Matchmaking algorithm too rigged, - Page 2 — Guild Wars 2 Forums

Matchmaking algorithm too rigged,

245

Comments

  • BMW.2951BMW.2951 Member ✭✭✭

    @Trevor Boyer.6524 said:

    @Delweyn.1309 said:
    It will never be possible to have perfect matchmaking and so, there will be either too easy or either too hard matches.

    But the problem is not that. It's the fact that sometimes you get stuck in those wins or loses streaks.

    And whe it's streaks, it's not just random. There is something that calculate to puts you with the best or the worse players numerous times in a row. This is the real problem imo. Something must be done about this.

    I wanted to point out this guy's post as he words it pretty accurately. It's not that anyone believes Arenanet employees are sitting back and having a good laugh while directly rigging matches by hand. When they say "It feels rigged" they are referring to whatever the algorithm is doing when it places us into a predetermined stature of winning or losing for that day, regardless of if you are peaking and breaking skill limits or maybe even playing a bit poorly. <- This isn't satisfying play and it makes it so actual player skill is taking a backseat to the algorithm's predetermination. This is frustrating and it does not feel good to play at all.

    Guild Wars 1 random arenas "that had absolutely no match making algorithm in place" felt like more balanced match making than Guild Wars 2 and it still does to this day. If the algorithm in Guild Wars 2 is having so many problems with a lower population, maybe it's time to consider removing it or heavily altering it's functions.

    YES! Make matchmaking completly RNG with only the people from your tier. If it can't find enough people in your tier it needs to throw in a few players that are in equal skill to even out the matches.

  • apharma.3741apharma.3741 Member ✭✭✭✭

    There is another thing many might not be considering.

    You may be at the rating you should be at.

    That doesn’t mean you won’t get loss streaks, some days you won’t be playing at your best, some days you will, you may need time to get into the flow, you may lose focus. The algorithm will do its best to match you but it cannot account for a 1800 rating player being tired, it can’t account for a 1600 rating player playing their best or being a pro at rotation but lacking in combat. It can’t account for you playing a different class or an off meta build.

    There will always be an element of random chance, you’re playing with 9 other strangers in the match and people aren’t robots who play the most optimal thing possible at maximum effectiveness all the time.

  • Dreddo.9865Dreddo.9865 Member ✭✭✭

    So every season people complaining about the MM and those 'streaks' (make a forum search with matchmaking as keyword) are either bad players or people that have big imagination - because the possibility that the matchmaking algorithm is inadequate does not exist (in the perfect world of software we are living).

    The phrase that I believe describes what I also experience is this:
    "I can tell that the system expects her to need carrying, whereas on my my main I am being expected to carry."

    But then the question is why should there be a system (if there is one...) that asks you to carry people? And how does this system decides that carrying the rest of the team is even possible? We have all experienced many matches giving our best (eg. getting ALL top stats) and loosing dramatically like -200 pts or more. How would such a match be possible to win it alone? Because it isn't.

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

    @apharma.3741 Completely agree with you. But with such different gambles existing that match making cannot account for, this should be creating more actual randomization within wins or loses, not predictable win & lose streaks. The fact that I can point my finger to the sky like Babe Ruth hitting a home run and accurately predict when someone "if not myself" is going to be on a win or lose streak for the next 10-15 games, is by far the most questionable thing I have ever seen in any automated match making system, in any of the games that I have ever played.

    @Dreddo.9865 The problem occurs with how the system splits ratings on each team:

    • Say 1 mmr is low and 10 mmr is high for the purpose of this example
    • The system looks at 10x guys waiting in a que for a match. Only these 10 people are que'ing at this given point in time.
    • The mmr skew of the players in que looks like this - 10, 4, 4, 4, 3, 3, 2, 2, 2, 2
    • System takes these numbers and attempts to balance an average party mmr vs. average party mmr.
    • It's going to look like this: RED 10, 2, 2, 2, 2 = 18 vs. BLUE 4, 4, 4, 3, 3 = 18 <- or at least it should be doing this as patch note say. Plenty of evidence showing that it's often doing stuff like splitting this as 10, 4, 4, 4, 3 vs. 3, 2, 2, 2, 2 just stacking strong players on one side, then lower players on other.
    • Anyway, if it is doing what it is supposed to be doing with 10, 2, 2, 2, 2 vs 4, 4, 4, 3, 3, it looks good on paper "average 18 vs. average 18" but this is terrible in applicable due to how conquest is actually played. The reality is that the 10 player will hold whatever node he is at, never die, and kill everyone. But while he is doing this, his 2s are getting crunched on other two nodes by 4s and 3s on enemy team. So enemy team just easily holds two nodes aaand GG. 10 player can't be on 3 or even 2 nodes simultaneously.
  • bravan.3876bravan.3876 Member ✭✭✭✭
    edited April 16, 2018

    @Ben Phongluangtham.1065 said:

    @mortrialus.3062 said:
    Hasn't there been statements by developers stating that that average distance in player score between highest and lowest players per game is less than 50?

    No. The average skill rating between teams is less than 50. The deviation is higher than that. Don't have the number off the top of my head though.

    We do our best to get the standard deviation as close as we can though.

    And no, I say it often, we don't rig matches to end loss or win streaks. But many people will never believe it.

    In S2 there was a mm system makes accounts with already high mmr or new accounts starting with average mmr but ofc still start in amber get winstreaks like 50+ until legend. There clearly was something implemented that made it obvious from the start which team is supposed to win. I would never say anet itself cause win- and losestreaks on purpose but i have the feeling that there are some parts left from the s2 self-fulfilling prophecy matchmaking. I don't know enough about programming but maybe some good computer scientists should take a closer look at the current matchmaking. There is something fishy when i can predict my win and lose streaks aside from running into wintrader and other afk trolls. In the end you can climb over time, but you need way more games for it than i would say were normally needed because you run into matches uncarrybalke no matter how good you are and that is frustrating. While other matches are so ez to win that it is boring.

