Guild Wars 2

Rieselle.5079:

They should modify the matchmaking system based on the assumption that most people tend not to switch classes, and most people only have small number of classes they will prefer to play.

So a subset of the population doesn’t have a mindset of playing to win. And we should construct a matchmaking and ranked system around them? Ya, bad idea.

Since double-thief is the most common example, here’s my counter. Most of those thief players are mediocre (or worse). Doubling them up only shows how bad they are compared to one or both playing a more forgiving and simple build like dragonhunter. Again, why should the matchmaking algorithm care? Eventually those bad thieves will slide down in rating (and then come to the forums and whine).

The amulet system just needs to be scrapped in favor of a system which allows players to mix and match without the power creep of 4-stat amulets. For example, if you had 3-stat amulets and then split them into a choice of 3 items each at 1/3 of the original (or some other similar ratios), you’d have what you need. You could also put some degree of restriction (like only 1 of cleric stats item).

However, doing that is predicated on reigning in the HoT power creep.

Sooooo, the rating system worked?

You played teams well under your personal rating (seems like they were closer to your friend’s rating). When you won, you don’t get many points. When you lose, you lose a lot because the system expects you not to lose.

A sample size of 2 is not enough to draw a proper conclusion.

JTGuevara.9018:

Because the system is a sham, that’s why. Matchmaking is just a see-saw. ANet keeps our MMR secret for a reason, you know.

ROFL. You know that number they display on the PvP tab? That’s your rating. So much for keeping it hidden…

461 isn’t that large of a difference. And if you play together often, isn’t less rating for you and more for him (assuming you’re the higher one) more representative of the players you should face? And unless you’re a top player, does it really matter what rating you end up with? This seems more like a psychological barrier.

Many people would rather complain than identify/admit their flaws and fix them. The pile of rating complaints on this forum is a prime example. The people who move up in rating are the ones who “play to win”.

Example 1: Blue team loses a fight at 350 points, it snowballs, and red team gets the map with 200 points. Red can hold that easily to 400 points. But instead, blue goes straight for lord with the full team and kills it. Game ends at 500-300. The game was close, but the score doesn’t look that way.
Example 2: Blue team and Red team are even on Temple of the Silent Storm. At 8:30, both buffs appear. Red overcommits to bottom, loses the fight and blue gets both buffs. Blue then gets a 200 point advantage and wins 300-500. The game was close, but because of a couple mistakes at a key point, they lose by a large margin.
Example 3: Blue team clearly outclasses red team and beats them in every large fight. But blue team has a very good thief who de-caps and harasses constantly. As a result, the score ends 400-500 in favor of blue. It would seem a lot closer than it was.

I’ve read a little bit about TrueSkill. It’s compared to Elo and shown to be better. However, Elo was designed with ease of calculation over performance, since it was developed before calculators were common (1960). Glicko and Glicko2 are better than Elo, but hard to run without a computer database. TrueSkill also works better on large PUG teams and not so well on small PUG teams.

10-15 points, not percent.

I’m sure others have seen increases in the 20’s. If their rating deviation is high and they beat stronger opponents, it will happen.

1.2 Playing with a Friend
Conversely, when the lower rated friend wins, he gets a huge rating increase and when he loses, it means very little – better known as “boosting”. So there’s reward in addition to risk.
But lets look at it a different way – how often do you see groups of friends who are dramatically different in rating – more than a few hundred points? In my experience, friends who PvP often are typically similar in skill. Your example is likely an outlier.

2.1/2.2 Proposed System
Win/Loss has to be most of it. I’d say no less than 80%.
Here’s an example of how to break your system which marginalizes win/loss: I queue with a friend as a rev/thief pair. All we do is double-team gank squishies off-point. We can’t care about actually capping. But we’ll kill a few people near points anyway because that’s where we catch them. We end up having high kills, a good K:D, and picked up some offense points. But what if the team loses? How much did we actually help? In your system, what we did is equally as good as winning.

