Quick Question — Guild Wars 2 Forums

Quick Question

Shaogin.2679Shaogin.2679 Member ✭✭✭

When you duo queue, how does it calculate your mmr? To my understanding it just calculates both players in the duo as whoever has the highest mmr. Does it also inflate the duo's mmr or work any differently?

Doc Von Doom

Comments

  • Zenix.6198Zenix.6198 Member ✭✭✭

    Yes, it takes the MMR of the higher rated player for both players.

    Personally, I also believe that having a Duo on a team (regardless of their actual rating) also increases the average team MMR by a small margin.
    I obviously can't prove this, but personally I just find duoQ games to be harder to carry.
    Might also just be my subjective perception, who knows : shrugs :

  • Shaogin.2679Shaogin.2679 Member ✭✭✭

    Actually on the wiki it looks like you get a +50 mmr, unless I'm reading it wrong.

    Doc Von Doom

  • Exedore.6320Exedore.6320 Member ✭✭✭
    edited June 22, 2019

    It's average rating of the players. What ANet said several years about using the highest is wrong or has since changed.

  • Zenix.6198Zenix.6198 Member ✭✭✭

    @Exedore.6320 said:
    It's average rating of the players. What ANet said several years about using the highest is wrong or has since changed.

    any sauce for that?

  • Exedore.6320Exedore.6320 Member ✭✭✭

    @Zenix.6198 said:

    @Exedore.6320 said:
    It's average rating of the players. What ANet said several years about using the highest is wrong or has since changed.

    any sauce for that?

    I tested it several seasons ago. I'm platinum and I had a friend in silver. When we queued together for a few games, opponents were noticeably worse and win adjustment was much lower compared to when I queued solo. Meanwhile his win adjustment was higher than when he solo queued.

  • Idunno about that whole +50 thing, I went on the wiki and couldn't find anything related to DuoQs and how that effects MMR. Without any confirmation from devs, it's impossible to know for sure.

    But Ben threw us a little bone here when DuoQ was first removed: https://en-forum.guildwars2.com/discussion/12161/season-9-queue-changes/p1

    Basically saying that the higher rated two players are queuing together, the less likely it is those two players will be matched against an appropriate team of similarly rating and structure; which is understandable in a game with such a low population.

    From personal experience DuoQs typically just break the matchmaker regardless. There's so few people playing, that DuoQs of any average rating can be thrust into just about any game with rating differences upwards of 200. Ironic, because that would be the same reason it was removed in the first place, but it doesn't seem to have changed at all since.

    Remove Ranked DuoQ pls&ty

  • Shaogin.2679Shaogin.2679 Member ✭✭✭
    edited June 23, 2019

    In the code on the wiki I noticed this line, but not sure what it means:
    GuildTeam affinity="50"/

    Doc Von Doom

  • Trevor Boyer.6524Trevor Boyer.6524 Member ✭✭✭✭

    @Exedore.6320 said:

    @Zenix.6198 said:

    @Exedore.6320 said:
    It's average rating of the players. What ANet said several years about using the highest is wrong or has since changed.

    any sauce for that?

    I tested it several seasons ago. I'm platinum and I had a friend in silver. When we queued together for a few games, opponents were noticeably worse and win adjustment was much lower compared to when I queued solo. Meanwhile his win adjustment was higher than when he solo queued.

    ^ He's right you know.

    I also had tested it myself within the past couple seasons, same type of scenario. Everything in the match results concerning gains/losses indicates that it is using average rating of both players for the duo, not just the higher rated player's mmr/rating.

  • Zenix.6198Zenix.6198 Member ✭✭✭
    edited June 23, 2019

    @Trevor Boyer.6524 @Exedore.6320

    I dunno about that.
    Im pretty sure, I read a dev comment somewhere, that says that the "rating gain/loss system" is its own separate thing and calculates for each player individually.

    That would imply, that its pretty normal (in case of a win lets say) that the higher rated player gets lower rating and the lower rated player gets higher rating.
    Since the calculation basis just compares your individual rating vs enemy team average rating.

    Edit:
    Found it: https://en-forum.guildwars2.com/discussion/19539/pvp-discussion-matchmaking-and-leagues/p2

  • Zenix.6198Zenix.6198 Member ✭✭✭

    @Shaogin.2679 said:
    In the code on the wiki I noticed this line, but not sure what it means:
    GuildTeam affinity="50"/

    could be, that this refers to this.
    https://en-forum.guildwars2.com/discussion/comment/303123#Comment_303123

    Not sure myself, tho

  • Trevor Boyer.6524Trevor Boyer.6524 Member ✭✭✭✭
    edited June 23, 2019

    @Zenix.6198 said:
    @Trevor Boyer.6524 @Exedore.6320

    I dunno about that.
    Im pretty sure, I read a dev comment somewhere, that says that the "rating gain/loss system" is its own separate thing and calculates for each player individually.

