Guild Wars 2

Can we do away with the malaise of server set win/lose streaks that force players in to a 50/50 ratio? I understand that this renders the ability for most players to eventually hit legendary division because losses only remove 1 pip and wins give 2 on a streak but honestly, aren’t balanced competitive matches that are unpredictable and fun, more important? Does everyone really deserve to be labeled as a legendary competitor within a ranked game that is aiming at esports?

In fact, allowing the match making to create balanced team vs. balanced team, rather than “Team A favored” vs. “Team B unfavored”, would solve a lot of the problems that people are complaining about in this forum, concerning match making:

• Immediately begins sorting out MMRs where they actually should be within divisions.
• Resulting in division icons actually meaning something in terms of player skill levels.
• Legendary would only be achieved by the upper tier competitors in the community.
• Lower skilled players who grind hard, won’t get win streaks in to divisions that they shouldn’t be in, which often reflects in the forum as “Why am I getting 30 losses in a row?” or “How are people making it in to legendary who don’t understand conquest rotations at all?” <- This is no over exaggeration by the way.
• Higher skilled players with less time to play will be less prone to MMR Hell situations.
• All of this, resulting in less blow out matches and fewer situations where players MMRs are being tanked due to starting a season late.

I assume ArenaNet has their reasons for not already doing this. These are reasons I’d like to hear and understand so that I could give better feedback that is more along the plans that they have for structured player vs. player in Guild Wars 2. But never the less, I feel this suggestion is crucial to at least consider for the sake the Guild Wars 2 competitive scene.

I had a 40-game win streak last season, many of the games were slaughters. I’m not sure that was better.

Sure let’s go back to where the game was already over at the start, free pips to the winning side while the losers just sat and waited

Josh Davis.6015:

Hi there,

First, here’s the a recent post on the topic:

Gaile Gray.6029:

With the start of Season 4, we’ll be returning to Season 1’s style of matchmaking which focuses on creating the most balanced teams from the players queueing within your pip range. This change may widen the skill range of players in each division, but we feel that increased match quality is worth the trade-off of reduced prestige for this season. During Season 2 and 3 we added two league features that greatly impacted your ability to climb: start-of-season division placement based on previous season performance and win/lose streak pips, and both of these features will remain enabled for Season 4.

Season 4 matchmaking is already using MMR to create the most balanced teams possible. Teams are no longer “stacked” as they were in Season 2 and 3.

The problem (that we’re solving for S5) is that we’re currently searching for players close to your division to match you against. This means that the pool of players that we can pull from to create a match is much smaller than if we were using the entire pool of players. A player may have a high-enough MMR to put them in the top 100, but if they’re in Sapphire they’re only going to get matched with players that are in or around Sapphire. Season 5 will use the entire pool of players for creating matches – this should help the system create more balanced matches and improve queue times.

As for the “forced 50/50”, there’s nothing on the backend that forces a certain win-rate. This is a byproduct of your MMR settling.

Hopefully things can start go better from now on.

I don’t feel like playing pvp anymore and only played 6 games total and still in sapphire. The queue is long(even in sapphire) and after the long queue you have a good chance of getting some kittenty games. I have completed the wing reward achievement back in season 2. There is simply no incentive to play when there is neither reward or fun involved.

Pips are the issue with leagues. MMR cannot sort effectively and provide a sense of progression if league progression is tied to pips.

MMR system – provides relatively balanced games overall and tends to even out to a 50/50 win rate once your MMR stabilises.

Pip progression – based on win streaks, which are unlikely if your MMR is stable.

Result – 1/-1/1/-1/1/-1/1/-1/1/-1/1/-1/1/-1/1/-1/1/-1/1/-1/ pip progression.

Suggested solution – leagues are MMR based, higher MMR results in higher league standing, relative to the rest of the population and assigned on a normal distribution.

The Ascension(or it’s successor) is tied to PvP achievements that are not linked to league standing, but linked to participation.

However greater league standing lowers the time to complete the various achievements…eg, 150 wins on a class would be baseline and thus applied to amber, emerald and sapphire wins, wins in ruby and diamond would count as two ticks on the achievement progression (so only needs 75) and wins in legendary would count as three ticks (so only requires 50 wins).

Basically the league system is a MMR ladder with cool rank symbols showing your performance relative to the rest of the population and the PvP legendary is a fancy reward track over the length of the year that is faster to complete if you compete in higher leagues.

Josh Davis.6015:

Hi there,

First, here’s the a recent post on the topic:

Gaile Gray.6029:

With the start of Season 4, we’ll be returning to Season 1’s style of matchmaking which focuses on creating the most balanced teams from the players queueing within your pip range. This change may widen the skill range of players in each division, but we feel that increased match quality is worth the trade-off of reduced prestige for this season. During Season 2 and 3 we added two league features that greatly impacted your ability to climb: start-of-season division placement based on previous season performance and win/lose streak pips, and both of these features will remain enabled for Season 4.

Season 4 matchmaking is already using MMR to create the most balanced teams possible. Teams are no longer “stacked” as they were in Season 2 and 3.

The problem (that we’re solving for S5) is that we’re currently searching for players close to your division to match you against. This means that the pool of players that we can pull from to create a match is much smaller than if we were using the entire pool of players. A player may have a high-enough MMR to put them in the top 100, but if they’re in Sapphire they’re only going to get matched with players that are in or around Sapphire. Season 5 will use the entire pool of players for creating matches – this should help the system create more balanced matches and improve queue times.

As for the “forced 50/50”, there’s nothing on the backend that forces a certain win-rate. This is a byproduct of your MMR settling.

so how in the world to i get paired with a group of people then the next match i fought against them and then we in the same team again? if there is no such thing as force 50/50 shouldnt u be in the same group of players together if 10 players only are Qing? not keep flopping back in forth making the w/l redundant.