1.4 / 2.4 Group Composition
In the case of GW2, having the matchmaker try and do this is an awful idea. Please see any of the myriad threads whining for class limitations and how the same counter-argument in favor of the current system destroys them all.
Basically, players are smarter (if they choose to be) than any matchmaker which also has to deal with queue times. Players will come up with a suitable team composition before the game starts (unless they’re not playing to win).

1.* Result Consideration
My point was “how close is close”? Is 400-500 close? 300-500? Where do you draw the line? Matches which are close can snowball off one mistake and not seem close based on score. If you want a sliding scale, how much is the highest bonus vs. no bonus? It’s not as easy as you’d think due to varied secondary mechanics.
Does this really matter? In Glicko2 if you’re properly rated, you should win half your close games, so your overall rating change will be minimal. If you aren’t properly rated, you win more than half of the close games and move up.

GW2 uses Glicko2, which is an advancement of the Elo system.

If your rating is higher than an opponent, you’re expected to win. If your rating is lower, you’re expected to lose. If the expected outcome happens, the rating change is smaller. If the unexpected outcome happens, the rating changes are large. The magnitude is linked to the strength of the expectation.

Glicko adds a rating deviation and Glicko2 adds a volatility element, which account for players whose ratings are unknown or are consistently getting unexpected outcomes.

In your example, your friend got less rating gain because he was more likely to win. When you beat a team which you know is worse, it doesn’t say much about how good you are.

Nercos don’t need buffs; they need nerfs. They just need less nerfs than other OP stuff.

The longer you’re in queue, the wider the matching range you have. For players at the fringes (very top and very bottom) at off-peak hours, there may not be anyone near them playing. After they sit in queue for a while, they’re matched with the closest people who are queuing.

Current system is better than your proposal.

1.1 Rating System
ANet has never stated how they handle rating gains losses. Is it: individual vs. each individual on the opposing team, individual vs. average team, or average team vs. average team. The first option of individual vs. each individual is the best of the three, but far from perfect.

1.2 Playing with a Friend
What you described is not a problem. If your rating is that disparate, expect to need to carry. The real problem is that it makes it annoying for enemies to fight the high rated person.

2.1 / 2.2 Proposed System
Win or Loss needs to be the dominant factor for a rating system. Period. It’s an aggregate of mechanical skill, map awareness, teamwork, etc. with a clearly measurable value.
If you begin to weight other factors more, players will go after those factors instead of trying to win. That is unhealthy. Some games, like CS:GO, have used a gauge of individual skill to give a small bonus to better players, but still use win/loss at the main rating adjuster.
Note: Top stats and player skill overlap a lot.

3. Personal Experience
Are you sure that you’re better? It’s likely your bias – you’re only looking at what you’re doing well, not what you’ve done poorly. How many games have you played and when did you play them? A low number of games early in the season has less certainty to rating values. They’re starting to settle now.

1.4 / 2.4 Group Composition
Because you can swap characters after the game is created, matching based on profession (class) is meaningless. Also, each profession may not have the same build or same role. It’s not like a MOBA where a character’s abilities and role are fixed.

1.* Result Consideration
So 50-500 is clearly mis-matched. But what about 400-500 vs. 499-500? 350-500 vs. 499-500? The 350 team could have lost a FoeFire match where the other team grabbed the lord. The game also needs regroups, so you typically see shifts of 100-150 points. Where do you draw the line?

B or C. But so is everything else from HoT.

The loot boxes are meant to mirror doing stuff in WvW or PvE. They get tons of crap loot and bagception too. If you’re going to clean up PvP, clean it up everywhere.

Not rose-tinted at all.

There’s are big differences between then and now which are being overlooked and discredited. First the difference between top builds and viable build wasn’t as large. If you played a viable build well, you could beat meta builds except when played by top players. Second, current meta builds are too forgiving of mistakes (which feeds into the first item). Missing a key ability or combo would kitten down your pressure for a while. You can’t just keep hitting the next ability and be fine like now. Third, there were less hard counters to CC, projectiles, and conditions. That fact has led to an arms race where CC or conditions are too extreme.