    That would imply, that its pretty normal (in case of a win lets say) that the higher rated player gets lower rating and the lower rated player gets higher rating.
    Since the calculation basis just compares your individual rating vs enemy team average rating.

    Edit:
    Found it: https://en-forum.guildwars2.com/discussion/19539/pvp-discussion-matchmaking-and-leagues/p2

    People believe that if a 1600 is duo with a 1200 that it places the duo as a 1600 duo. What we are saying is that it is averaging the two player's ratings and placing that duo as 1400.

    Anyone who has ever been at like 1600 rating, who solo qued for several games around this rating, would clearly notice the steady gains & loses they were receiving. And then if they qued with a 1200 duo, that 1600 player would immediately notice that his gains on wins were half of normal or less, and that his rating loss was triple of normal. He would also notice that the players he is against were clearly much less experienced than what he is put against as a solo que.

    The algorithm most certainly adjusts rating gain & loss based on the individual player's rating. But the reason why the 1600 begins getting less than half of normal rating gains and triple the rating loss, is because being qued with a 1200 is lowering his placement rating, which is why the opponents in the matches are easier, which is why his gains are less and his losses are greater. This is also why the 1200 player will begin getting jacked up gains and much smaller loses, because for him, the average placement rating is being elevated higher than normal vs. his actual rating. So he gets bigger gains on wins and much smaller loses, because the opponents have much higher ratings than he normally would be going against.

    Consequently I'd like to point out that if the algorithm took a 1600 and 1200 as a duo and placed it as a 1600 duo, the 1600 would be playing against all the same players as he would normally, and he would experience no difference in rating gain & loss. <- If you have ever actually taken the time to test all of this yourself, you'd see that this is just not what is happening at all.

    The only thing that explains the fall of opponent skill level when the 1600 ques with the 1200, and the individual player rating adjustments thereafter, is that the algorithm is using the average of the two player's ratings, for deciding the opponents they are against.

  • Zenix.6198Zenix.6198 Member ✭✭✭

    @Trevor Boyer.6524 said:

    @Zenix.6198 said:
    @Trevor Boyer.6524 @Exedore.6320

    I dunno about that.
    Im pretty sure, I read a dev comment somewhere, that says that the "rating gain/loss system" is its own separate thing and calculates for each player individually.

    That would imply, that its pretty normal (in case of a win lets say) that the higher rated player gets lower rating and the lower rated player gets higher rating.
    Since the calculation basis just compares your individual rating vs enemy team average rating.

    Edit:
    Found it: https://en-forum.guildwars2.com/discussion/19539/pvp-discussion-matchmaking-and-leagues/p2

    People believe that if a 1600 is duo with a 1200 that it places the duo as a 1600 duo. What we are saying is that it is averaging the two player's ratings and placing that duo as 1400.

    Anyone who has ever been at like 1600 rating, who solo qued for several games around this rating, would clearly notice the steady gains & loses they were receiving. And then if they qued with a 1200 duo, that 1600 player would immediately notice that his gains on wins were half of normal or less, and that his rating loss was triple of normal. He would also notice that the players he is against were clearly much less experienced than what he is put against as a solo que.

    The algorithm most certainly adjusts rating gain & loss based on the individual player's rating. But the reason why the 1600 begins getting less than half of normal rating gains and triple the rating loss, is because being qued with a 1200 is lowering his placement rating, which is why the opponents in the matches are easier, which is why his gains are less and his losses are greater. This is also why the 1200 player will begin getting jacked up gains and much smaller loses, because for him, the average placement rating is being elevated higher than normal vs. his actual rating. So he gets bigger gains on wins and much smaller loses, because the opponents have much higher ratings than he normally would be going against.

    Consequently I'd like to point out that if the algorithm took a 1600 and 1200 as a duo and placed it as a 1600 duo, the 1600 would be playing against all the same players as he would normally, and he would experience no difference in rating gain & loss. <- If you have ever actually taken the time to test all of this yourself, you'd see that this is just not what is happening at all.

    The only thing that explains the fall of opponent skill level when the 1600 ques with the 1200, and the individual player rating adjustments thereafter, is that the algorithm is using the average of the two player's ratings, for deciding the opponents they are against.

    Well ...we certainly have different experiences then.
    Admitedly, I have never actively tested this, but I have Qd with people with significantly lower rating than mine and I honestly can't remember there being a huge disparity to my regular soloQ experience when it comes to rating gains/losses.