    I also would like to see some changes in how the rank deduction points gets calculated. When you lose a match 499 to 501 and get like - 25 rankpoints than it is frustrating too, in particular when you carried the whole team and you lose because of another players mistake by helping the losing team to do the beast what our team didn't need to win (seriously my thief run to beast killed half its hp already before the other team even arrived to do it themself, we had a save 2 cap at this point), so the other team killed it sooner and right on time. I obviously was the highest ranked player in my team and it seems also higher than the average of the other team but i get punished that i only could carry 99 mistakes of my teammates but not the 100th right in the end (one old example from s10, no offhour q, and i had already like 80 games played). My solo mmr compared to average mmr of mine or opponent team is just unfair in most matchups in particular when the rating amoung the players in both teams is so different and the variance from the average is so high in both directions.

    "playing revenant is borderline exploiting" - up condimirage 2k18

  • Dreddo.9865Dreddo.9865 Member ✭✭✭
    edited April 16, 2018

    @Trevor Boyer.6524
    Yes I get your point but still why should the "10" be playing with "2" in the same game is what I don't understand. That means that "10" is not properly evaluated by the system and is put in a wrong division. So as I wrote before he either has to carry hard or stuck playing with '2'. That's what happened to me this season and had to duo Q to move along. And the matches at that level were so bad I was about to quit playing this season - the quality of players was of a "newcomer" - even played with a GS (!) firebrand (perma dead whole game) that when I asked him about his build he replied with irony of how we expect firebrands to save our *** lol.

  • apharma.3741apharma.3741 Member ✭✭✭✭

    @Trevor Boyer.6524 said:
    @apharma.3741 Completely agree with you. But with such different gambles existing that match making cannot account for, this should be creating more actual randomization within wins or loses, not predictable win & lose streaks. The fact that I can point my finger to the sky like Babe Ruth hitting a home run and accurately predict when someone "if not myself" is going to be on a win or lose streak for the next 10-15 games, is by far the most questionable thing I have ever seen in any automated match making system, in any of the games that I have ever played.

    @Dreddo.9865 The problem occurs with how the system splits ratings on each team:

    • Say 1 mmr is low and 10 mmr is high for the purpose of this example
    • The system looks at 10x guys waiting in a que for a match. Only these 10 people are que'ing at this given point in time.
    • The mmr skew of the players in que looks like this - 10, 4, 4, 4, 3, 3, 2, 2, 2, 2
    • System takes these numbers and attempts to balance an average party mmr vs. average party mmr.
    • It's going to look like this: RED 10, 2, 2, 2, 2 = 18 vs. BLUE 4, 4, 4, 3, 3 = 18 <- or at least it should be doing this as patch note say. Plenty of evidence showing that it's often doing stuff like splitting this as 10, 4, 4, 4, 3 vs. 3, 2, 2, 2, 2 just stacking strong players on one side, then lower players on other.
    • Anyway, if it is doing what it is supposed to be doing with 10, 2, 2, 2, 2 vs 4, 4, 4, 3, 3, it looks good on paper "average 18 vs. average 18" but this is terrible in applicable due to how conquest is actually played. The reality is that the 10 player will hold whatever node he is at, never die, and kill everyone. But while he is doing this, his 2s are getting crunched on other two nodes by 4s and 3s on enemy team. So enemy team just easily holds two nodes aaand GG. 10 player can't be on 3 or even 2 nodes simultaneously.

    Let’s say your rating is a solid 1500, how much better than a solid 1450 rating player does that make you and how much worse than a 1550 rating player? That’s the variance and I think you would be hard pressed to find a big difference except over time with the higher rated players getting more wins.

    Lets say you are 1500, win 5 games in a row because you’re playing your A game and you were against higher rated players late at night so they were tired and the enemy players just lost but could have been even, you’re now a 1550 rating but you could be fighting 1650 rated players where you’re the lowest rated player in the whole game. We keep the 50 either way variance for 4 players who are 1600 but obviously whichever team the 1550 (1500 actual because win streak altered your rating) are at a disadvantage. You lose that match but your rating hasn’t gone down enough to lower you truly back to the 1500 so you keep getting put in similar games, you keep losing till you’re back to 1500.

    Now is a good time to mention that even at your 1500 rating, if you’re meant to be here then you will win 50-60% of matches and likewise lose 40-50% of matches keeping your rating so you can still keep losing. Eventually you will lose too much rating then bounce back as you get put against players worse than you, think the inverse of above.

    Then there’s the elephant in the room.

    If you won one, lost one, and maintained your rating ANet is fixing a 50% win ratio.
    If you get win streaks and lose streaks, ANet is fixing the algorithm to put you with bad players or in landslide matches.
    If you get a bad team comp vs a perfect one, ANet rigged it.

    Remember when you point the finger at someone else there’s 3 more pointing back at you.

  • HadrianBlackwater.5736HadrianBlackwater.5736 Member ✭✭
    edited April 16, 2018

    Anybody who plays a lot sees behavior of this sort. I won like 9 out 10 the other day and then got two matches straight with two thieves on my team the other team had zero thiefs and 1 thief and were especially tanky and point dominant comps. I proceeded to lose both by an average score of about 500-200. I was ready to throw my computer out the window. They can say whatever they want but when you get on a win steak it definitely seems to look at it as a bogus event and tries to correct it by putting you in more difficult matches until you lose a few games. Its like your placement put you at 1450 good luck getting much past that for long.

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

    @apharma.3741 100% see what you're saying

    The thing that fuels my hard argument is when I am playing on alts or on my gf's account, then I am seeing myself taking various different positions around the leaderboard and the only thing that keeps me in those positions are the strange individual patterns that seem to be attached to each individual account.