Josh Davis.6015:

Hi there,

First, here’s the a recent post on the topic:

Gaile Gray.6029:

With the start of Season 4, we’ll be returning to Season 1’s style of matchmaking which focuses on creating the most balanced teams from the players queueing within your pip range. This change may widen the skill range of players in each division, but we feel that increased match quality is worth the trade-off of reduced prestige for this season. During Season 2 and 3 we added two league features that greatly impacted your ability to climb: start-of-season division placement based on previous season performance and win/lose streak pips, and both of these features will remain enabled for Season 4.

Season 4 matchmaking is already using MMR to create the most balanced teams possible. Teams are no longer “stacked” as they were in Season 2 and 3.

The problem (that we’re solving for S5) is that we’re currently searching for players close to your division to match you against. This means that the pool of players that we can pull from to create a match is much smaller than if we were using the entire pool of players. A player may have a high-enough MMR to put them in the top 100, but if they’re in Sapphire they’re only going to get matched with players that are in or around Sapphire. Season 5 will use the entire pool of players for creating matches – this should help the system create more balanced matches and improve queue times.

As for the “forced 50/50”, there’s nothing on the backend that forces a certain win-rate. This is a byproduct of your MMR settling.

This is soooo not true. My observation from 6+k matches say otherwise. If what you say was true, then i wouldn’t get constant blowout (500-100, 100-500, 500-100, 100-500) matches pre HoT and s1, s4. If those teams have equal win chances, then i am Japanese empress.

Also, please explain me how it is possible that we absolutely destroy 4 man premade and next match you are on the team of that 4 man premade.

Josh Davis.6015:

Hi there,

First, here’s the a recent post on the topic:

Gaile Gray.6029:

With the start of Season 4, we’ll be returning to Season 1’s style of matchmaking which focuses on creating the most balanced teams from the players queueing within your pip range. This change may widen the skill range of players in each division, but we feel that increased match quality is worth the trade-off of reduced prestige for this season. During Season 2 and 3 we added two league features that greatly impacted your ability to climb: start-of-season division placement based on previous season performance and win/lose streak pips, and both of these features will remain enabled for Season 4.

Season 4 matchmaking is already using MMR to create the most balanced teams possible. Teams are no longer “stacked” as they were in Season 2 and 3.

The problem (that we’re solving for S5) is that we’re currently searching for players close to your division to match you against. This means that the pool of players that we can pull from to create a match is much smaller than if we were using the entire pool of players. A player may have a high-enough MMR to put them in the top 100, but if they’re in Sapphire they’re only going to get matched with players that are in or around Sapphire. Season 5 will use the entire pool of players for creating matches – this should help the system create more balanced matches and improve queue times.

As for the “forced 50/50”, there’s nothing on the backend that forces a certain win-rate. This is a byproduct of your MMR settling.

I don’t believe this byproduct….

I win a few games with team mates who know what they are doing, they have a minimum of 12k AP each…. A few wins later I’m on a team that doesn’t even have a total of 6k AP…..

The system forces losses so everyone can have medal and it’s kittening bullkitten.

The flip flops are the result of low player population. When there are only so many to choose from, you will see the same players repeatedly.

The game is trying to make an even match and I suspect it doesn’t like to repeat matches. So it’s going to mix and match from a relatively small pool.

This is clearly why they want to pick from the whole population.

So, the root problem is low population. ANET attempted to solve that with the legendary wings, but then shot themselves in the foot by trying to prevent people from getting them.

High rewards aren’t going to solve the issue in and of themselves. I recently got an ascended accessory drop after losing a match badly. I couldn’t have cared less about the reward.

I was unaware of pvp/avoided it for the first period I was in GW2. A big step would be to reach out to the pve population. Perhaps a living world episode that included travelling to an arena?

@JOSH DAVIS Here’s an idea for you. Hold pvp matches in the Queen’s Arena in Divinity’s Reach and perhaps at the Arenas in Rata Sum.

This would make matches easy to spectate and perhaps join and would increase the awareness of pvp in the pve world.

Ithilwen.1529:

The flip flops are the result of low player population. When there are only so many to choose from, you will see the same players repeatedly.

The game is trying to make an even match and I suspect it doesn’t like to repeat matches. So it’s going to mix and match from a relatively small pool.

This is clearly why they want to pick from the whole population.

So, the root problem is low population. ANET attempted to solve that with the legendary wings, but then shot themselves in the foot by trying to prevent people from getting them.

High rewards aren’t going to solve the issue in and of themselves. I recently got an ascended accessory drop after losing a match badly. I couldn’t have cared less about the reward.

I was unaware of pvp/avoided it for the first period I was in GW2. A big step would be to reach out to the pve population. Perhaps a living world episode that included travelling to an arena?

@JOSH DAVIS Here’s an idea for you. Hold pvp matches in the Queen’s Arena in Divinity’s Reach and perhaps at the Arenas in Rata Sum.

This would make matches easy to spectate and perhaps join and would increase the awareness of pvp in the pve world.

i’ll agree w/ most of wut u wrote. sometimes u feel lik a nut u kno

@Cynz.9437

Cynz, two seasons of stacked matches have resulted in badly distorted MMR. This would be particularly true in the upper divisions. People win streaked and teamed well above there true ability. Now, they are expected to perform much better than they really are.

That’s the explanation for your blowout matches.

Ithilwen.1529:

@Cynz.9437

Cynz, two seasons of stacked matches have resulted in badly distorted MMR. This would be particularly true in the upper divisions. People win streaked and teamed well above there true ability. Now, they are expected to perform much better than they really are.

That’s the explanation for your blowout matches.

The problem existed before leagues were introduced….

Thanks for the reply Josh
By the way, I am talking about mostly solo ques.

I see how MMR settling can create win streaks or lose streaks. But then why does my MMR never settle? I’ve played since day one of release and through every ranked season to date, with somewhere around 8000 matches played to show for it. What I notice are consistent patterns of win streaks and then lose streaks. Very rarely do I see win/loss/win/loss rhythms. I also notice exactly when the shift happens. Whenever I begin a season, I start out on a hot foot and usually win the first 10-20 matches with little interruption during that win streak. By “little interruption”, I mean that I may lose 1-2 matches during this initial run. As soon as the win rate begins to push around 75% of the initial run, strange things begin to happen. I begin getting put in to teams with players that are clearly less capable than the opposing team we are against and it results in complete blowout matches. The thing is, this goes on for as many matches as it takes to drop my win rate back down to around 50%. As soon as it reaches around 50%, I will always go on another win streak directly afterwards. From this point forward until the end of the season, I’ll notice the same thing happens at around 60% win rate. The match making begins to throw me in to losing teams that are clearly not capable of handling the players on the opposing teams. When I’m on those win streaks, it’s the exact opposite, I end up on the winning team. It lasts until I hit around 60% win rate. As the season goes on, the win streaks and lose streaks seem to get longer and longer as the whole number based matches played takes longer to bounce back and forth between 50% and 60% win rate.

I believe what you’re explaining to me. I’m confused because this pattern is definitely happening and I’m trying to figure out why. Can the MMR have such volatile change at 8000 matches to create such a scenario?

Josh Davis.6015:

As for the “forced 50/50”, there’s nothing on the backend that forces a certain win-rate. This is a byproduct of your MMR settling.

Well that’s not really true, and you should know better than to be disingenuous in that way.

The system isn’t explicitly forcing a 50/50 win rate, it is a byproduct. You’re correct when you say that. However, it is not a byproduct of your MMR settling. In fact, the current system will prevent your MMR from settling as another byproduct of the way it functions.

The system attempts to make every match perfectly balanced. Another way to say this is that the system attempts to create matches that have a 50/50 chance for either side to win. If every match played is a 50/50 balanced match, then this necessarily produces a 50/50 result for the individual players that participate in these matches.

This results in behavior where players that increase their MMR by winning matches will then tend to decrease their MMR by losing matches. This MMR seesaw does not have a damper on it. There is no functionality in the system to make this oscillation settle to a true value.

The reason this is happening is that the system is artificially trying to create balanced matches from a diverse population. If your selection population varies severely in skill, then the only way to create a balanced match is to split the high skill players into separate teams, resulting in “balanced” matches that have no relation to actual skill level.

In a true competitive setup, the system still aims to create balanced matches, but it does so by attempting to sort players into groups based on skill. It then can choose from a pool of players with similar skill, producing a balanced match because everyone in the match has a similar skill level. Why is this important? Well, in this system, a player that consistently wins is identified as an outlier that has a higher true skill than their MMR suggests. This results in that player moving to a higher skill selection pool. In our current system, however, a player that consistently wins is identified as a strong outlier that needs to be paired with weak outliers to make an average-balanced team.

This all stems from the basic problem of the division system being treated as an additional reward track. A choice needs to be made here: either the division system is merely a reward track with no relation to skill, or it needs to be a purely skill based progression system that is not guaranteed for everyone that plays. You can’t have both.

Josh Davis.6015:

Hi there,

A player may have a high-enough MMR to put them in the top 100, but if they’re in Sapphire they’re only going to get matched with players that are in or around Sapphire. Season 5 will use the entire pool of players for creating matches – this should help the system create more balanced matches and improve queue times.

But.. doesn’t this change amplify the biggest problem with s1 matchmaking which pits high MMR vs high MMR and low vs low in every division, making people with good ratings to struggle even in amber? This would just promote MMR tanking even more, right?

@Fay.2357

You missed a vital point, namely: If each team comes in with roughly equal chances, then it’s up to the players to decide there own fate. It’s no longer an issue of hoping the server will grant you a winning team. Now it’s up to the players to perform.

That’s exactly a “skill based” system. Players who play above expectations advance. ( I am currently running a little better than a 50% win ratio and advancing slowly )

No Ithilwen, you’re missing the whole basic function of the system. I’ll try and explain it a bit better.

We can start by saying skill is represented as a number from 1-10, 10 being high. An “even” match occurs when the total skill on both sides is the same.

In the proper system, the matchmaking is drawing from a pool of people that all have similar skill, let’s say between 7 and 8. So you’ll have one team that’s something like 77888 and the other team is 77788, so total skill is very similar. However, one of those people actually has a higher skill than the system thinks. Instead of being a 7, they’re actually. 10. This unbalances the match and causes them to win. Here’s where the key part is. As a result of winning, the system moves their number from 7 up to 9, and begins matching them with a pool of people between 9 and 10. This way balance is restored AND everyone is playing people of similar skill.

Let’s look at what happens in our system. The pool of players is more diverse. You’ll have teams that look more like 23489 and 25558. If the person with skill 4 wins a lot, the system will then change their number higher, let’s say to 8. This is where the difference happens. As a result of this increase, they now are matched with far worse players in order to keep the teams balanced. Instead of moving up to a higher division of competition, they are kept in the same pool and matched with worse players.

A good system rewards winning by matching the winning player with other higher skilled players until your competition is good enough that you stop winning. Our system rewards winning by attaching bigger and bigger weights to your feet until you stop winning.

Fay.2357:

No Ithilwen, you’re missing the whole basic function of the system. I’ll try and explain it a bit better.

We can start by saying skill is represented as a number from 1-10, 10 being high. An “even” match occurs when the total skill on both sides is the same.

In the proper system, the matchmaking is drawing from a pool of people that all have similar skill, let’s say between 7 and 8. So you’ll have one team that’s something like 77888 and the other team is 77788, so total skill is very similar. However, one of those people actually has a higher skill than the system thinks. Instead of being a 7, they’re actually. 10. This unbalances the match and causes them to win. Here’s where the key part is. As a result of winning, the system moves their number from 7 up to 9, and begins matching them with a pool of people between 9 and 10. This way balance is restored AND everyone is playing people of similar skill.

Let’s look at what happens in our system. The pool of players is more diverse. You’ll have teams that look more like 23489 and 25558. If the person with skill 4 wins a lot, the system will then change their number higher, let’s say to 8. This is where the difference happens. As a result of this increase, they now are matched with far worse players in order to keep the teams balanced. Instead of moving up to a higher division of competition, they are kept in the same pool and matched with worse players.

A good system rewards winning by matching the winning player with other higher skilled players until your competition is good enough that you stop winning. Our system rewards winning by attaching bigger and bigger weights to your feet until you stop winning.

from the bolded parts, i lik this much

Josh Davis.6015:

Hi there,

First, here’s the a recent post on the topic:

Gaile Gray.6029:

With the start of Season 4, we’ll be returning to Season 1’s style of matchmaking which focuses on creating the most balanced teams from the players queueing within your pip range. This change may widen the skill range of players in each division, but we feel that increased match quality is worth the trade-off of reduced prestige for this season. During Season 2 and 3 we added two league features that greatly impacted your ability to climb: start-of-season division placement based on previous season performance and win/lose streak pips, and both of these features will remain enabled for Season 4.

Season 4 matchmaking is already using MMR to create the most balanced teams possible. Teams are no longer “stacked” as they were in Season 2 and 3.

The problem (that we’re solving for S5) is that we’re currently searching for players close to your division to match you against. This means that the pool of players that we can pull from to create a match is much smaller than if we were using the entire pool of players. A player may have a high-enough MMR to put them in the top 100, but if they’re in Sapphire they’re only going to get matched with players that are in or around Sapphire. Season 5 will use the entire pool of players for creating matches – this should help the system create more balanced matches and improve queue times.

As for the “forced 50/50”, there’s nothing on the backend that forces a certain win-rate. This is a byproduct of your MMR settling.

I guess you’re prepared for the Posts that look like sthis: “WHY AM I FIGHTING ZAN AND TEXBI?” “Why is my team fighting legend players when I’m in Ruby?” “This matchmaking is stupid!!??!!!”

witcher.3197:

Josh Davis.6015:

Hi there,

A player may have a high-enough MMR to put them in the top 100, but if they’re in Sapphire they’re only going to get matched with players that are in or around Sapphire. Season 5 will use the entire pool of players for creating matches – this should help the system create more balanced matches and improve queue times.

But.. doesn’t this change amplify the biggest problem with s1 matchmaking which pits high MMR vs high MMR and low vs low in every division, making people with good ratings to struggle even in amber? This would just promote MMR tanking even more, right?

They should just reset MMR, keep the division placement (legend → sapphire), award more pips for beating better players, remove safe nets.

@Fay.2357

Your argument sounds good in theory, but it ignores a basic reality. There aren’t enough players in the various skill bands to run the system the way you are talking about.

( that’s not to even touch on the distortion of skill created by two seasons of stacked matchmaking and the problems of teams carrying people )

Hence, your argument is a non starter.

-edit- I think that the stacked matches may have been an attempt to make and end run around this problem. Unfortunately the “level out” wouldn’t have happened for months or years, if at all. Players used the system to exploit.

I’d also add that this concern with being matched with people who are more or less “good enough for you” is symptomatic of the kinds of attitudes that are helping keep pve players out of pvp.

Josh, although you’re attempting to create the most “balanced” teams possible, it just isn’t working very well. Perhaps this is due to low population, but my idea of a balanced game is one where my MMR is similar to that of everyone on my team, and similar to that of everyone on the other team. Another possibility is highly distorted MMRs from S2/S3 matchmaking style.

However, what I’m seeing is there are rather extreme skill variances. These extreme skill variances are not fun, and don’t lead to very balanced matches, even if 50/50 win/loss is being achieved as a byproduct of the matchmaking. There are way too many matches that are blowouts.

And the events that take place in the matches are often baffling, to say the least. I’ve had a match where the other team triocapped (? – they capped their home, and the other 2 players were nearish there and not doing anything else, so ???) their home and sent 2 mid; I capped our home, and by the time I was halfway to mid (on Legacy, as a Thief) our 4 players at mid were all dead. These kinds of matches, where you need to anticipate the mode of failure from your team being extremely clueless or otherwise bad at PvP in order to win, are unfortunately the majority of all matches I’ve played thus far in the season. And the flipside of this is, you also have to try and find the bad players on the other team and figure out how to exploit them in order to win, as well.

I’d really like to be playing against people of similar skill level to me, and with people of similar skill level to me. Creating balanced teams with such wide skill gaps simply isn’t fun, and it’s also not really helping me improve as a player in any meaningful sense.

Perhaps an MMR reset is in order? Perhaps expanding the player pool will help with this? I don’t know. But if the matchmaking system’s goal is just to create “balanced” games like it currently is doing, without regard to wide skill gaps, I fear that nothing will change. And if nothing changes, you’re going to bleed more competent players who simply cannot enjoy the current style of matches that are being produced.

Oh Grouch, can you guys help us with those with coming changes while you are at it? They just pile up and pile up and pile up

Josh Davis.6015:

Cynz.9437:

Oh Grouch, can you guys help us with those with coming changes while you are at it? They just pile up and pile up and pile up

glory shards go for nothing

can’t sell tomes and with key changes tomes pile up faster than i can use them

Cynz.9437:

Josh Davis.6015:

Cynz.9437:

Oh Grouch, can you guys help us with those with coming changes while you are at it? They just pile up and pile up and pile up

glory shards go for nothing

can’t sell tomes and with key changes tomes pile up faster than i can use them

Save till next season to buy the stuff for season 5.

Josh Davis.6015:

(…)

MMR being adversely affected by S2/S3 style matchmaking is something we’ve been discussing – we’re evaluating a few different options to address it.

As for your other points, take a look at my recent post history – I’ve been discussing some of the problem areas we’re looking to hit in various threads.

Are you ever gonna change the way premades affect mmr and matchmaking? As it is, the way premades’ mmr gets calculated is such a rough approximation that it only really works in matching two full premades against each other. As soon as there are soloqueuers in the mix, the result is neither pretty, nor balanced, and for soloqueuers nearly always a total pain.

Also, queueing with friends that are leagues above someone in terms of skill, inflates mmr, and vice versa. This results in skewed mmr for players that enjoy playing with guildmates/friends. Something that always kept me from queueing with friends.

this matchmaking is the worst. all my games go according to this scenario. (bear in mind ive been legendary x 4 all 3 seasons). so i started late and im stuck in sapphire. so the last game i had in the last 5 minutes dictates all the games ive had thus far…(currently 20). im playing warrior and i decide to go far since the enemy team has 2 thieves and i know 1 of them will take home. i go far and proceed to kill the 2 teefs…a warrior comes while i fully cap the point and i drop him. dont get to stomp because low and behold the 3 teefs come back. im cleaving on the warrior and manage to down 1 teef. the warrior decides to vengeance and i get downed a few seconds later. as him and the teef are stompimg…the warrior loses the vengeance and i get rezzed. i proceed to down the teef and look up at the map…the only point that is capped in blue team favor is my point. so i kill the teefs 2 times each and the warrior once…f lives….and yet my team cant take our 2 other points. so i change my startegy and i include myself in the teamfights but the nodes i wasnt on…were lost..we didnt lose bad..only by 50 points but…with me winning 3v1s..my team should have capatalized much better than that…this has happened on on 20 of my games and im becoming bored of this crap fest.

MMR settling you say? 50% win chance you say?

MMR settling really? I’ve done over 5000 games of PvP now and even in S2/3 where supposedly distorted MMR happened I never got these insane win streaks (my win rates were around 55-65% on main at most).

Still I am getting these trash games with people that shouldn’t even be in sapphire, in diamond. I literally lost a game with Toker on my team because he couldn’t carry 2 trash players that we had. These are players that don’t talk, don’t have a build, keyboard turn and afk on points.

I wouldn’t even feel bad if my team lost cuz the other side was better but this pretty much never happens. Every game this season if I lose it’s because the trash on my team can’t even play PvP at a mediocre level (aka my level). I had another game where they had 3 people camping home so obviously play the other 2 nodes but no, my team would stream in 1 by 1 and feed then complain we shouldn’t push far and focus on home. QQQQQQQQQQ

Ithilwen, there is no roof to hit in season1/4 matchmaking, due to even matchmaking (mind you, Im NOT saying balanced matchmaking).

In season2/3 you’d hit a roof where you always get matched against higher mmr teams.

Totally different things.

Even if Cynz screenshot would show close matchups, in this matchmaking system he’d just have to wait a few days for better players to move on and then get back on winstreak.

If Cynz were where he belongs, that would mean he would be somewhere on the ladder with players of similar skill. This would result in close matchups. I only see two matches that arent blowouts -> your argument about Cynz being where he belongs isnt valid no matter the perspective used to analyse it.

Josh Davis.6015:

The problem (that we’re solving for S5) is that we’re currently searching for players close to your division to match you against. This means that the pool of players that we can pull from to create a match is much smaller than if we were using the entire pool of players. A player may have a high-enough MMR to put them in the top 100, but if they’re in Sapphire they’re only going to get matched with players that are in or around Sapphire. Season 5 will use the entire pool of players for creating matches – this should help the system create more balanced matches and improve queue times.

I can’t understand why it’s taken 5 seasons to understand and fix this obvious issue. Nevertheless, kudos on finally addressing a huge problem with only matching people near the same progression (which has almost nothing to do with skill).

Unfortunately, it doesn’t seem like you guys understand even more obvious issues that will make Season 5 only marginally better than the failures that preceded it.

The first obvious issue is matching premades with solo players. Stop pretending that an algorithm can tell…

1. If premades have optimized profession comps and complementary players builds
2. If they have TS or not,
3. If they have experience playing together

Any of those 3 factors will override the limits of what an algorithm can address, so it’s sheer folly and making you guys look foolish thinking it can.

It’s time to admit the obvious and go back to solo and team ques ASSUMING you really are interested in fair matchmaking and making the game popular/fun to play for solo players.

The second obvious problem relates to a matchmaking algorithm that pays no attention to profession/builds in a conquest only type competition. Again, it’s sheer folly that you guys pretend that an algorithm solves this problem. Your algorithm may say a matchup is “fair” based on skill, but it gets sabotaged when you divide them up into totally mismatched groups of professions/builds.

How are these things so hard to notice/fix??? They aren’t even debatable.

You guys had 3 seasons of major failures that by your own actions required “fixing”. Yet… the “fix” for Season 4 was to go back to Season 1. Wow!

Please just address the obvious and stop thinking algorithms solve your problems. Use common sense.

Ithilwen.1529:

@Cynz.9437

Cynz, two seasons of stacked matches have resulted in badly distorted MMR. This would be particularly true in the upper divisions. People win streaked and teamed well above there true ability. Now, they are expected to perform much better than they really are.

That’s the explanation for your blowout matches.

So explain this.
-I started in the last 11 days of S3, fought my way tooth and claw to mid-Ruby (very few win streaks)
-I started the very minute I hit rank 20 in pvp, so I had the default beginner MMR at the start
-I did zero pvp over the offseason
-I am predominantly soloQ due to being in a PvE-focused guild and not having a presence in the sPVP community due to being new
-I’m routinely seeing blowout games (500-less-than-100, including a full shutout of 500-0 with no DC’s). I’m occasionally seeing 5-man premades vs. 5-man soloQ despite the promises of no 5-man vs full-soloQ. More than 70% of my games are “blowouts” in which the losing team scored less than 200 points, around 30% the losing team scores less than 100.

Now sometimes I’m on the up side of this trend, and sometimes I’m on the down side of it. I’m at about a 45% win rate this season and am 2/3 through sapphire.

But if MMR was to blame I wouldn’t be seeing this.

Josh Davis.6015:

Hi there,

First, here’s the a recent post on the topic:

Gaile Gray.6029:

With the start of Season 4, we’ll be returning to Season 1’s style of matchmaking which focuses on creating the most balanced teams from the players queueing within your pip range. This change may widen the skill range of players in each division, but we feel that increased match quality is worth the trade-off of reduced prestige for this season. During Season 2 and 3 we added two league features that greatly impacted your ability to climb: start-of-season division placement based on previous season performance and win/lose streak pips, and both of these features will remain enabled for Season 4.

Season 4 matchmaking is already using MMR to create the most balanced teams possible. Teams are no longer “stacked” as they were in Season 2 and 3.

The problem (that we’re solving for S5) is that we’re currently searching for players close to your division to match you against. This means that the pool of players that we can pull from to create a match is much smaller than if we were using the entire pool of players. A player may have a high-enough MMR to put them in the top 100, but if they’re in Sapphire they’re only going to get matched with players that are in or around Sapphire. Season 5 will use the entire pool of players for creating matches – this should help the system create more balanced matches and improve queue times.

As for the “forced 50/50”, there’s nothing on the backend that forces a certain win-rate. This is a byproduct of your MMR settling.

TL ; DR : Tank your MMR if you want any kind of advancement in the pips.

Thanks ANet.

Protip: Don’t use a pip system for achievements if this is how you’re going to “balance” the PvP. It makes it agonizing and kittenin irritating to play PvP, because you can’t advance. It’s literally the worst of both worlds — I’m getting a 50% winrate (so I can never have a “good” day or a bad day or some kind of frigging variety), and I can’t advance. I know that if I ever have a “winning streak,” that it’s all gonna go away unless I get really lucky.

This means I’m stuck and not having fun. At least with seasons 2 and 3 you could advance fast and have fun that way. You got some variation, and when you hit the ceiling you stopped playing. But I’m already at the ceiling, and I just started the season. So what now? It’s not fun.

SPESHAL.9106:

Josh Davis.6015:

The problem (that we’re solving for S5) is that we’re currently searching for players close to your division to match you against. This means that the pool of players that we can pull from to create a match is much smaller than if we were using the entire pool of players. A player may have a high-enough MMR to put them in the top 100, but if they’re in Sapphire they’re only going to get matched with players that are in or around Sapphire. Season 5 will use the entire pool of players for creating matches – this should help the system create more balanced matches and improve queue times.

I can’t understand why it’s taken 5 seasons to understand and fix this obvious issue. Nevertheless, kudos on finally addressing a huge problem with only matching people near the same progression (which has almost nothing to do with skill).

Unfortunately, it doesn’t seem like you guys understand even more obvious issues that will make Season 5 only marginally better than the failures that preceded it.

The first obvious issue is matching premades with solo players. Stop pretending that an algorithm can tell…

1. If premades have optimized profession comps and complementary players builds
2. If they have TS or not,
3. If they have experience playing together

Any of those 3 factors will override the limits of what an algorithm can address, so it’s sheer folly and making you guys look foolish thinking it can.

It’s time to admit the obvious and go back to solo and team ques ASSUMING you really are interested in fair matchmaking and making the game popular/fun to play for solo players.

The second obvious problem relates to a matchmaking algorithm that pays no attention to profession/builds in a conquest only type competition. Again, it’s sheer folly that you guys pretend that an algorithm solves this problem. Your algorithm may say a matchup is “fair” based on skill, but it gets sabotaged when you divide them up into totally mismatched groups of professions/builds.

How are these things so hard to notice/fix??? They aren’t even debatable.

You guys had 3 seasons of major failures that by your own actions required “fixing”. Yet… the “fix” for Season 4 was to go back to Season 1. Wow!

Please just address the obvious and stop thinking algorithms solve your problems. Use common sense.

This.

I should never see a team of 3 guards vs. a team of 3 thieves, regardless of whether I benefit from it or not.

HOW kittenING HARD IS IT to at least attempt to not class-stack teams? Sure, when 6 guards are up for queue, there’s gonna be stacked classes. But would it really be that difficult to do 3 and 3 instead of 5 and 1?

Ithilwen.1529:

@Cynz.9437

To use the parlance of S2/S3… perhaps you are where you belong? (this isn’t an attack, I’m sincere.) That looks like the record of a fairly even series of matches.

Now you need to exceed expectations to move up.

If i was where i belong, then
- blowouts wouldn’t happen, but they do, a lot (doesn’t 505-37 strike you odd?). Let’s assume i did reach my personal max MMR, then, logically, my teammates and enemies should be around same MMR, which means the matches should be relative even (400-500 range, outside of dcs/afks etc) – THEY ARE NOT. Not trying to sound arrogant, but I consistently get players (not all of them ofc) that are worse than me as enemy and as teammates, regardless the division i am in (alone the fact that i could solo revs/mes/some DHs/druids in late diamond as thief speaks for itself, i am not even going to start on map awareness/rotations)

- that screenshot was taken mid of diamond, 1-2 days later i moved half division and got in legend, where you face pro league players almost every match. If i was, where i “belong”, it wouldn’t be possible.

- my team of pugs absolutely destroys 4 man premade+1 pug (500-100), next match i end up on the team of that 4 man premade and obviously can’t do much as they were just getting spawn camped all match. Where does it fit?

- ok, let’s assume people got carried in s3 and their MMR was inflated. Why am i still having those players in late diamond/legend now? Shouldn’t have they been punished with lot of losses due to inflated MMR and them underperforming?