    So I'm sorry to say, but "personal testing" just doesn't make the cut for being a reliable source for me.

    It also doesn't make sense from a design PoV, considering how easily exploitable this would be.
    2 legend players could just Q together. One on their main; the other on their purposely downplayed alt.
    They get gold opponents and just farm easy rating.

  • Exedore.6320Exedore.6320 Member ✭✭✭
    edited June 23, 2019

    @Zenix.6198 said:
    So I'm sorry to say, but "personal testing" just doesn't make the cut for being a reliable source for me.

    It's kitten math! You can't make it lie!
    You have two possible rule sets and you know the mathematical behavior of each. Then you do testing - which anyone can do - and see which rule set is supported by the data. The conclusion is that that data does not support a duo being matched at the higher player's rating.

    The feeling of players being not as good is subjective evidence which reinforces the objective math.

  • Zenix.6198Zenix.6198 Member ✭✭✭
    edited June 23, 2019

    @Exedore.6320 said:

    @Zenix.6198 said:
    So I'm sorry to say, but "personal testing" just doesn't make the cut for being a reliable source for me.

    It's kitten math! You can't make it lie!
    You have two possible rule sets and you know the mathematical behavior of each. Then you do testing - which anyone can do - and see which rule set is supported by the data. The conclusion is that that data does not support a duo being matched at the higher player's rating.

    The feeling of players being not as good is subjective evidence which reinforces the objective math.

    "subjective evidence". Made my day.
    Not to mention that a lot depends on the parameters of your testing.
    Time(s) of day you Qd.
    Weekday(s).
    Sample size.
    Actual ratings of your teammates and the enemy players.

    Hard to judge the quality of the results of your testing without those informations.
    Otherwise it's just my personal experience vs your personal experience....which objectively leads nowhere.

  • Trevor Boyer.6524Trevor Boyer.6524 Member ✭✭✭✭

    @Zenix.6198 there is nothing mentioned in the algorithm notes that I could find, concerning whether it uses the higher rated player's rating for placement, or if it averages both player's ratings for placement. So I first decided to strum back through the patch notes, to see when any of this changed, but I couldn't find anything in any of the past notes either. Then I figured I'd strum back through the pvp sub-forum and look at every single Arenanet tagged thread that had something to do with matchmaking, to find some answers. Here are the results:

    This first quote was from a thread discussion where they were testing a new method of matchmaking in unranked. Ben had made the thread to get community feedback on how it was working.

    @Ben Phongluangtham.1065 said:

    @NICENIKESHOE.7128 said:
    Curious on what happen if you teamup during this experiment though, does MMR just kinda take the average of two?

    It takes the average of the pair.

    There must have been good feedback on the test because not long after, Ben announced this.

    @Ben Phongluangtham.1065 said:
    Hey folks. Just a quick update. We're moving the new matchmaker to offseason ranked within the next few minutes. Additionally, current plan is that this new matchmaker will also be used for Season 10 matchmaking.

    Edit - One additional note: We moved back to the old standard that a match can have +2 of a class on one team rather than +1. We felt that the restriction was, in the end, causing more imbalanced matches. As always, we'll consider to monitor metrics. It's something very easily adjustable for us going forward.

    During that very following season, which was season 10, someone had made this thread -> https://en-forum.guildwars2.com/discussion/24901/season-10-matchmaker/p1 and the OP post was this.

    @Alatar.7364 said:
    I'd just like to ask if the new/fixed Matchmaker is in effect this Season?

    Which was responded to by Cal with,

    @Cal Cohen.3527 said:
    yes­­­­­­­­

    ^ Best I can do for you. I couldn't find any other quotes on the topic at all. Make of it what you want.

    On a side note, after rereading through some of my old postings, I realized how much of a jerk I sound like sometimes, as if I were attacking Arenanet. Just wanted to apologize about that to these two guys specifically, because they have gone out of their way to be incredibly helpful in their responses over the years.

  • Zenix.6198Zenix.6198 Member ✭✭✭

    Holy kitten @Trevor Boyer.6524 .
    This is something I can actually respect. Good job on that. +1 from me.
    Guess I owe you a genuine apology after all.

    On the other hand this makes me question my faith in the Devs.
    Like, how can this be in the game.
    It's clearly exploitable.

©2010–2018 ArenaNet, LLC. All rights reserved. Guild Wars, Guild Wars 2, Heart of Thorns, Guild Wars 2: Path of Fire, ArenaNet, NCSOFT, the Interlocking NC Logo, and all associated logos and designs are trademarks or registered trademarks of NCSOFT Corporation. All other trademarks are the property of their respective owners.