Right, what we have in season5 is how a real rating system works. Just need to lift the solo/duo restriction and we’re fine.

A lot of these “problems” aren’t really problems and are just people not understanding the system or over-estimating their skill.

Famine.7915:

I believe majority already ruled.

The vote was slightly over 50% if I recall correctly. This solo/duo only is a trial for season5. It needs to pass a 75% threshold to stay for future seasons.

Your guildies are only partially correct. Personal skill and contribution only carries you so far. You also need to have a well-rounded team and coordinate strategy and target calling.

Know what role your build should be doing (team fights vs. solo, damage vs. support/bunker, etc). Watch the map and don’t stagger with your team, especially if you’re a team fight build. Resist the urge to be greedy for the third node when you have two secure.

Class stacking will likely never be changed – mostly because it’s not a problem; it’s a symptom of builds with a low skill threshold which are over-performing at low and medium skill levels (dragonhunter being the primary example).

At higher levels, you tend to see a more rounded set of classes anyway. Typically an ele for team fight support, a necro for team fight damage, a mesmer or thief to fight side points, a medi DH or scrapper to fight side points, etc.

Condi warrior and trap DH are popular because they perform well with a low skill threshold. Condi necro can melt teams if left alone, but are also drop quickly when focused.

S1-S4 did screw up ratings a bit, but not enough to warrant a full reset.

The point is that you have to improve and play better in order to move up.

For the umpteenth time, rating is based on who you win against or lose to. Game count and win percentage alone does not matter.

You have to beat people higher rated than you in order to move up. If you’re losing half your games to people around your rating and not beating people much higher than you, you’re at your correct rating and will not move up.

The pro smurfs have no trouble getting to platinum in a few games. Pretty sure OP would never run into them.

Only argument I see for “too easy to solo” is giving the lord group some type of condi clear. The supporting NPCs are designed around stopping ranged and sustained power dmg, but not spike condi damage.

Moonsinn.5382:

TheMountain.6204:

Exedore.6320:

sigh

Please understand the system before you make false remarks.

Win percentage does not dictate rating. Rating is based on who you played when you won or lost. People with high win percentages are either improving or were initially placed lower than they should have been. Players near 50% win percentages are at the correct rating. Players with lower win percentages are ranked too high and are moving down.

Perhaps you don’t understand this yourself.

If 2 guys can go 6-4 in their first 10 matches but end up 500+ points apart on their mmr then something is wrong. At that point the guy who ended up higher could go 3-7 and the lower guy could go 7-3 and you end up with a big difference in W/L but the higher guy is STILL much higher rating wise. So what is the real problem? It’s the initial MMR placement.

It really boils down to how lucky someone is in those first 3-10 matches. I get how the common person doesn’t understand this I just don’t get why the guy running the PvP part of this game doesn’t get it.

You do know that everyone started their placement matches with different ratings, right? They basically put everyone closer to 1200 rating (and each other) and increased the amount that the initial matches would change your rating. The 2 people with the same win/loss in the first 10 matches get to different ratings because they started with different ratings that were based on their previous matches. Seems fair to me.

In addition to this, rating depends on who you play, not just whether you win or lose. If you beat people at 1700, but lose more games to people at 2000, your rating will end up around 1800 (exaggerated examples, but that’s the idea). Similarly, if you beat people at 900, but lose to people at 1500 and 1300, your rating is going to be somewhere around 1000-1100.

Solo queue is silly. It’s a crutch for people who want to be good without working as a team.

sigh

Please understand the system before you make false remarks.

Win percentage does not dictate rating. Rating is based on who you played when you won or lost. People with high win percentages are either improving or were initially placed lower than they should have been. Players near 50% win percentages are at the correct rating. Players with lower win percentages are ranked too high and are moving down.

Before you make whine posts on the forums, you should understand how the system works.

Tier/badge is based on rating. Your rating is not absolute; it’s where you place relative to other players. When a rating system starts, everyone is in the middle and it takes time to spread out. As people toward the middle spread out, it forces people on the fringes further out. This is typically manifested by losing more games than winning if you’re on the lower end or winning more games than losing on the higher end.

If your particular case, if you’re losing more than winning and losing games by a wider margin, you’re likely over-rated compared to where you should be. If you stopped playing frequently for a few days, this is not unexpected.

It’s a problem with solo queue.

In any rating system, you’ll have a larger portion of the population near the middle rating. That middle section is inherently more volatile during the first few weeks of the rating system. If the parameters are well-tuned, it should sort itself out once people play more games.

Lol.

If this is a big fight and you’re cleaving, then chances are that your enemies are also reviving, which is counter-acting most, if not all of your cleave. Stomp ends it outright.

Sometimes it’s better to cleave or let people bleed, but it’s usually best to stomp.

Queue system for reward tracks would be nice, but seems like a waste of resources when there are bigger things to tackle.

Simpler solution is to just repeat the current one (if possible) and otherwise, go back to first in the list.

New players have to start at the midpoint for the Elo and Glicko2 algorithms to work properly. If you start new players at some other value, that new value becomes the midpoint.

New players already have maximum rating deviation and volatility (part of Glicko2).

This really just sounds like observation bias. You notice bad players more than average or good players and think they account for a larger amount of players than they actually do. In reality, I would guess that many new players are actually average and starting them at the midpoint is just fine.

Revenant is essentially the opposite of Dragonhunter.

Revenant has a very high skill floor but high ceiling. In the hands of most players, you’ll be destroyed. However, once you’ve master it, you can do exceptionally well and compete with other skilled opponents on equal footing.

Dragonhunter has a very low skill floor, but also a low ceiling. You can get the build from metabattle, faceroll it, and beat most people at low and medium skill levels. But at higher skill levels, the playstyle of DH becomes too narrow and predictable, allowing it to be easily countered.

Revenant needed the nerfs and probably needs more down the road. It only seems weak because other elite specs are too strong and haven’t yet been nerfed.

JTGuevara.9018:

We actually used to have them. This was back on 2012. I participated in a few. This was back when you had both solo and team q.

Pretty sure that’s incorrect. As I recall, the automated tournaments were removed at the same time they added solo queue and team queue.

Saiyan.1704:

Know of any reason as to why they removed them?

Shortly after the game first released, there were three automated PvP options:

• Hotjoin (might have gone by a different name) 8v8 maximum, but otherwise it’s the same as today. 5v5 was added later and 8v8 phased out.
• Free Tournament 8 team best-of-1 single elimination with teams of 5 players. You could queue as any size and full team rosters were assembled by a simplistic matchmaker. If your team won a round, it stayed together for the next round (no substitutions). Rewards increased the higher you got. You could queue at any time of the day and the tournament started as soon as it had 8 full teams.
• Paid Tournament Same format as Free Tournament, but required a ticket to enter. Tickets were purchased with gems or given as rewards for winning free tournaments. Rewards were considerably large for that point in GW2’s history.

Paid tournament was removed very quickly because good players could farm lots of rewards while lesser players felt robbed of gems. The skin rewards were folded into free tournament loot boxes.

There wasn’t a clear reason(s) given for removal of free tournament. However, I’m guessing it had to do with blow-out matches in earlier rounds and queue times (have to wait for 40 people instead of 10).

The solo queue and team queue system initially used the same matchmaking and rating algorithm as Season5, but some matchmaking parameters had different weights. Team queue was any size. There was a leaderboard for each queue. The leaderboards were based on rating with a primitive decay mechanic.

In a later “test” season, the leaderboard was changed to number of wins. This was presumably done to counteract people who were 10-1 being at the top of the leaderboard (you didn’t display until you played an 11th game). The win-based leaderboards got the nickname of “farmboards” and no one cared about them anymore. That’s likely why they were removed going into season1.

I’d like to see automated tournaments return, but in a more refined manner:

• Include more teams.
• More matches per round, at least for later rounds.
• Limited availability – set day and time to enter.
• Restrict based on rank/rating, such as “gold and below” on one day and “platinum” on another.