Yes, it sucks if you can’t progress, but what really bother is me is that even in higher divisions i get absolute blowout matches (regardless in which team’s favor it is). It shouldn’t be happening in higher divisions, the only thing that is even higher is Pro League, i mean, when will it stop being like that? IT IS NOT FUN.

Lucred.1802:

(…)
HOW kittenING HARD IS IT to at least attempt to not class-stack teams? Sure, when 6 guards are up for queue, there’s gonna be stacked classes. But would it really be that difficult to do 3 and 3 instead of 5 and 1?

You could just as easily ask, how hard can it be for those thieves to log onto something else and play it passably well?

Its a trade-off. Either anet allows character swapping or they include a matchmaking rule to prevent profession stacking.

How would you like having to queue with the character you are gonna play during match?

And who says those dhs all run the same build? One could be a bunker, the others could be dps… Ive even seen a passably well roamer dh. Those thieves could be specced for roaming or for teamfighting.

Its not as easy as a lot of people seem to believe. Personally, I like being able to swap characters. I queue with whatever Id like to play, but if I see that the setup is gonna be very unfavorable, I have no trouble swapping to a different role.

Yasi.9065:

Lucred.1802:

(…)
HOW kittenING HARD IS IT to at least attempt to not class-stack teams? Sure, when 6 guards are up for queue, there’s gonna be stacked classes. But would it really be that difficult to do 3 and 3 instead of 5 and 1?

You could just as easily ask, how hard can it be for those thieves to log onto something else and play it passably well?

Its a trade-off. Either anet allows character swapping or they include a matchmaking rule to prevent profession stacking.

How would you like having to queue with the character you are gonna play during match?

And who says those dhs all run the same build? One could be a bunker, the others could be dps… Ive even seen a passably well roamer dh. Those thieves could be specced for roaming or for teamfighting.

Its not as easy as a lot of people seem to believe. Personally, I like being able to swap characters. I queue with whatever Id like to play, but if I see that the setup is gonna be very unfavorable, I have no trouble swapping to a different role.

If thief players are forced to reroll due little viability of stacked thief team (note: stacked dhs, wars, druids is no problem), then it is either due to crappy balancing or crappy matchmaking design (your pick). Both of those are poor dev choices. Why should player be punished for dev’s poor design decisions and play something they don’t enjoy?

Glad, you like to switch. Some don’t.

Imagine you had to regrind all your leagues progress every day because devs kept pulling the power cord and all data of your progress would be lost from the server on daily basis. THIS is how it feels being forced to reroll because of dev’s poor decisions.

Yasi.9065:

Lucred.1802:

(…)
HOW kittenING HARD IS IT to at least attempt to not class-stack teams? Sure, when 6 guards are up for queue, there’s gonna be stacked classes. But would it really be that difficult to do 3 and 3 instead of 5 and 1?

You could just as easily ask, how hard can it be for those thieves to log onto something else and play it passably well?

Its a trade-off. Either anet allows character swapping or they include a matchmaking rule to prevent profession stacking.

How would you like having to queue with the character you are gonna play during match?

And who says those dhs all run the same build? One could be a bunker, the others could be dps… Ive even seen a passably well roamer dh. Those thieves could be specced for roaming or for teamfighting.

Its not as easy as a lot of people seem to believe. Personally, I like being able to swap characters. I queue with whatever Id like to play, but if I see that the setup is gonna be very unfavorable, I have no trouble swapping to a different role.

Thief and DH were examples. 3+ anything on one team is broken (either for better or worse).

Not everybody has 2-3 characters they’re good on, nor should they really be forced to have them.

People across the board have been calling for no character swapping once queue pops since launch. I really don’t know why it’s even still in the game other than to mitigate the effects of class stacking.

Implement a no-class-stacking policy and get rid of character swapping after queue pop and you solve both class stacking issues and a main-line complaint that’s been a main-line complaint for 3+ years. If the matchmaker was even remotely close to fair you’d never have to/want to swap classes anyway.

I played necro since beta, so I know exactly how it feels being forced to re-roll due to bad balancing.

Nevertheless, I posted it several times already… I prefer a team with 4 good thieves to a team with perfect comp but bad players any day. And I can tell you… that team with good players will always win, no matter the teamcomp.

Because teamcomp isnt as important as people tend to think. Skill is way more important. You can compensate a lot with player-skill.

So you have 3 thieves on your team? Well, then you probably wont win by trying to hold 2 points the entire match. Instead you will have to go for kills, beasts, decaps/caps, always disengaging and keeping the enemy team running.

Lucred, please at least attempt to read the whole post you are quoting. What has class-stacking to do with role stacking? You can have 5 dhs and still get 1 roamer, 2 dps, 2 support. Why shouldnt they be all allowed to play together?

You are trying to solve bad balancing (op build is getting stacked) and lack of build diversity in pvp with some obscure no-profession-stacking-rule. You are only treating the symptoms, but not solving the problem.

I can tell you an easy way to solve a lot of complaints.

Hero builds.

PvP team creates lets say 3 hero builds for each role in pvp. The roles are fixed for each team… lets say 1 roamer, 2 dps, 2 support. You choose a hero build, queue with it, and get put in a team with always the same comp, since its fixed.
No role-stacking, pvp team has dominion over hero builds balancing, even mmr can be hero build dependent instead of general.

Of course, you also at the same time have to give up build diversity and general complexity in pvp.

Thats pretty much the only way to prevent role-stacking.

Lucred.1802:

People across the board have been calling for no character swapping once queue pops since launch. I really don’t know why it’s even still in the game other than to mitigate the effects of class stacking.

And we had that for a tiny period (less than a month) (Aurgust or Semptember 2013 – Xerph and other top players asked that)
It didnt go so well :PPP
edit:(i am sorry….. it was a bug ….. where ppl couldnt join any other PvP ques with different characters , if their maincharacter havent finished that match ….. bug…. tricky bug …mucho lotsa bugs…)

If you and Cynz can find a way to persuade other ppl to reroll to form a more balanced team , then you can have Soloque back , and not be named by pro players ‘’hotjoins version2’’ – where ppl dont listen to others more skilled players advices+tactics
(like in every online game -welcome to the internet!)

@Yasi.9065: yes, team of good players would/should win, even with 3 thieves. Problem is, assuming both teams have decent players of similar skill, the team with stacked thieves is at huge disadvantage.

Switching does make sense in games like OW or LoL because characters themself are not so complex and many of them have very similar gameplay, where in GW2 classes are very complex in comparison. Certainly, we always have that faceroll class per meta, but generally gw2 classes is not something you can simply jump into and do decent with.

Josh Davis.6015:

The problem (that we’re solving for S5) is that we’re currently searching for players close to your division to match you against. This means that the pool of players that we can pull from to create a match is much smaller than if we were using the entire pool of players. A player may have a high-enough MMR to put them in the top 100, but if they’re in Sapphire they’re only going to get matched with players that are in or around Sapphire. Season 5 will use the entire pool of players for creating matches – this should help the system create more balanced matches and improve queue times.

Are you also going to be tying MMR to Division in Season 5? I can see the proposed method working IF division was tied to MMR. However, if it isn’t, the division players end up in will literally be the result of RNG and grind.

The other alternative would be to alter pip gains/loss depending on a player’s MMR. The higher the MMR, the more pips they gain per win, and the less they lose on a loss.

The lower a player’s MMR, the less they gain on a win, and the more they lose on a loss.

Reikou, two words… self-fulfilling prophecy.

Why not just scrap seasons and hand out rewards according to mmr?

The solution to a non-grindy ladder is -> decay + placement matches

Yasi.9065:

Reikou, two words… self-fulfilling prophecy.

How is it a self-fulfilling prophecy?

If you want to get a higher division, improve you game, get higher MMR.

The goal here obviously being to have Division represent how your MMR ranks compared to everyone else.

Yasi.9065:

Why not just scrap seasons and hand out rewards according to mmr?

Personally I don’t see the issue here. Specifically a situation where rewards are based on the highest MMR achieved during the season, or even MMR at the end of the season. If there is one, could you please elaborate on it?

There is also one more issue. Early legend players have 1h que because system is not allowing diamonds. There could be an exeption to allow highest diamonds to play with legends.

This has always been my main suggestion to sort out the problems:

https://forum-en.gw2archive.eu/forum/game/pvp/Best-Suggestion-Ever-For-Ranked-Spvp/first#post6197575

But each time I post this, it doesn’t seem to get much response.

Zui.9245:

Josh, although you’re attempting to create the most “balanced” teams possible, it just isn’t working very well. Perhaps this is due to low population, but my idea of a balanced game is one where my MMR is similar to that of everyone on my team, and similar to that of everyone on the other team. Another possibility is highly distorted MMRs from S2/S3 matchmaking style.

There’s also the (not immediately obvious) problem that the whole MMR system is unlikely to work well when the skill disparities between players in a match is large.

MMR has both a rating and a model of how different MMRs translate into expected winning percentages. You might have, for instance, player A with a rating of 1800, B with a rating of 1700, and C with a rating of 1600; the model would use those numbers as relative points on a curve and predict that perhaps A would beat B 60% of the time, B would beat C 60% of the time, and A would beat C 70% of the time.

Crucially, the model of how MMRs turn into winning percentages is just an assumption – we don’t actually know what that relationship is. It makes a series of normality assumptions and calculates percentages on a logistic curve, but the real world is unlikely to be so simple. However, as long as the skill disparity between players in a match is small, the deviation between the model and reality is also going to (very likely) be small.

From the example above, it means that the model is going to be really accurate when it says A will beat B 60% of the time; it will also be very accurate when it says B will beat C 60% of the time. The prediction that A will beat C 70% of the time, on the other hand, is likely to be less accurate (though probably pretty good).

This usually ends up not being a big issue, because matchmaking will pair players against opponents that it expects to have equal skill; doing so minimizes the impact of the errors in the model, and thus updating of MMR stems mostly from errors in the MMRs themselves – people not being at the right MMR and moving up or down.

Now GW2 is a team game that usually has matchmade teams. You can still apply Glicko to a team game, but there’s another set of assumptions baked in there – how individual MMRs translate into a team MMR. This follows the same sort of rules as above. If everyone on the same team has a very similar MMR, then the assumptions you make about how individual MMRs translate into a team MMR don’t matter too much – if everyone on the team is 1600 MMR, then saying the team has a 1600 MMR isn’t a stretch. However, when there’s a wide variety of MMRs on a team, how does that translate? If you have a team of people with 1200, 1400, 1600, 1800, and 2000 MMRs, what is the MMR of that team? Is it a simple average? Are there carry effects, where the 2000 player is actually more important to the outcome of the match because they will dominate lesser skilled players? Are there feeder effects, where the 1200 player actually matters more to the outcome of the match because they’ll be wrecked repeatedly and the team will have to essentially fight 4v5? Are there more complex interactions? I don’t know, but the general rule is that when there are big disparities on a team, you should have very little confidence in your prediction of what is going to happen in the match.

In a smart MMR system, it would look at the big disparity within a team and assign that match a very low k-value, because the model should recognize it has no idea what is going to happen; that means that MMRs will not update much as a result of that match. I don’t think GW2 has anything like that implemented. More problematic is the pip system, where you may have a team with a radical mix of MMRs, meaning you have no idea if the match you have made will be close at all – but you still give or take a pip for a win or a loss.

tl;dr – Glicko works well for team games like GW2 solo queue when applied under a pretty narrow set of conditions (close match-ups with players of similar skill on each team). Glicko works poorly when match-ups are lopsided or when players on the same team do not have similar skill to each other. Thus, the kind of matches GW2 is making in season 4 are known (or should be known) to be very low quality, unpredictable messes, and the number of unsatisfying blowouts should be expected.