    • My main account - Generally plays around 1600-1650. Bad lose streak always starts in between 1600 &1650, it will happen two or three times in a season, and it will not stop until my w/l ratio goes down under 50%. At that point, I'll be plat 1 or gold 3 and then it stops. It lets me climb back to around 1650, then the exact same thing happens again, so on and so forth.
    • My gf's account - For some reason this account allows me to play 1700+ and the bad matches just seem to never happen at all. Most of her matches are easy or balanced at the worst, rarely feel the kind of heat and expectation to carry that is tossed at my main account. The strange thing here is that this is a core account, she doesn't even have HoT or PoF. So when I am on this account I have to play core builds and I guarantee you that I am not as strong as core builds as I am on my main account with my Druid. I've been using her account for 3 or 4 seasons now, so it's not that this account hasn't adjusted to my personal skill level. Does this make any sense? uuuuh yeah sure.
    • My Alt account - For some reason this account is not allowed to make it past gold 2. Each time I breach gold 3 and near plat 1 on this account, it gets slammed with an epic lose streak of impossible to carry games until It's dropped down into gold 1 again. This happens in the exact same fashion, every time. This is also a core account. This pattern I see is the same as what happens to my main account, but it begins happening much earlier to this account, than it does to my main account. This account gets slammed in late gold 3 ,every time. Whereas my main account would be given one easy match after another until around plat 2. <- True story.

    But yeah, being the same player on 3 different accounts and seeing 3 very different pattern tags attached to each account, is why I am quite confident in my argument. If you have not experimented with alt accounts like this, you should go do it. Within about a week's time of stabilizing your skill level and MMR on the new account, you may find yourself just as confused as some of the other people in this thread.

  • Highlie.7641Highlie.7641 Member ✭✭✭

    Oh it clearly functions perfectly...

    When you hear a thief state "We need to go far, we can't kill the necromancer" you know your in a good place...

  • Delweyn.1309Delweyn.1309 Member ✭✭✭
    edited April 16, 2018

    @Trevor Boyer.6524 said:

    @Egorum.9506 said:
    So when people with top 250 titles are saying the mmr system is off, is it still a 'git gud! U reached ur cap!' Issue? Because I'm getting sick of uncarryable games, with 40% of the teams total damage, or having a 20:0 kdr and still losing because 3 of my teammates chased around a sb with bunker builds.

    I think the system is using hidden mmr to determine matchups instead of using the shown mmr, and the streaks happen because it's slow to adapt. This would cause the yoyoing

    Exactly, the system uses a mixture of current rating and actual MMR to determine match making. This explains a great deal of poor algorithm behavior in extremes such as:

    • My main account with 12,000+ matches is made to carry every match I'm in. I almost never get paired with people in my same division. 9/10 matches I am carrying guys in mid gold and I might see a legend on my team 1/100 matches I run "that is no exaggeration, it almost never happens." The struggle is real to maintain plat 1/plat 2 when I'm always against teams of full plat or almost full plat.
    • On my gf's account with less than 150 matches played, I can tag plat 2 right out of placement almost every season, with 9/10 or 10/10 wins. One season I came right out of placement at 1710 on that account and it enraged me. Her matches are easy, she almost always gets paired with people who have higher rating than her because her historical MMR is so low. I can tell that the system expects her to need carrying, whereas on my my main I am being expected to carry.

    My only account has nearly 12k matches played too and I experience the same thing that you.
    The worst is in unranked. Nearly all my matches are very hard with a lot just impossible to carry. Even at plat 1 ranked I had easier matches and a better win rate than in spvp.

    But what we talk about can't be understood by players who don't play a lot is pvp. We suffer of those problems only when we play many matches in a row + after a high pvp lvl (428 atm).

    I remember my old days, I didn't suffer so much of those streaks.

  • Bossun.2046Bossun.2046 Member ✭✭✭

    @Ben Phongluangtham.1065 said:

    @mortrialus.3062 said:
    Hasn't there been statements by developers stating that that average distance in player score between highest and lowest players per game is less than 50?

    No. The average skill rating between teams is less than 50. The deviation is higher than that. Don't have the number off the top of my head though.

    We do our best to get the standard deviation as close as we can though.

    And no, I say it often, we don't rig matches to end loss or win streaks. But many people will never believe it.

    TBH Ben, I don’t believe that Anet ever left the matchmaker on season 6 or 7(not sure) that tried to force a 50/50 win ratio on players. I see a lot of people on my guilds/ friend list with near perfect 50/50 including me.

    sugoi monogatari oniichan

  • @Ben Phongluangtham.1065 said:
    I asked for numbers for this season to be pulled since people are always interested:

    Average skill rating difference between teams: 11.866
    Average standard deviation difference between teams: 11.643
    Average rating difference in a match: 98.203 (min rating vs max rating across all players in the match)

    One thing to keep in mind that end score difference never means that the match didn't start off even. Scores tend to snowball in our game for a number of factors. Some due to map layout/mechanic design. Some due to human nature, as people tend to tilt or give up after getting behind by a certain number. Sometimes people play above or below their potential. That's just part of human performance.

    Thanks for the numbers Ben!

    Do you happen to have similar numbers for the time periods where you changed the algorithm in unranked to have more strict matchmaking? (i.e. MMR difference was smaller, but longer queues)

  • cptaylor.2670cptaylor.2670 Member ✭✭✭✭

    @BikeIsGone.8675 said:

    @Ben Phongluangtham.1065 said:
    I asked for numbers for this season to be pulled since people are always interested:

    Average skill rating difference between teams: 11.866
    Average standard deviation difference between teams: 11.643
    Average rating difference in a match: 98.203 (min rating vs max rating across all players in the match)

    One thing to keep in mind that end score difference never means that the match didn't start off even. Scores tend to snowball in our game for a number of factors. Some due to map layout/mechanic design. Some due to human nature, as people tend to tilt or give up after getting behind by a certain number. Sometimes people play above or below their potential. That's just part of human performance.

    Oh silly Ben, how could individual player performance possibly have anything to do with winning/losing streaks. Tilting is a fake-news rumor that has been spread by the media. Nah, rigged matchmaking seems far more likely. /sarcasm

    Pleasant.

    Anyway, is there a cap? Does the average get skewed by duo queues? I noticed the other day someone was mentioning that they had worse look duo queuing which might explain why some matches I’ve had significantly higher rated opponents maybe? I didn’t really ask anyone if they were duos.

    Once again, I simply addressed the fact that the current system seems to keep you at a certain rating whether it’s through loss streaks or win streaks against highly skilled people or people that sometimes seem to not even have keybinds. Just sometimes either impossible at the personal mmr cap and trying to progress or so easy it’s laughable if you’re at the end of your losing streak. But sure, it’s solely skill based every single time. Please feel free to duo queue with me later and teach me the way. I’d love to see such a skilled professional in action with 4 other teammates being blown up 2 seconds into a fight in mid and then spawn camped the remainder of the match.

  • Rufo.3716Rufo.3716 Member ✭✭✭

    The biggest problem is this system is trying to rate individual players in a team oriented system and it fails miserably. I'm usually a higher level gold player, every once in awhile cracking plat. I don't know how many times this season I've seen players sitting on treb, camping beasts, or abandoning point caps for artifacts. This is not gold level play. This is low silver caliber play at best.

    I know many players feel like they are trying to carry matches 1v5 because their teammates actually are that bad. They do nothing but run mid and die or make bad decisions all game long that only help the other team.

    Either somebody is lying about the actual average player ratings, or there needs to be a much better way to measure a player's skill level. I'm really getting sick of having to explain why it's bad to spawn camp beasts and then I get flamed and told I'm an idiot for my trouble. Seems like I'm trying to explain why X action isn't advisable and they would be better off doing Y action instead in 80% of my games. This is why players are mad, and this is why your high level players have left the game. Players don't want to continually play in matches where their skill means nothing.

  • Master Ketsu.4569Master Ketsu.4569 Member ✭✭✭✭
    edited April 16, 2018

    It's not so much "rigged" as it is the result of how arithmetic works with statistical analysis.

    In any team game where rating is taken from the entire team for losing or given for winning, you are going to have players who are ranked slightly lower or slightly higher than they should be since even the #1 player in the world has a limit to what can be carried in a team game. We have even done experiments as a community in the past with legendary players placing themselves in bronze on purpose and it was discovered a highly mismatched player will go on to win about ~80% of their games.
    One way to understand this is to simulate probability:
    http://www.shodor.org/interactivate/activities/ExpProbability/
    Use the spinner that has 1/3 more red than blue to simulate a "good" player that contributes more to winning a 5v5 than losing. You will notice that in roughly 20% of 10 count spins, the result is often not as expected. So while it is unlikely, it is technically within mathematical possibility for a player to be mismatched just through pure factor of luck.

    So how does this contribute to unwinnable games? Simple. Because of the inaccuracies caused by the above occasional irregularity, what will also occasionally happen ( In roughly 10% of games ) is one team gets a bunch of players who are rated higher than they should be all on the same side and the other team gets a bunch of players rated lower than they should. This results in a total steamroll of a match, and there really is no way for the matchmaking system to predict it. This isn't really the fault of Anet, since outliers is just a natural product of any statistical issue where the probability is less than 100%.

    Rather than fixing the matchmaker, Anet can actually fix this issue better through balancing of the classes and making the game more reliant on player skill rather than hard counters and matchup luck. The worse the balance and the worse the influence of luck is, the more statistical "freaks" you are going to end up with that have a chance of triggering a steamroll game if they end up in the same match together. But one thing the community must accept is that even if Anet does balancing absolutely perfect, these games will still occasionally creep up for much the same reason that freak ocean waves happen. Nothing is perfect, and when those imperfections happen to collide it's gg.

  • apharma.3741apharma.3741 Member ✭✭✭✭
    edited April 16, 2018

    @Trevor Boyer.6524 said:
    100% see what you're saying

    The thing that fuels my hard argument is when I am playing on alts or on my gf's account, then I am seeing myself taking various different positions around the leaderboard and the only thing that keeps me in those positions are the strange individual patterns that seem to be attached to each individual account.

    • My main account - Generally plays around 1600-1650. Bad lose streak always starts in between 1600 &1650, it will happen two or three times in a season, and it will not stop until my w/l ratio goes down under 50%. At that point, I'll be plat 1 or gold 3 and then it stops. It lets me climb back to around 1650, then the exact same thing happens again, so on and so forth.
    • My gf's account - For some reason this account allows me to play 1700+ and the bad matches just seem to never happen at all. Most of her matches are easy or balanced at the worst, rarely feel the kind of heat and expectation to carry that is tossed at my main account. The strange thing here is that this is a core account, she doesn't even have HoT or PoF. So when I am on this account I have to play core builds and I guarantee you that I am not as strong as core builds as I am on my main account with my Druid. I've been using her account for 3 or 4 seasons now, so it's not that this account hasn't adjusted to my personal skill level. Does this make any sense? uuuuh yeah sure.
    • My Alt account - For some reason this account is not allowed to make it past gold 2. Each time I breach gold 3 and near plat 1 on this account, it gets slammed with an epic lose streak of impossible to carry games until It's dropped down into gold 1 again. This happens in the exact same fashion, every time. This is also a core account. This pattern I see is the same as what happens to my main account, but it begins happening much earlier to this account, than it does to my main account. This account gets slammed in late gold 3 ,every time. Whereas my main account would be given one easy match after another until around plat 2. <- True story.

    But yeah, being the same player on 3 different accounts and seeing 3 very different pattern tags attached to each account, is why I am quite confident in my argument. If you have not experimented with alt accounts like this, you should go do it. Within about a week's time of stabilizing your skill level and MMR on the new account, you may find yourself just as confused as some of the other people in this thread.

    I guess you’re special then.

    Edit: not meant in a malicious way, can’t believe I have to put this before someone (not Trevor) flips a table and reports it.

  • Master Ketsu.4569Master Ketsu.4569 Member ✭✭✭✭
    edited April 16, 2018

    TL;DR of my above post: Any system that relies on probability and matching with enough chaotic factors involved will eventually produce a sample that shows a highly exaggerated result. In the case of GW2, this manifests as an unwinnable/unlosable match.

    https://journals.ametsoc.org/doi/full/10.1175/JPO2922.1
    -Since Players MMR and chance to win or lose can also be graphed the same as waves, it can be subjected to mostly the same analysis as this page.

  • Rufo.3716Rufo.3716 Member ✭✭✭

    @apharma.3741 said:

    I believe he said in another post they were considering putting peoples ratings after the match but wouldn’t want to do it before the match because it has a high chance of causing toxicity.

    I share his sentiment that the player base on the whole cannot be trusted to use that information properly. I personally don’t think showing rating at the end will be the miraculous discovery anyone thinks it will be.

    People are going to flame people for not knowing how to play anyway. How is this going to change anything? Ya right now, I'm very embarrassed at my rating. But in the end people will actually realize I'm better than what my ranking represents. But people would just think they got paired badly without being able to see those actual numbers.

  • cptaylor.2670cptaylor.2670 Member ✭✭✭✭

    Experiences of matches so far tonight alone, win, lose, win, lose lose lose. So the three losses we had two eles on the team. What does ele bring to this meta that we need two of them? Didn't even pay attention to their build because they were fighting one scrapper at home for the majority of the match and didn't kill him once.

    Second loss, downed two people and had a third run off, I was down, and two of my team ran off to chase the remaining person off point while the other was down on the point. (Keep in mind that before they even came I was fighting two of them and had already downed 1 before I got ambushed by the thief. Later in the match a scourge was down and a scourge from my team was auto attacking them to death. Not using shade skills, AUTOATTACKING. And of course the one on my team got downed without even using barrier once whether it was on cooldown or they just didn't know they had actual skills to press, and finished resing their scourge and wiping the rest. (I was dead after helping down the scourge and running in from having just recapped home for the 10th time because the thief could not kill their Mesmer and of course ambushed by the salty thief. Honestly, I couldn't kill a chrono as a thief either so I don't blame the thief for not wanting to deal with it in this case. Then again, I don't main thief.)

    Third loss, team constantly dies at mid. I go far and am fighting two people, we still don't have mid or another point, and the team still wipes. People rush to bell one at a time. One of them, yet again, an ele, rushes in with 3 people on bell.

    There is nothing I can do about this matchmaking. Sorry. But, matchmaking is failing. I'm sure I'm terrible. Probably the worst player in the game, but even I have enough common sense to not make the mistakes these people are making and they should NOT be in my team or even at this level.

    Even if I were some brilliant strategist and could figure out exactly what I needed to do to "carry" the match be it capping another portal or running a different utility like portal giving me an advantage capping two different points and somehow magically 1v3'ing people as my team wipes, why should this burden be upon me instead of just being matched evenly and with people that are competent enough to not make brainless mistakes? Why should matchmaking force you to have to outskill 9 people, including your own teammates, in order to progress?

    I've had 3 losses now, so I'm sure my next one I will auto attack and enemies will fall over dead, there will be a dc, or two will rage quit and afk in the middle of the fight for a complete 500-100 blowout. Because matchmaking has decided it's time for me to win again.

    I'm sorry, I don't care how many people want to accuse me of being whiny and or insisting that I'm just redirecting my frustration because I'm "terribad," but this matchmaking system isn't working or people have been so completely "tilted" by the current meta that everyone is just auto-attacking or running subpar builds or trying out new classes and not caring whether we actually win or not.)

    Seriously though, what does ele actually bring to ranked right now? What ele build is superior in the current state of this game that it is worth taking over something else that performs the same function more efficiently? Of course this falls back on the balance team, but I guess I need to check metabattle and see what the current build is and why it's favorable to something else.

    Anyway, on to most likely win a game by a landslide and maybe a couple after only to repeatedly lose over simple mistakes or because I have two thieves, and two eles, or two scourges that both go mid and somehow die in the first 10 seconds of the fight.

  • TheGrimm.5624TheGrimm.5624 Member ✭✭✭✭
    edited April 17, 2018

    Last season MM was pretty good, balanced quite often, closer matches, large number of matched teams in regards to classes and numbers. Whatever changes were made for this is one has moved it back by 6 seasons. Way to early to see this many blowouts one way or another. Still small sample size but again, too early to see that many, would reconsider whatever logic was added to be re-evaluated and potentially removed.

    Envy the Madman his musing when Death comes to make fools of us all.
    De Mortuis Nil Nisi Bonum.
    TheGrimm PoTBS/GW1/WAR/Rift/GW2/MWO/ESO/WoT/WoW/D2/HoTS/Civ6/CU/AoC

  • Rufo.3716Rufo.3716 Member ✭✭✭

    Wonder if I said something to kitten some people off. Haven't won a game tonight and keep getting two aholes who Q as other classes then both switch to scourge. I'm sure they must be a duo since they are always on the same team. Getting real sick of this real quick. Fun is just quickly draining from this game all together.

  • Dreddo.9865Dreddo.9865 Member ✭✭✭

    @Ben Phongluangtham.1065
    Thanks for posting these numbers but how do people get their rank is what bugs me. There are numerous reports in these forums bragging about 'getting gold in my first season' which translates to either we have a lot of natural talented players or the system is so 'generous' promoting players to higher divisions. How are players evaluated and are being put into a division? Does total ranked matches count? Does individual match performance count? Does number of deaths count? etc.

  • @Trevor Boyer.6524 said:
    @FlOwMaKeRs.8623 Ty much sir. Another post stating the exact same things that I've been talking about, to contribute to @Illconceived Was Na.9781 and his evidence -> https://en-forum.guildwars2.com/discussion/36523/matchmaking-still-bad-and-disgusting#latest

    Flowmaker mentioned two bad matches, but not any of the others. That really doesn't help make a case that the system is rigged. It just proves that humans are good at noticing results that favor someone else. Actual evidence would include all the matches someone has over a fixed period of time, ideally for many someones, so that the results don't depend on coincidence. And then, we'd still want to compare those numbers to what might happen randomly.

    Here's an example of someone letting their personal experience (literally) color their impression of whether results are random:

    And here's a (non-PvP) example of someone who collected sufficient evidence to start drawing conclusions:

    It's not hard to collect data on PvP matches; it's just tedious to keep track of before & after scores, ratings of team members, ratings of opponents, and so on. Still, if you want to demonstrate to ANet that the algorithm is failing, it's worth doing.

    If you want to vent about how bad some of the matches are, that's fine. But if you want to make an extraordinary claim about the algorithm being rigged, then it's reasonable to expect more than a bunch of anecdotes from like-minded people.

    Hype is the path to the dark side. Hype leads to unfulfilled expectations. Disappointment leads to anger. Anger leads to disgust. Disgust leads to "oh, new shinies! I'm back!"

  • Delweyn.1309Delweyn.1309 Member ✭✭✭
    edited April 17, 2018

    @Ben Phongluangtham.1065 said:
    I asked for numbers for this season to be pulled since people are always interested:

    Average skill rating difference between teams: 11.866
    Average standard deviation difference between teams: 11.643
    Average rating difference in a match: 98.203 (min rating vs max rating across all players in the match)

    One thing to keep in mind that end score difference never means that the match didn't start off even. Scores tend to snowball in our game for a number of factors. Some due to map layout/mechanic design. Some due to human nature, as people tend to tilt or give up after getting behind by a certain number. Sometimes people play above or below their potential. That's just part of human performance.

    C'mon Ben, your pulling out numbers without taking into account the real specific problem. You're more intelligent than that, pls.
    It's like judging a student with his average score in school. We all know that an average score of 12 means nothing if the person has 18 and 17 in some subjects and 4 or 6 in others.
    So if 95% of players are"casual" pvpers, their numbers will dilute the numbers of the 5 other % who are dedicated pvers.

    The report about the streaks is made by players who play a lot in pvp and have this profile :

    • high pvp level (more than 400)
    • a lot of matches played (+ 10 000)
    • can play a lot of matches in a row each day (+ 10)
    • have a lot of matches played in ranked

    Yes, maybe in GW2 it's not the majority of players. But this is probably this kind of player who make pvp living who are the most accurate in their observations.

    Most players probably do only their daily or the minimum of matches required to stay in the rank. And this is this kind of players who say that all is fine with the matchmaking.

    Edit :
    Ben, I suggest that you take a look on accounts of numerous players with more than 10k matches played (in pvp) and have played a lot in ranked this season, and see their matches history since the beginning of the season.

  • Nappa.1904Nappa.1904 Member ✭✭

    That's another thing. I rarely have even matches (maybe 2-3 in 10 that are won/lost within 100). Vast majority of them are 200-300 point steamrolls. If everyone's being paired up with evenly matched skill rating players, why do this happens. Believe me, I am not talking about dcers, afker's throwers. I mean where everyone is actively playing.

  • Frostball.9108Frostball.9108 Member ✭✭
    edited April 17, 2018

    pointless comment, misread

    Frostball - twitch.tv/frostball99

  • Bossun.2046Bossun.2046 Member ✭✭✭

    @Ben Phongluangtham.1065 said:
    I asked for numbers for this season to be pulled since people are always interested:

    Average skill rating difference between teams: 11.866
    Average standard deviation difference between teams: 11.643
    Average rating difference in a match: 98.203 (min rating vs max rating across all players in the match)

    One thing to keep in mind that end score difference never means that the match didn't start off even. Scores tend to snowball in our game for a number of factors. Some due to map layout/mechanic design. Some due to human nature, as people tend to tilt or give up after getting behind by a certain number. Sometimes people play above or below their potential. That's just part of human performance.

    Let's say for one moment that your Matchmaker is working perfectly fine. It grabs people that it believes should be playing together via their rating. At that point, what would be the biggest offender to getting teammates who don't play as if they belonged in that rating?

    I see you mention plenty of reasons as to why a game would snowball and how it's basically the person's fault, but don't mention how currently some of them are getting carried by their class and due to that end up on a match with high rating vs people who truly deserve that rating via good decision making and mechanical skills.

    Don't you agree that having players with an inflated rating that doesn't match their skills make the Matchmaker feel useless?

    sugoi monogatari oniichan

  • Frostball.9108Frostball.9108 Member ✭✭
    edited April 17, 2018

    @Ben Phongluangtham.1065 said:
    I asked for numbers for this season to be pulled since people are always interested:

    Average skill rating difference between teams: 11.866
    Average standard deviation difference between teams: 11.643
    Average rating difference in a match: 98.203 (min rating vs max rating across all players in the match)

    One thing to keep in mind that end score difference never means that the match didn't start off even. Scores tend to snowball in our game for a number of factors. Some due to map layout/mechanic design. Some due to human nature, as people tend to tilt or give up after getting behind by a certain number. Sometimes people play above or below their potential. That's just part of human performance.

    Im curious about what the average rating difference in a match is when there is a legend tier player in it? This season or last season or whatever

    Frostball - twitch.tv/frostball99

  • HadrianBlackwater.5736HadrianBlackwater.5736 Member ✭✭
    edited April 17, 2018

    Averages doesnt mean anything. The poor matches that people are pissed about are hidden behind the good ones. Does 5 matches where everyone is within 35-50 MMR of each other justify 2 where there is a couple hundred MMR range between some players. I would say it does not and you need to tighthen up the allowable MMR range more and if the queues go up so be it. I dont want to play with people 100+ MMR below me ever. If the MMR rating is accurate at all they are not as good as me. I want to sink or swim playing with people as teamates that are exactly the same MMR as I am. I know it cant be exactly so I would say -50 below and +50 above. Why can't you at least try stricter rules and see what it does to the queues. You would be surprised I think how many would be willing to wait 5 - 7 minutes. Right now its 1-2 for me at least most of the time it doesnt need to be that fast.

  • @Trevor Boyer.6524 said:
    I don't think you're quite understanding how extremely difficult it is for a single individual player to provide the kind of sheer raw hard evidence that YOU as a single user are requesting, concerning spvp.

    I'm not requesting a thing. You're making a claim; it's up to you to support that claim with evidence. Otherwise, you're just speculating based on a few bad experiences.

    And incidentally, as a collector of data, I am fully aware of just how difficult it is for a single, individual player to collect that data. However, one player can collate a lot of data from a lot of different people.

    You also go on a bit about all the video that might generate and you're right: no one in their right mind would watch it all. Which is fine; I'm not asking anyone to watch it all. I'm even willing to accept (most) people's word about their match results.

    What I'm not willing to do is accept that any of us can draw conclusions about the overall system based on a few dozen matches.

    So again, if you want to rant about a series of bad results, go ahead. Just don't expect anyone else to agree with the claim that the system is rigged, without actual evidence.

    Hype is the path to the dark side. Hype leads to unfulfilled expectations. Disappointment leads to anger. Anger leads to disgust. Disgust leads to "oh, new shinies! I'm back!"

  • apharma.3741apharma.3741 Member ✭✭✭✭

    @Frostball.9108 said:

    @Ben Phongluangtham.1065 said:
    I asked for numbers for this season to be pulled since people are always interested:

    Average skill rating difference between teams: 11.866
    Average standard deviation difference between teams: 11.643
    Average rating difference in a match: 98.203 (min rating vs max rating across all players in the match)

    One thing to keep in mind that end score difference never means that the match didn't start off even. Scores tend to snowball in our game for a number of factors. Some due to map layout/mechanic design. Some due to human nature, as people tend to tilt or give up after getting behind by a certain number. Sometimes people play above or below their potential. That's just part of human performance.

    Im curious about what the average rating difference in a match is when there is a legend tier player in it? This season or last season or whatever

    Now this is a good question, do the figures change at the higher/lower end of the spectrum.

  • Pato.9320Pato.9320 Member ✭✭

    @cptaylor.2670 said:

    Another thing I've noticed about this system is that if I have a bad match with particular people and choose to not queue for 15 minutes, I still somehow wind up with them on my team. I don't think this is coincidence because the queue time is significantly longer when it wants you to lose. I assume it's intentionally waiting for those other people to get out of the match. Not sure if this is because it wants you to improve with those particular people or because it knows you work poorly together and is assuring that it is a loss. Sometimes after said breaks, even if they're only 5 minutes or so which you would still think would be long enough, you queue up and it instantly pops, because it those people you were avoiding were waiting in queue this entire time and it was just waiting for you before it would generate the match making sure that you're on the same team.

    Matches last an average of 10 minutes. Queues last approx 2 mins. If you want to be in a different group, just wait 6 minutes and you will get different people, because they will all be in another match already (unless they all do the exact same thing, which isn't very probable).

  • ugrakarma.9416ugrakarma.9416 Member ✭✭✭✭

    I believe that the problem is tied to a small number of players at certain times.

    What they should do is make the algorithm perform a different calculation when it has few players available, in a way that guarantees greater variability.

    Let's suppose, that has only 10 players at a certain time, 3 players with very good skilled and 7 very bad. this means that there will never be a balanced match, a team may have 1, 2 or 3 of the good players. the team that has 2 or more will tend to be the winner. there will still be chances of a bizarre match where the 3 good players will be on the same team.

    The only way to ease this problem is to randomize to the maximum, the team that will have 2 or more good players. Ensuring that they do not repeat the same partners. If they repeat, the phenomenon of snow ball loses streak will occur.

    "It's a testament to the folly of the humans and their gods. They say Arah was sacred, but all I see is one big dragon nest."(Rytlock Brimstone)

  • apharma.3741apharma.3741 Member ✭✭✭✭

    @Cal Cohen.3527 said:

    @Frostball.9108 said:

    @Ben Phongluangtham.1065 said:
    I asked for numbers for this season to be pulled since people are always interested:

    Average skill rating difference between teams: 11.866
    Average standard deviation difference between teams: 11.643
    Average rating difference in a match: 98.203 (min rating vs max rating across all players in the match)

    One thing to keep in mind that end score difference never means that the match didn't start off even. Scores tend to snowball in our game for a number of factors. Some due to map layout/mechanic design. Some due to human nature, as people tend to tilt or give up after getting behind by a certain number. Sometimes people play above or below their potential. That's just part of human performance.

    Im curious about what the average rating difference in a match is when there is a legend tier player in it? This season or last season or whatever

    The sample of games with a legend player is pretty low, so I extended it down to 1700 (plat 3+). Here's the data including last season and this season:
    Average skill rating difference between teams: 14.16
    Average standard deviation difference between teams: 13.55
    Average rating difference in a match: 189.71
    Percent of games with average skill rating difference >50: 0.4%

    There's a significant increase in rating range at this level, but as seen the skill rating mean and standard deviation differences between teams are pretty similar. The rating range is always going to be higher at the edges of the rating curve as a tradeoff with keeping reasonable queue times , but it doesn't stop the matcher from making fair teams which is the most important.

    As some others have pointed out, the average game doesn't always tell the whole story. So I also grabbed all matches from last and this season and paired it down to the set of games with a rating range over 200 (these account for about 10% of all matches). Here's the data for that set:
    Average skill rating difference between teams: 19.83
    Average standard deviation difference between teams: 33.67
    Average rating difference in a match: 279.42
    Percent of games with average skill rating difference >50: 7.5%

    While the skill rating and standard deviation differences are a bit higher in this set, these numbers aren't too bad overall.

    I'm not trying to say that the matcher is perfect, but the vast majority of games are pretty balanced.

    Thanks for the data, it’s been interesting to see. Maybe it will put this matchmaking is rigged stuff to bed?

    Who am I kidding we will have another thread about it within a week.

  • Thanks for the data, @Cal Cohen.3527

    Hype is the path to the dark side. Hype leads to unfulfilled expectations. Disappointment leads to anger. Anger leads to disgust. Disgust leads to "oh, new shinies! I'm back!"

  • zoopop.5630zoopop.5630 Member ✭✭✭✭

    300 is flipping huge .... for match making difference in rating like even 150 is big imo.