I’d like to see that essay against solo queue only. Probably for another thread though.

The point of rating to decay is to stop people from being able to sit at the top of the leaderboard until the end of the season. It’s not intended to hurt anyone else.

The idea behind it is that the people at the top have to play every few days. If they got there with luck or binging early in the season (when rating is more volatile), later games may hand them a loss and knock them down the leaderboard. It also is intended to address the related problem of “almost top” stagnation where there isn’t enough rating to boost players higher later in the season because the higher players have hoarded it.

Playing or not playing in past seasons has no impact on rating. I played one game in S3 (to get stacked rewards) and didn’t play at all in S4. I went 6-4 in placement and ended up in gold tier3 and moved up to platinum.

Pache.9406:

The solution to this problem, is the opposite of your tittle. Core specs should be brought up to par with elite spec, not elite spec brought down to core specs level (like the tittle suggest).

No, you are wrong and OP is correct.

The problem with elite specs is not only that they’re too powerful compared to core, but more that they’re too forgiving and do too many things well. The only way to fix that is to nerf elite specs and the associated power creep.

No. The point of the ranked system change was to remove the double-restriction of pips and rating which were competing against each other.

Restricting matches by both pips and rating caused the pool of eligible matches to be tiny. This led to long queue times, and with longer time in queue, the rating differential increased, which led to lop-sided matches.

I would say that goal was accomplished. Queue times are much shorter and blow-out matches are much less common.

Tiefsee.3647:

I agree, but isn’t the main complaining about class stacking?

Whining – not complaining. People are whining about the rating system and placement matches mostly because they think they’re better than they actually are.

Duo queue is also the topic of many whines.

The penalties for leaving or DCing are supposed to get worse the more frequently you do it. But it had a few bugs each time ANet tried to fix it, so I think they just gave up and stuck with a flat penalty.

Threather.9354:

Is it worth? I guess ill be placed around 1500s due to only playing up til legend on seasons. I see considerably less qq than last seasons.

Your rating will not align with your rank from past seasons, especially from season4. Don’t be surprised if you’re lower.

The matching based solely on rating is much better than the awful pip-based matching from the last four. Games are closer and queues are 1-2min long.

How long does it take to take ascended armor/weapon if you play like 10 hours a day?

You can probably reach the max tokens for ascended items within a week. The rewards used to buy ascended items are from the pip track are predominately based on time, but playing better gets them faster. Because you’ll be playing people near your rating, you should win about 50% of your games. Win gives 10 pips, loss gives 3.

Is power rev viable? Ele still thrash for solo I guess? I can pretty much play all classes except thief.. I just need to choose my flavour of the month.

Power Rev works, but you have to be good at it or you’ll be farmed. Ele is good for solo, but depends on your team to know to fight with you and not stagger in solo. I would say DH and warrior give you the most flexibility between 1v1 and team fights.

Did you ask one of thieves to switch at the start?
Did you tell the thieves that their role is de-capping or picking off low HP enemies?

Unfortunately not every match is great and not every game can be carried. Hopefully in a week or two, player ratings will be better aligned; S1-S4 probably messed them up quite a bit.

akaCryptic.2389:

That’s all good but ranked is supposed to represent skill while the social aspect is only secondary. Voice comm is esentially a huge advantage.

Playing as a team is part of skill. Again, for most skill levels voice comm is not the massive advantage people make it out to be.

One solution that wouldn’t kitten off both sides would be to have the same number of duo Qs on each team. So 2 2 1 vs 2 2 1 would be fine whereas 2 1 1 1 vs 1 1 1 1 1 is not fair.

I’ve won games with more solo queued players on my side. You’re just complaining to complain or to try and legitimize your losses.

Hesacon.8735:

The Bad:

• Placement matches are weighed far too heavily, I went 3-7 in the first 10 matches because I was testing out new builds trying to get one I liked after abandoning spvp due to queues and other issues. I started with a rating around 900. I have won 8 or my last 10 and got it up to 1000. People on my friends/guilds list (see attached) have similar or lower win/loss rates but 500 more rating.

