argument against such a thing is insufficient population.

    Quote Originally Posted by Ocho View Post
    Or to put it more simply, it doesn't matter if you're topping the charts if you're killing the wrong people.

  14. #59
    Sword of Telara
    Join Date
    Sep 2014
    Posts
    888

    Default

    Quote Originally Posted by Kronos v View Post
    I believe the argument against such a thing is insufficient population.
    Simple: adjust the band size to suit the number of players online.

    At its worst, it cannot be any more terrible than what we are currently playing.
    Last edited by Foolio; 07-02-2015 at 03:58 PM.
    Spellbook@Greybriar
    Rejuvenation@Greybriar

  15. #60
    Plane Touched detpack's Avatar
    Join Date
    Nov 2011
    Posts
    297

    Default

    When I play COD, there is a robust points system which I believe to be better but similar to the favor system. Assists, kills, objectives, and score streaks, and winning the last round are some of the factors in how many points you receive. The server considers both how well your average score is but weights a large factor on how you did in the previous game with the same players if you are queueing again.

    Here is a screenshot where you see all the points from the last round of a game. The server will attempt to get you into another game after one minute and gives the players a choice to vote between two maps or random map. You can leave that group of players if you wish at any time but if you leave a game, you get nothing including the loot boxes that usually drop midway through the game which are based on activity and achievements whether or not you win. The players open the loot boxes after the game.

    By dividing up players in part on the points they received in the last game, it seems fair to the players involved. There are premades sometimes but it's not the end of the world because you know they are making you a better player and the pugs in Advanced Warfare tend to rally against the premades and call them names if they beat um (it's COD lol).

    The idea that you remain in a lobby with a playing pool of people more instantly accounts for confounding variable factors such as weapon selection or how you are playing that day. Sometimes your hot, and sometimes your not.

    Anyways, I thought I'd share this because it's interesting. I don't think our ELO system is perfect as it is but I don't share my ELO with anyone so it doesn't really matter.

    ELO, and other PvP MMOs-s1_mp64_ship-2014-11-11-22-15-39-28.jpg

+ Reply to Thread
Page 4 of 6 FirstFirst 1 2 3 4 5 6 LastLast

Bookmarks

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts