Guild Wars 2

An ideal match occurs within the algorithm.
(Plat, Gold, Silver, Silver, Bronze) is against (Plat, Gold, Silver, Silver, Bronze)
The class comps are even balanced.
(Tempest, Herald, Reaper, DH, Daredevil) vs. (Tempest, Herald, Reaper, DH, Daredevil)
The algorithm has actually done it’s job perfectly from the perspective of sorting ratings/classes amongst two separate teams. This is the most balanced match ever. What could go wrong here?:

• One of the players is a guy who normally plays low plat. He has signed into his gf’s bronze account to mess around for a few hours. He isn’t even trying to match manipulate. He’s just bored.
• One player has reached his gold on the only character he’s proficient with, but today he has decided to grind for pvp wings so he is playing on a class that he has never played before.
• One of the plats signed in after a bad day, he’s playing like kitten and can’t carry his assumed weight within the algorithm.
• One of the silvers signed in on a good day, learns a meta spec and becomes effective. The silver player is now playing more like a high gold.
• One of the silver players entered the match and forgot to change his pve weapons to his pvp weapons.
• Someone smells smoke in their kitchen and must afk to check out what’s going on.
• Someone is honest to god lagging.
• One of the plats is duo with a bronze and they are smurfing with one of the silvers.
• ect.. ect.. ect.. the list of random circumstances goes on.

Given even a couple of the above circumstances in play, the match could easily turn into a blowout and feel like “bad match making” But this is not the algorithm’s fault.
Could any of us have done a better job balancing this match, considering we were made to create it amongst a pool of accounts being played by people that we do not know? No, we couldn’t.

The algorithm does it’s theoretic job perfectly but it cannot see or identify the types of circumstances listed above. Ranked mode is one big gamble, especially with solo/duo only. Between those three or four open PUG slots, the circumstances listed above and flat out match manipulation, it is one big gambit. The above circumstances are real and can never be sorted out from ranked mode, not entirely anyway. The only way to help eliminate such gambits and frustration is to que five man teams, whether ranked is fixed back to 1-5 or we simply go and start doing automated tournaments. That way, we know exactly who is on a team and what to expect.

This season, even more people have lost interest in ranked. With solo/duo only, due to the above circumstances, the match outcomes feel about 20% due to individual contribution and 80% circumstance gamble. If you’re talking duo que, it’s more like 40%/60% but still the point being is that players feel like their individual contribution takes a backseat to the gamble. This is an unsatisfying feeling for any gamer in any game that would lead the gamer to believe that the game’s resolve would be based on skill and not gambling. This isn’t good for attracting players to the game mode.

So what is it that I’m discussing here? Match making really, how it gets botched even though it works perfectly and why it feels bad most of the time. Are there any ways to fix this? Is it even fixable? Is there even anything wrong with the algorithm? I don’t know, you tell me. I can say this though, in a solo/duo only ranked conquest mode, where an individual’s participation represents only 20% of his team, where an individual cannot avoid circumstantial gambits that effect match outcomes, it would feel good to be ranked fully or mostly on personal performance. Leave team win/lose progression to automated tournaments. This way, no matter how bad a match is, an individual can still perform well and have it be counted for. This would make worrying about gambits negligent and conveniently discourage any and all reasons to create match manipulations for the sake of win/loss progression. It even encourages players to keep trying instead of give up and AFK. It all makes sense.

~ That’s my opinion

Having played in an automated tournament, it feels even worse. My team’s score was only double digits at the end. The other team was positively gleeful in curb stomping us.

Long ago, 5 player teams were a separate queue. Perhaps because of the high volume of complaints about match waits, that queue was combined with the solo queue.

Those “premades” then exploited and stomped solos for several seasons. We are now at the point where full teams are strongly disliked. In addition, the farming of solo players has driven enough away that there isn’t population for full teams.

I’d also personally argue that full teams would be blatantly elitist and drive away the last few holdouts, like me.

Fixing PvP has several requirements. One of the biggest ones is more people playing. To achieve that; rewards would have to be increased and wins distributed better.

Distributing wins better would mean a pure solo queue with anti syncing policy. Legendary armor would be a possible option for attractive rewards.

Ithilwen.1529:

Having played in an automated tournament, it feels even worse. My team’s score was only double digits at the end. The other team was positively gleeful in curb stomping us.

Long ago, 5 player teams were a separate queue. Perhaps because of the high volume of complaints about match waits, that queue was combined with the solo queue.

Those “premades” then exploited and stomped solos for several seasons. We are now at the point where full teams are strongly disliked. In addition, the farming of solo players has driven enough away that there isn’t population for full teams.

I’d also personally argue that full teams would be blatantly elitist and drive away the last few holdouts, like me.

Fixing PvP has several requirements. One of the biggest ones is more people playing. To achieve that; rewards would have to be increased and wins distributed better.

Distributing wins better would mean a pure solo queue with anti syncing policy. Legendary armor would be a possible option for attractive rewards.

There’s a saying, when you point the finger at someone else there are 3 more pointing back at you.

When are you going to own up to the idea that you aren’t particularly good and that the matchmaker is trying to get you decent matches? The issue is at low levels as listed above these factors can have a much bigger impact on the game. The better question is, are you being prevented from obtaining all the items and goals higher ranked players get? Not really except winning an AT or getting a fancy title for the leaderboard.

We also missed personal grudges and rivalries from the list. There’s some people who will flat out refuse to play with each other because one of them has a history of being a salty kitten. Even heard people refusing to play because someone is a core spec, certain class, 2 thieves mainly but can be other classes.

Just to say, I entirely own up to the fact that I am terrible at using mesmer in PvP (much better in WvW) and the meta isn’t kind to the power mes I have more experience in playing in WvW. I wouldn’t dream of playing mesmer in ranked because of this, because I know it’s not half as good as a face roll class or my main ele.

Your replies are not relevant to what I said or to the OP. Therefore they merit no further answer.

I don’t honestly think ANET is seeking to make fair matches. I think they believe that “competitive” matches in which one side is heavily favored are better in some way.

If this is the case, expect population to go further down.

Anet had a chance long ago to make pvp decent but they crapped it away. They could have gone with just pure solo que so everyone is the same. that fixes all smurfing unless they really really try hard to help friends with alt accounts that are trash. They could have accounts for ranked play require a even higher rank or even a phone number to make it legit.

Then there is the game play it’s self yeah dancing in and around circles is all fun and good it actually it’s even close to competitive pvp. Courtyard old and new is actually one way to save pvp next to the 2v2 arena. Tho anet failed by not making those que maps so they are doomed to fail already.

For the last 4 years anet has basically done all they can to ruin gw2 pvp. Hope can only come from season 9 (8 will be just as bad) or wait for gw3

Some of these replies are very off topic.
Let me clarify a few things:

• The point of automated tournaments is to purposely form the strongest team possible. We can and should expect blowout matches in automated tournaments. That is what happens in 5 man premade vs. 5 man premade.
• My post was to point out that solo/duo que ranked mode also feels like that, despite having an algorithm that is creating balanced matches and that maybe it shouldn’t have to feel like that
• My post then points out why the algorithm’s match making in ranked que will often create blowout situations, due to circumstances that are impossible for the algorithm to read and sort.
• I then explain what it is about bad match making that frustrates players and that is the aspect of individual contribution taking a back seat to the four other PUGs that roll randomly with the individual. Why is it frustrating? Well it is because an individual can be the highest MMR on a team, carry hard and still lose/get punished by the system, despite outperforming everyone on his own team and the enemy team. This is because progression and ranking is completely based on winning or losing the match. It feels bad when the individual realizes that his own personal skill truly takes the back seat to the luck of the draw on which PUGs he will get on his team. This feels less like an esport and more like egambling. Again, this is not a problem with 5 man premade vs. 5 man premade.
• Then I encourage the idea that a solo/duo only ranked game mode, which is still a 5 man vs. 5 man, but doesn’t allow us to form full teams, should rank players based on personal performance entirely or mostly. Why? Because this is not an automated tournament that feels competitive due to being able to eliminate gambles. No, it is ranked mode solo/duo only and it is one big gamble with who we are que’d with and it feels bad to have poor luck and lose ranking due to a day of bad dice rolling, despite performing at a high level. The last thing you want in a game, is for the gameplay to feel bad because of an automated system rolling dice for you. Thus I proposed the idea of ranking based on individual contribution and one’s own performance. It would eliminate ranking based on dice rolled circumstances and turn it into a solid example of how good a player actually is, despite the kittenty luck that the player may have in que.
  Ranking for once, would solidly mean something.
• I also point out that if ranked leaderboards ran off personal performance, this would actually remedy the cheating epidemic. There would be no reason to run smurfing tactics to win/throw matches when the system is ranking you on personal performance. This would also defuse the frustration that players have concerning match making in general because it wouldn’t matter if we won or lost the match, it would only matter how we performed as individuals. Thus no one needs to get mad at anyone for losing a match. The list of pros for this transition is huge. I honestly cannot think of a single con. It would even better help distinguish the flavor between a ranked season and automated tournaments. Ranked bases on individual performance but Automated is still centered around high performance team wins.

It all sounds good to me anyway.
Not sure why this thread isn’t getting more attention.

I understand what you’re getting at with changing rating gain to a more personal basis rather than dependant on win/loss and while it seems like a good idea, how would you accomplish it? What criteria would you go off of?

In my opinion there are just far too many variables in trying to determine if you pulled your weight, atleast from an algorithm based stand point. In a match it seems easy to judge who is doing well and who isn’t, but thats not always because of a numbers stand point. Many times it is simply being at the right place at the right time, so to repeat my question, what would your criteria be?

Ithilwen.1529:

Your replies are not relevant to what I said or to the OP. Therefore they merit no further answer.

I don’t honestly think ANET is seeking to make fair matches. I think they believe that “competitive” matches in which one side is heavily favored are better in some way.

If this is the case, expect population to go further down.

Matchmaking needs to put similar players on the same team to make the rating algorithm work. The ranked system works as well as any system can. The biggest issue with any system is going to be the people playing it. There will always be people that try to exploit the system to get easier matches, and there will always be griefers that like messing up the matches for others.

@ Rednova
I actually sat last night with pen & paper and tried to flatten out a couple ideas for a system that is based on individual performance. Amongst many problems I stumbled upon that make this almost impossible to do, the single biggest problem is that a personal rating, no matter how it is implemented, has almost nothing to do with winning a match. It will ultimately result in players playing for the individual rating and not the win of the team, which kind of defeats the point of a conquest match. Even if you were to implement a vote/rate system at the end of a match, where all the players rate each other according to performance, which was my first idea actually, there are still just too many problems. If one were to go as far brainstorming as I did last night, they would eventually run in to and see the same sorts of problems. I guess the idea of individual rating goes to the trash can. Even mathematically, there is no practical/reliable way to make this work.

I do still think something needs to be changed though, between the chemistry of the given game modes and that ranked is just broken with how far cheating tactics have evolved, how advantageous and encouraged they are and with how there are too many gambits while in que of ranked for it to be a realistic gauge of a player’s skill level. Either return ranked to 1-5 man que or remove it entirely. Enable year of ascension rewards for unranked and let automated tournaments handle the difference between “Unranked Conquest and Actual Team Conquest”. Automated tournaments is clearly how conquest was meant to be played in the first place. Anyone who has ever been on a solid 5 man team understands this. If they want to make ranked mode solo/duo only, maybe we should test a 2v2 season? Sounds like a lot of fun to me. Other than that, patching & fixing the current ranked conquest truly lies within eliminating this NA cheater mentality and figuring out how to minimize circumstance gambits as listed in the OP post. However it is anet can do this, that’s what needs to happen.

Ithilwen.1529:

Long ago, 5 player teams were a separate queue. Perhaps because of the high volume of complaints about match waits, that queue was combined with the solo queue.

I’d love to see solo duo be replaced with a forced 5 man que system.

Faux Play.6104:

Ithilwen.1529:

Your replies are not relevant to what I said or to the OP. Therefore they merit no further answer.

I don’t honestly think ANET is seeking to make fair matches. I think they believe that “competitive” matches in which one side is heavily favored are better in some way.

If this is the case, expect population to go further down.

Matchmaking needs to put similar players on the same team to make the rating algorithm work. The ranked system works as well as any system can. The biggest issue with any system is going to be the people playing it. There will always be people that try to exploit the system to get easier matches, and there will always be grievers that like messing up the matches for others.

Ok I hear you very well. However out of all of the PvP systems I’ve experienced recently. LoL, DoTA2, and Smite, I don’t even see anywhere close to these problems. As they are removed and quick order even Unranked matches in the games I mentioned above. Have much much higher match quality then GW2 ranked ever had. Yet they have to balance much much much more items and characters and skills then GW2 does.

So sorry, I don’t see that as a good point or excuse for GW2. There is two huge differences I’ve noticed between actual PvP titles and GW2.

-The constant work and effort that goes into the different game’s PvP systems, and keeping grievers and cheaters out or to a very low stand point.

-The biggest one is the difference of business models. Which make all MMO’s PvP systems suffer in comparison to competitive PvP titles.

zealex.9410:

Ithilwen.1529:

Long ago, 5 player teams were a separate queue. Perhaps because of the high volume of complaints about match waits, that queue was combined with the solo queue.

I’d love to see solo duo be replaced with a forced 5 man que system.

That would be the end of PvP in GW2. Sorry, there are not enough elitists to support such a system.

Ithilwen.1529:

Your replies are not relevant to what I said or to the OP. Therefore they merit no further answer.

I don’t honestly think ANET is seeking to make fair matches. I think they believe that “competitive” matches in which one side is heavily favored are better in some way.

If this is the case, expect population to go further down.

Part of it is, more so if we realise the matchmaker is doing what its intended to do, create games for people to play that are balanced within a certain parameter.

You go on loss streaks till you get to where you should be then you get wins and losses in different amounts. This fluctuation can be down to many of the reasons mentioned by the OP. I also added that there are some people who may intentionally throw a game because they see your account name and remember it from the last 5 seasons of PvP posts.

As for the OP, the fact remains that the matchmaker is trying to do as good a job as it can. The issue is the meta changes at a snails pace, there’s no diversity or variety for a lot of classes and many are still stung by the rewards nerf from previous seasons. This leads to a much reduced population playing which only makes the matchmakers job even harder.

The only way I think it can be improved (without shafting the top 100 players like Ithilwen wants to do) is to cap how much you can win or lose from bottom to top. Most complaints about matchmaking come with complaints about the loss of pips.

This argument isn’t going to go anywhere we could have the best match making in the universe people will still complain about losing. All it is are people complaining there’s only so much the Devs can do to hold your hand.

apharma.3741:

Ithilwen.1529:

Your replies are not relevant to what I said or to the OP. Therefore they merit no further answer.

I don’t honestly think ANET is seeking to make fair matches. I think they believe that “competitive” matches in which one side is heavily favored are better in some way.

If this is the case, expect population to go further down.

Part of it is, more so if we realise the matchmaker is doing what its intended to do, create games for people to play that are balanced within a certain parameter.

You go on loss streaks till you get to where you should be then you get wins and losses in different amounts. This fluctuation can be down to many of the reasons mentioned by the OP. I also added that there are some people who may intentionally throw a game because they see your account name and remember it from the last 5 seasons of PvP posts.

As for the OP, the fact remains that the matchmaker is trying to do as good a job as it can. The issue is the meta changes at a snails pace, there’s no diversity or variety for a lot of classes and many are still stung by the rewards nerf from previous seasons. This leads to a much reduced population playing which only makes the matchmakers job even harder.

The only way I think it can be improved (without shafting the top 100 players like Ithilwen wants to do) is to cap how much you can win or lose from bottom to top. Most complaints about matchmaking come with complaints about the loss of pips.

This post illustrates a basic myth that is perpetuated by apologists for “competitive” matches. That is to say, matches that are made easy for high ranking players.

There is no finding a level. There is no balancing out. That theory was tried for two seasons straight. Those seasons were characterized by extreme blowouts. It’s directly responsible for the current population issues.

No, it isn’t my duty to be fed to high ranking players to make them feel good. The vague hope of better matches some time in an undefined future isn’t enough motivation, sorry.

Ithilwen.1529:

apharma.3741:

Ithilwen.1529:

Your replies are not relevant to what I said or to the OP. Therefore they merit no further answer.

I don’t honestly think ANET is seeking to make fair matches. I think they believe that “competitive” matches in which one side is heavily favored are better in some way.

If this is the case, expect population to go further down.

Part of it is, more so if we realise the matchmaker is doing what its intended to do, create games for people to play that are balanced within a certain parameter.

You go on loss streaks till you get to where you should be then you get wins and losses in different amounts. This fluctuation can be down to many of the reasons mentioned by the OP. I also added that there are some people who may intentionally throw a game because they see your account name and remember it from the last 5 seasons of PvP posts.