    That kinda explains a majority of games being blow outs

  • Crinn.7864Crinn.7864 Member ✭✭✭✭
    edited April 18, 2018

    @Dreddo.9865 said:
    @Ben Phongluangtham.1065
    Thanks for posting these numbers but how do people get their rank is what bugs me. There are numerous reports in these forums bragging about 'getting gold in my first season' which translates to either we have a lot of natural talented players or the system is so 'generous' promoting players to higher divisions. How are players evaluated and are being put into a division? Does total ranked matches count? Does individual match performance count? Does number of deaths count? etc.

    The rating systems distributes across a curve with a mean of 1200. Gold starts at 1200, so a player that is placing Gold is merely slightly above average. Your rating placement is done with the exact same glicko2 algorithm that rates you after placements. Placements is merely a mechanic for hiding your rating change for the first 10 matches, the reason the first ten are hidden is because glicko2 has enormously high deviation in early matches, and devs don't want players flipping out over the rating swings during early matches.

    Sanity is for the weak minded
    YouTube

  • Rufo.3716Rufo.3716 Member ✭✭✭

    I've pretty much come down to ranked as being a farm fest and having fun. I seriously just got out of 2 matches as a silver player (been high gold for pretty much ever) got 8 points and 10 points.

    Another question I have is, is MMR and ranked rating the same, or are there 2 different rating for ranked and unranked matchmaking. I seem like I play with higher level players in unranked vs. ranked.

  • cptaylor.2670cptaylor.2670 Member ✭✭✭✭

    @Cal Cohen.3527 said:

    @Frostball.9108 said:

    @Ben Phongluangtham.1065 said:
    I asked for numbers for this season to be pulled since people are always interested:

    Average skill rating difference between teams: 11.866
    Average standard deviation difference between teams: 11.643
    Average rating difference in a match: 98.203 (min rating vs max rating across all players in the match)

    One thing to keep in mind that end score difference never means that the match didn't start off even. Scores tend to snowball in our game for a number of factors. Some due to map layout/mechanic design. Some due to human nature, as people tend to tilt or give up after getting behind by a certain number. Sometimes people play above or below their potential. That's just part of human performance.

    Im curious about what the average rating difference in a match is when there is a legend tier player in it? This season or last season or whatever

    The sample of games with a legend player is pretty low, so I extended it down to 1700 (plat 3+). Here's the data including last season and this season:
    Average skill rating difference between teams: 14.16
    Average standard deviation difference between teams: 13.55
    Average rating difference in a match: 189.71
    Percent of games with average skill rating difference >50: 0.4%

    There's a significant increase in rating range at this level, but as seen the skill rating mean and standard deviation differences between teams are pretty similar. The rating range is always going to be higher at the edges of the rating curve as a tradeoff with keeping reasonable queue times , but it doesn't stop the matcher from making fair teams which is the most important.

    As some others have pointed out, the average game doesn't always tell the whole story. So I also grabbed all matches from last and this season and paired it down to the set of games with a rating range over 200 (these account for about 10% of all matches). Here's the data for that set:
    Average skill rating difference between teams: 19.83
    Average standard deviation difference between teams: 33.67
    Average rating difference in a match: 279.42
    Percent of games with average skill rating difference >50: 7.5%

    While the skill rating and standard deviation differences are a bit higher in this set, these numbers aren't too bad overall.

    I'm not trying to say that the matcher is perfect, but the vast majority of games are pretty balanced.

    I should probably be able to understand what this means, but it's not making a whole lot of sense to me.

    Average skill rating difference between teams being the average of the ratings of the 5 person on one team being a 20 point difference between the other team?
    Deviation difference between teams? Not sure what this is.

    Rating difference in a match between the lowest rated player and the highest? Is this on the same team or just everyone in both teams? Either way that's a pretty huge difference.

    Percent of games with average skill rating difference? Not sure what this is either, but it's weird that it jumps from .4% to 7.5% just in one season? I know that's still less than 10% but it's huge in comparison to the previous. This might not mean anything given that I'm not entirely sure what it is to begin with. lol

  • rwolf.9571rwolf.9571 Member ✭✭✭

    @Cal Cohen.3527 said:

    While the skill rating and standard deviation differences are a bit higher in this set, these numbers aren't too bad overall.

    I'm not trying to say that the matcher is perfect, but the vast majority of games are pretty balanced.

    I wish I had your optimism. But when over 50% of my matches are 1 sided (Win or Lose, mostly losing). I would hardly call that balanced. Maybe the "Big Picture" everything looks fine. But on the individual scale it isn't.

    (In Tier 3 Gold/ T1 Plat for frame of reference)

  • Namless.4028Namless.4028 Member ✭✭✭

    @rwolf.9571 in gold3/plat1 your impact on your matches is very low since you get matched regulary with plat3/legend tier players

  • cptaylor.2670cptaylor.2670 Member ✭✭✭✭

    @rwolf.9571 said:

    @Cal Cohen.3527 said:

    While the skill rating and standard deviation differences are a bit higher in this set, these numbers aren't too bad overall.

    I'm not trying to say that the matcher is perfect, but the vast majority of games are pretty balanced.

    I wish I had your optimism. But when over 50% of my matches are 1 sided (Win or Lose, mostly losing). I would hardly call that balanced. Maybe the "Big Picture" everything looks fine. But on the individual scale it isn't.

    (In Tier 3 Gold/ T1 Plat for frame of reference)

    I was staying in that rank for awhile. Then got down to the beginning of gold 3 and got discouraged so I stopped even trying. Now I've dropped below that a bit and am losing games with a 100-200 point lead because my entire team gets farmed standing in one scourge shade with a thief fighting a warrior off point and 3 people fighting a sword ele for 10 minutes.

    I'm sure I'm absolutely terrible, but I'm not this terrible. I don't know how these people are making it to this level, whether it's imbalance and playing the flavor of the month easy class in placements and getting higher than they should due to serious imbalance and bad class design like scourge in a point capture game, or what the deal is. I honestly don't know. But I'm infuriated at some of the things these people are doing in my matches.

©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.