So the system put you at a low rating because you were playing poorly during placement? And now that you’re playing better, you’re moving up? Sounds like it’s working just fine!

Remember, win/loss ratio by itself only means if you should be moving up or down. Who you win or lose against determines what your rating should be. You’re winning against lower rated opponents than your friends.

1a. Not exactly. Top stats are too easy to game and a single one isn’t reflective of good performance. Having multiple top stats or a high percentile in them all is more realistic.

1b. You can just give raw points; it breaks the Glicko2 system, which has to be zero-sum. Read up on the CS:GO MVP system. It’s a workable solution for recognizing a player much stronger than their team.

2. This is pure opinion. Placement matches work for the most part and if you’re poorly placed (seems that DCs are causing a lot of it), you should quickly move up or down to your proper rating.

3. See 1a and 1b.

4. Nothing is wrong with bunker or condi in theory. The problem is that power creep has taken these to the extreme. Condition builds apply multiple types and numerous stacks far too quickly. Bunker builds are generally too forgiving either from passives or abilities just being too strong.

Your tier from last season means absolutely nothing in this season because S4 tier was not tied to rating. You could have been a 1000 rating player in legendary last season and a 2000 in Sapphire.

Placement is based on the ratings of opponents you played and the outcome (win or loss) more than anything.

The argument for solo queue is basically this: “I want to win, but I don’t want to coordinate or use teamwork in a team game.” If you don’t play to win, then don’t cry about losing to someone who does.

• If your team comp isn’t good, then switch or ask someone else to switch! Group queues have players who already bit the bullet and played something to help the team even if it wasn’t their first choice.
• Know your role. If you’re not good at solo fights, don’t run out solo. If you’re a support, stay glued to someone who needs it. If you’re a thief, don’t try to take every 1v1; only take the ones you know you can win. If you’re not fast out of the gate with good sustain, don’t push far alone at the start.
• Focus/assist, chat, map pings, and basic map awareness are all tools you can use. If your team is staggered, call it out. Call incoming with numbers or ping before the fight starts. Coordinating before the match starts helps a ton. I rarely see PUG teams do any of this. In my experience, voice comm doesn’t improve this for most group queues (coordination in-game is not strong enough); they’re just not shy about communicating.

A rating system has to ultimately be driven by winning and losing. It’s by far the best measure of skill over time.

However, many other games have supplemental bonuses for high performing team members which use things like this. Adding a few points to a win or cushioning a loss for a “carry” team member would be okay.

No matchmaking system is perfect. Every so often, you’ll have lop-sided matches.

StickerHappy.8052:

So they say The placement matches will place you on your appropriate MMR? How does the system decide on that if all the matches and teammates you have are utter trash..? So if we lose the game is that on me?

This highlights a problem with using pure Elo/Glicko for an individual player in a system with constantly shifting teams. Many games use a modifier to try and account for individual contribution to help above average players from being weighed down by significantly weaker teammates.

The main arguments against this kind of compensation are: 1. the method of determining it is a grey area and 2. eventually you’ll stabilize while the weaker players go down – i.e. you can’t be unlucky forever.

The argument against team queues in competitive is predicated on a random solo team of keyboard turners being matched against a full ESL team. It’s simply ludicrous.

Do honestly think the vast majority of teams can call targets, coordinate burst, rotate appropriately, track respawns, etc.? The main advantage for most teams in queue is a well-rounded team composition. And that shouldn’t be such an advantage because any solo queue team has ample time to achieve a well-rounded comp quickly with minimal communication. Most solo queue players simply don’t put in the effort. And for that, they should lose and have lower ratings.

The biggest problem from past seasons with team vs solo queuing in the combined competitive queue was the awful pip system which restricted the matchmaking pool. Teams and solo players were both sitting in 7-10min queues and at that point, rating isn’t going to be anywhere close when a match does occur. THAT is why you saw so many blow-out matches in Season1 thru Season4.