As for the OP, the fact remains that the matchmaker is trying to do as good a job as it can. The issue is the meta changes at a snails pace, there’s no diversity or variety for a lot of classes and many are still stung by the rewards nerf from previous seasons. This leads to a much reduced population playing which only makes the matchmakers job even harder.

The only way I think it can be improved (without shafting the top 100 players like Ithilwen wants to do) is to cap how much you can win or lose from bottom to top. Most complaints about matchmaking come with complaints about the loss of pips.

This post illustrates a basic myth that is perpetuated by apologists for “competitive” matches. That is to say, matches that are made easy for high ranking players.

There is no finding a level. There is no balancing out. That theory was tried for two seasons straight. Those seasons were characterized by extreme blowouts. It’s directly responsible for the current population issues.

No, it isn’t my duty to be fed to high ranking players to make them feel good. The vague hope of better matches some time in an undefined future isn’t enough motivation, sorry.

Put the tin foil down, back away from the hat. ANet is not out to get you and just you. There is no special marker on ESL player accounts which make them paired up with each other (outside of the manipulators) and so feed them easy wins.

Go watch Frostball’s streams, he’s been playing core specs, fresh accounts and getting them all into platinum solo. Not payed much attention to the stream (sorry Frosty) so dunno if he got to legend but still it shows there’s no special circumstances or barrier to progression that skill does not overcome.

Ah well I guess there’s just no getting through to some people especially when they are so hypocritical.

Also that last bit made me laugh, welcome to competition, the best stand on the shoulders of the defeated.

The “best” would be able to do it on there own merits. In this instance we have “the best” being fed easy wins by being placed on teams that are stacked to be better than there opponents.

That’s neither honorable nor competition. A competition, by it’s very nature and definition, requires a level playing field.

Ithilwen.1529:

The “best” would be able to do it on there own merits. In this instance we have “the best” being fed easy wins by being placed on teams that are stacked to be better than there opponents.

That’s neither honorable nor competition. A competition, by it’s very nature and definition, requires a level playing field.

You have yet to prove the system does feed them easy wins, it is just your extremely warped vision that sees this. Go watch Frostball’s stream, he loses plenty of games even though he worked hard to carry, same with Helseth and Sindrener. They are not fed easy wins any more than everyone else gets the blow out matches in their favour.

Competition especially where there’s a ladder means some winning, some losing and the winners always stand on the shoulders of the defeated, it’s just the way it works.

Duo Q is a compromise to allow play with a friend. It´s not suited for the current conquest setting but it´s useful as long as it´s not exploited …

If conquest is used for a ranked system that means something, then it must be solo or 5 man premade. For solo Q we should get class ranking and thus class locking.
And that not each match will be perfect is quite shure. Lots of human factors as Trevor showed …

@ apharma
Gotta stop you right there. You need to read this and consider that many users are beginning to notice clear evidence of what Ithilwen is talking about:
https://forum-en.gw2archive.eu/forum/game/pvp/matchmaking-algo-BROKEN-w-evidence/first#post6635823

Trevor Boyer.6524:

@ apharma
Gotta stop you right there. You need to read this and consider that many users are beginning to notice clear evidence of what Ithilwen is talking about:
https://forum-en.gw2archive.eu/forum/game/pvp/matchmaking-algo-BROKEN-w-evidence/first#post6635823

Yes I have read that however what information was given?

We got the names of top tier players and that these interesting matches happened to the guy twice in the season.

The thing is what is not said speaks just as loudly to me as what is said.

In the first example he neglects to mention 2 very important things as well as a 3rd issue you raised yourself.

What time was it? If this was at some obscure hour of the night or at say 10am when a lot of people are either at work, school or other activities this would reduce the pool of players significantly.

What was his and CJ’s rating? He didn’t say, he could have the same or similar rating to zeromis and kitten making it fairly equal on paper especially if the other solo players were in the same rating.

The third thing is, were they the real accounts of these people? I only mention this because there has been such a plague of alt accounts in NA that while you might know it’s these people behind the computer we do not know if the rating and MMR is accurately being taken into account by the match maker.

The second example is easy enough to figure out, he queued with a platinum, last I heard it queues a duo as the highest rated player. This was to prevent platinum’s getting easy wins by queuing with bronze and losing the average rating of the duo. So while his alt account was 1480 he was being treated as 1620 to prevent manipulation.

Funny thing is he won the match because of the match maker treating him as 1620 which is probably a better representation of his skill.

Edit: Everyone complains about these mismatches but how often does it come up with a game that’s on paper OK? You listed above tons of external factors the maker has no control over or knowledge of and any of these will tip the game massively. That’s before we even talk team comps and the lack of any way to make that fair without extending queue times to excessive lengths where you will see waits of 15-30 mins across the whole spectrum of players.

I don’t think the time of day, his and CJ’s rating or mains/alts has anything to do with what we are looking at here unless there is something that I seem to be completely misunderstanding. Let’s take a look at this again:

RED team
first duo: 1790 + 1750
second duo: 1700 + 1740
solo: 1680
avg: 1732

BLUE team
duo: 1620 + 1510
solos: 1710 + 1630 + 1510
avg: 1596

So when I read this thread, here is what I see:

• Algorithm takes 3x duo ques and looks at them mathematically. We are told it is supposed to create mathematically balanced matches or so they say.
• Algorithm decides to take the two mathematically strongest duo que teams and stack them together. These are not only the stronger duos but these duos are also comprised of the highest rated players out of the 10 players in the que. That right there alone, doesn’t make any sense mathematically.
• Then it takes the 2nd highest rated solo player, which is only 20 points away from the 1st “virtually the same rating” and stacks it together with the two strongest duos. The red team has been created.
• Now it takes clearly the weaker of the duos, which is ranked significantly lower than the two stronger duos, and puts it on a team with all of the weaker solo players. Blue team is created.

The algorithm, which we are told is designed to create mathematically balanced splits, given the rated numbers in the que, should have done something more like this:
RED
1st Duo – 1790 + 1750
Solo – 1710
Solo – 1630
Solo – 1510
Avg – 1678

BLUE
2nd Duo – 1700 + 1740
3rd Duo – 1620 + 1510
Solo – 1732
Avg – 1660

That makes much more sense in every way, from the perspective of mathematics and balance. The algorithm however, did not do that. That system designed around creating mathematically balanced splits, decided to stack every strongest player on the red team.

I mean, when I look at this, I don’t understand why there are people trying to defend the algorithm here. Whether it was match manipulation, server favoritism or a burp within the algorithm, something is wrong with that match making.

Trevor Boyer.6524:

I don’t think the time of day, his and CJ’s rating or mains/alts has anything to do with what we are looking at here unless there is something that I seem to be completely misunderstanding. Let’s take a look at this again:

RED team
first duo: 1790 + 1750
second duo: 1700 + 1740
solo: 1680
avg: 1732

BLUE team
duo: 1620 + 1510
solos: 1710 + 1630 + 1510
avg: 1596

So when I read this thread, here is what I see:

• Algorithm takes 3x duo ques and looks at them mathematically. We are told it is supposed to create mathematically balanced matches or so they say.
• Algorithm decides to take the two mathematically strongest duo que teams and stack them together. These are not only the stronger duos but these duos are also comprised of the highest rated players out of the 10 players in the que. That right there alone, doesn’t make any sense mathematically.
• Then it takes the 2nd highest rated solo player, which is only 20 points away from the 1st “virtually the same rating” and stacks it together with the two strongest duos. The red team has been created.
• Now it takes clearly the weaker of the duos, which is ranked significantly lower than the two stronger duos, and puts it on a team with all of the weaker solo players. Blue team is created.

The algorithm, which we are told is designed to create mathematically balanced splits, given the rated numbers in the que, should have done something more like this:
RED
1st Duo – 1790 + 1750
Solo – 1710
Solo – 1630
Solo – 1510
Avg – 1678

BLUE
2nd Duo – 1700 + 1740
3rd Duo – 1620 + 1510
Solo – 1732
Avg – 1660

That makes much more sense in every way, from the perspective of mathematics and balance. The algorithm however, did not do that. That system designed around creating mathematically balanced splits, decided to stack every strongest player on the red team.

I mean, when I look at this, I don’t understand why there are people trying to defend the algorithm here. Whether it was match manipulation, server favoritism or a burp within the algorithm, something is wrong with that match making.

If those are indeed the ratings of the players at that time then yes the latter way of making each team have an average of ~1660 should have been done. However when I wrote my post about the event I didn’t have all the user rating and the guy certainly didn’t mention the ratings at the time in his original post, the one you linked. Still a rating difference of 140 should not on paper make a complete blow out match.

How often does the match maker do this though?

The guy said it happened twice in the season, now I don’t know how much he’s played or the sampling but everyone is complaining about these match ups but don’t mention the ones that were mathematically balanced.

I’ll agree that for that match something could be looked at, doesn’t the match maker make a team of 5 then make another team as close as it can to the average rating of the other? If that’s the case it could instead change to a system of getting 10 players in a certain range and creating 2 teams which should make for less anomalies like the example.

You’ll still get the odd game with a 140 average rating difference every now and then but as I said, if it’s not a common occurrence people need to accept it’s good enough.

First, about the match maker:
I had a dev explain this to me once. He used this exact term and said: “the match maker ping pongs players back and forth until it finds the balance”. I assumed that meant the match maker looks at those 10 people in que and starts arranging them until it finds the closest average party mmr vs. average party mmr. Similar to what I did manually when making that recent example. But hey what do I know? Maybe I assumed wrong.

About the difference of average party rating being only 140:
This is deceiving imo. Just as deceiving as looking at the difference between the sums of the party ratings which would be 680. Now that makes the difference sound huge. Regardless, if you were to look at just the average party mmrs and compare them in the same way we would compare individual players, the difference between a 1596 player vs. 1732 is actually huge.

How often is the match maker doing this?
I won’t even go into historical evidence/evaluation. Let’s just talk about s7. So during s7 I ran about 130 matches. The day I decided on leaving GW2 to take a break was the same day I ran almost 20 matches in a row while taking down quick notes of player ratings on each side of the teams, their account names as to identify throw alts and well identified top players so I could see the frequency of old algorithm behavior when they were present. During the first 100 matches, I could feel when something was wrong with a match but I had no evidence. During those 20 matches with notes, I could see evidence when something was wrong with a match and it was happening quite frequently actually. I was noticing that certain players would much more often than not, have a ridiculously stacked team. Whether I was with them or not. Sometimes I would get lucky and be with them and get to steamroll a team of all players who were 200 rating or bellow us or sometimes I would be against them and have to try and carry a team of players who were all 200 rating or lower, than the recognized players we were against. Some of these are people that I duo with and when I duo with them, the match making evidentially seems to favor their presence. This is no embellishment. I don’t know if it’s simply match manipulation, server favoritism or a bit of both but actually seeing it with evidence, ruined the experience for me. You could call it luck but for how long of a duration of time actually seeing this can you keep calling this luck? That’s my testimony anyway and even though it is more detailed than others, it isn’t that different than the words of other forum users who are noticing the same sorts of things.

As far as balanced matches:
Honestly when I played unranked, the game seems to toss a much higher frequency of balanced matches. Like win 500 to 450 or lose 450 to 500. When I played ranked, I very rarely see balanced matches. It’s usually I win on a blowout or a lose on a blowout. Something is very wrong what that.

I just want to add that, if you are a player that does inhouses/scrims, you’ll see what REAL matches feel and look like. You’ll learn that these top players, although they are the best players, they are not indomitable and their presence certainly does not warrant an absolute blowout. You’ll see that they are just a hair better than you in combat and they are just a hair faster than you in rotations. Their presence will warrant situation where they decide to 1v2 at home point, barely survive long enough to allow their team to get just barely ahead in points on other 2 nodes to win the match. They are great players! but they do not create 500 to 50 win situations on a consistent basis during inhouses/scrims, not unless the team you are scrimming is just terribly new.

Point being: To see these players getting one mathematically favored match after another in ranked mode is just… very curious indeed when those matches are being made by an algorithm that is supposed to be creating balanced matches. If they were winning 500 to 450, I wouldn’t question it. But when they are repeatedly getting smash matches vs. a bunch of guys 200 rating or lower, it’s not right. I mean with the way the algorithm says it’s supposed to work, the higher MMRs/Ratings in the game should be pulling in the lower PUGs, not being stacked with other high MMR/Rated players go reread the algorithm notes.

Haven’t logged in for a couple weeks.
Anything changed since the MM dishonors?

One could argue there is nothing “perfect” about any match that involves plat and bronze players together regardless of whether the numbers on each side are equal. We have to get past the idea that as long as the average MMR on both sides are close to equal that means good. Bronze should only play with other bronzes and parts of silver. This is a weakest link type game meaning you are generally only as good as the weakest players on your team because the weak will be targeted and culled. They will not be able to win fights where they are the majority of the players involved for one side. Its not fair to have players in with people much more skilled then they are. Conversely its not fair for better players to be teamed with people much less skilled then are. Until they are willing to trade increased queue times for better matches better matches are not going to happen consistently. MMR averaging teams is a flawed concept when you allow too big of a range of players to be involved.

Just to chime in quickly here with something that I seem to recall but am too lazy to sift through pvp changelogs to look for: The matchmaking algorithm used to create teams with as close of an average rating as possible, i.e. pick ten players then sort them so that the resulting average team ratings were as close as possible. Then at some point when the MMR was made visible, if memory serves, the system was tightened up to force teams of roughly equal skill to face other teams so as to better reflect the nature of the game mode, i.e. the algorithm was changed to first pick five players for a team and then match the teams together as closely as possible while maintaining reasonable queue times. Thus far what I remember. Now the addendum:

This however resulted in (some) people complaining that they were getting dragged down by having a random noob thrown into their team … and for making the good players thus get dragged down to the average while the bad players got dragged up to the average. Which is quite the opposite of what the matchmaker should be doing – its main purpose is to create reasonable matches that get you to your skill rating the quickest. Those matches are not always going to be good quality matches by default because unless you’ve played over a hundred games or so, your rating is still fairly volatile in terms of accuracy, and as far as I’m aware the system does not sort together people with a similar number of games played.

In my personal experience after this change went through, the match quality did improve overall – more matches that were not blowout matches and matches that did have swings, which was not something I had been seeing very often prior to this. HOWEVER, this has generally only begun being the case after me having played somewhere between 50 and 75 games in that particular season.

So to conclude, you really can’t make the system suiting everyone. People at the top, if they queue together, by that very virtue of being at the top, will not have enough people within a reasonable skill bracket to queue with/against and thus will be matched against lower-rated players. Doesn’t matter who, that’s just the nature of the top 1% (let’s say it’s 1%). Those matches will either be what we have now, blowout matches that confirm the rating of the top-rated players (while ideally also not giving a big bump to the lost rating of the losing team) or matches where those top players are split and the match is just random chaos until someone identifies a weak link and then a blowout happens, and the weakest link is (possibly) ridiculed over team chat for being so weak and dragging everyone down. Personally? I prefer the system we have now when you can’t really blame any single player on the team for dragging everyone down and if you do end up on the losing side in a serious blowout match, it’s usually not going to cost you too much (barring smurfing and match manipulation, both things which the matchmaker can never account for). And yes, I’ve been on the receiving end of said blowout matches. Despite that my rating has remained stable over the past two weeks with a deviation of roughly 50 points (or about 3 games’ worth of MMR) either way (lucky winstreaks or loss-streaks moving it either way).

So in my opinion, if anything needs changing, it’s not the baseline algorithm, it’s the details of how closely the two teams should be in terms of rating (directly relating to queue times) as well as things not related to the matchmaker directly, namely class balance, map design and game modes available.

Just be like me and carry yourself to legendary by doin 1vXs, quick finishing off team fights and decapping blah blah idk gw2 is ez money ey

Olterin Fire.5960:

So in my opinion, if anything needs changing, it’s not the baseline algorithm, it’s the details of how closely the two teams should be in terms of rating (directly relating to queue times) as well as things not related to the matchmaker directly, namely class balance, map design and game modes available.

Pretty much this.

The effects of any form of “naughty list” bias or favor given to a particular account (that may not even exist) would get minimized if matching had a tight spread. The same thing goes for random team assignment. Players who are not where they belong in terms of rating will have a greater impact on the match outcome due to their performance rather than the outcome being a matter of predetermined imbalance.

I’m under the impression that the priority of matching opposing teams class for class is mostly responsible for the massive spreads we are seeing, and it’s these massive spreads that are randomizing rating progression.

People kept spamming the forum saying that games are lopsided because of class imbalance (such as one team getting an ele/nec while the other getting two thieves), but in higher levels, most everyone by then should know how to play different classes and be amending their team compositions. Matching by class was a mistake, not only for dropping match quality but also because it’s opened the doors for more creative forms of match manipulation. We’ve seen no improvements from the change.

As long as wait time in queue is part of the equation,the result won’t be the best possible match but the best possible match within a time frame.

Evan Lesh.3295:

I’ll just leave this here:
https://wiki.guildwars2.com/wiki/PvP_Matchmaking_Algorithm