Guild Wars 2

Madara.7435:

I think its a very good idea to poll this and not just dump it on people. Thank you for that!
Question is, if we have our classes locked will matchmaking force tournament rules on us, I mean 1 class per team only? I would love to try this. yes qtime will most likely increase, but Id rather wait 2-3 minutes more in q than having 10 minutes of a frustrating match with 5 thieves/guards. unranked could still stay the way it is now.

It’s hard to say what will happen to queue times if we enforce a hard limit on professions. The compromise is to let the profession limit increase over time.

Supreme.3164:

Evan Lesh.3295:

Supreme.3164:

You should stop people from changing class after pressing “accept” on the queue

This is what the poll is for.

You said that it would not be possible to change class once the 10s countdown starts, but before that there is a whole min where people can change build/class freely and I’ve thought the poll was used to decide whether or not people should be allowed to change class during that minute.

You are correct. Once the final 10 second countdown starts, there is no reason to swap characters, so we’ll lock that down to prevent exploits. The 90 second warmup is the period during which characters could additionally be locked if the poll succeeds.

Kir Sakar.3647:

Do I get it correcty that with the “YES” option, there is no way for me as a player to adapt to team composition? Say, i see we have DPS with no sustain before the match starts, there is no way for me to swap to support, even if this swap would not cause class stacking?

Then I find it a bad idea, it takes away an important aspect of the game and makes Solo/DuoQ even more a lottery because class is not identical with role.

You are correct, but there are trade-offs to both options. This is why we created a poll! If we lock characters on queue, we can do things like use profession-specific MMR, have profession leaderboards, etc.

Supreme.3164:

You should stop people from changing class after pressing “accept” on the queue

This is what the poll is for.

bsandro.7428:

Does it mean that class swithing during the match will be unavailable in both cases?

Yes. We will disallow swapping characters once the 10 second countdown begins. This prevents things like playing 95% of a match and then swapping for profession achievements.

If point gains were arbitrary as you describe, then there would be no upper bound on MMR, and you would see people who play a lot with 10k rating at the top of the leaderboard. This is not accurate and just takes us back to seasons 1-4.

The reason you are gaining very little is because you are expected to win all those matches and your rating is already accurate. When you lose a match you are expected to win, you take a decent hit. Your history shows that you can easily maintain your rating, which is good. If you’re already one of the best players, why do you need your rating to continually rise?

I can understand the argument that high-skilled players might want some sense of progression past being number 1, but that is outside the scope of MMR. This is one of the reasons you consistently get more rewards while highly rated.

It averages your ratings. In previous iterations of the league system, a party would be match-made with the highest badge tier.

Sounds like you lost connection to one of our servers. Your friends list and guilds might have also been broken. Your only option is to wait or try relogging.

Rating deviation will no longer increase in future season to stop situations like this.

Victories after gaining desertion status will be granted assuming you are still connected at the end of the match after the next update.

Wargameur.6950:

Khorg.2195:

So….regarding matchmaking/teams… There’s multiple people posting threads on pvp teams in ranked.. yet you guys address everything but that. Can you at least acknowledge that a lot of your player base is upset about this?!?!

It has been addressed already. They recognized there was some flaws in matchmaking and they would change some things during s6.

Evan, I think you said our rating would reflect our skill around ~200 games ( If it’s wrong tell me ). Do you know when the majority of the population reached that threshold ? I mean, in a perfect world, everyone would place themselves in the first quarter of the season then enjoy the game at their level of skill for the rest of the season.

There are changes to matchmaking coming for off-season. I’ve updated the wiki with the changes.

I just checked and it takes on average 43 (+/- 10) games to hit minimum deviation.

alemfi.5107:

Chase.8415:

Evan Lesh.3295:

I actually just updated the matchmaking wiki to be up-to-date if you’re interested, but I think your question has more to do with rating accuracy than matchmaking. Time of day definitely has an impact as well as day of the week. If we had an infinite population always playing, ratings would be wicked accurate.

https://wiki.guildwars2.com/wiki/PvP_Matchmaking_Algorithm

If it makes you feel any better, I’ve also been getting dumpstered this last week. Maybe more people are trying harder towards the end of the season.

Is that true where you get a separate rating for each profession? It would be awesome having a unique rating for each class because I feel stuck in rolling just one class sometimes.

Also, can we get the percentiles of each rating bracket?

Not Evan (obviously), but from what I know, it is tracked, but not used. Tracked so that in possible future adjustments/implementations of the match making, they can use their existing data. (Like queuing with a select pool of characters, or being able to queue from char selection screen etc.)

This is correct. We have them, but do not use them.

I actually just updated the matchmaking wiki to be up-to-date if you’re interested, but I think your question has more to do with rating accuracy than matchmaking. Time of day definitely has an impact as well as day of the week. If we had an infinite population always playing, ratings would be wicked accurate.

https://wiki.guildwars2.com/wiki/PvP_Matchmaking_Algorithm

If it makes you feel any better, I’ve also been getting dumpstered this last week. Maybe more people are trying harder towards the end of the season.

BeepBoopBop.5403:

has the leaderboard been updated?

It will update when decay happens on daily reset.

Spartacus.3192:

Evan Lesh.3295:

Spartacus.3192:

I really hope min games for next season will be at least 100 matches.

but thanks for doing this now.

We were thinking something like 10 per week, but it is configurable. We could to 1 a day, 20 per 2 weeks, etc.

The only problem with 10 per week is if people going on vacations for a week and failing the requirement,

So it could be every 2 weeks at least with a minimum total amount of at least 100 please.

The cutoff works just like dishonor in that if you come back and play those 10 games, you’ll go right back on the leaderboard. If you go on vacation a week before the season ends, well, then decay will remove you anyways.

Spartacus.3192:

I really hope min games for next season will be at least 100 matches.

but thanks for doing this now.

We were thinking something like 10 per week, but it is configurable. We could to 1 a day, 20 per 2 weeks, etc.

Bio Flame.4276:

Evan Lesh.3295:

Cynz.9437:

Evan Lesh.3295:

Found a couple minor bugs in addition to the lack of reshuffling. I’ve also worked on a new version of the matchmaker that will improve roster size and profession matching. We will likely trial it during the off-season.

If I remember, I will update the wiki!

So we have to endure the bugs and flaws of the current system (that certain players abuse) until next season? Lovely… :/

Perhaps you’d understand if you researched a bit on game development. Should be tons of resources if you feel like googling; things like quality assurance, build processes, release management, etc.

Since your reply was so disrespectful and pratronizing, I’ll answer on the same tone:

Perhaps you’re understand if you researched yourself on PvP-centric games and their balance schedule, such as GW1, LOL, DOTA2, Overwatch, etc…
There’s tons of resources if you feel like googling or better yet, just ask ANET’s Devs how they managed to do fast updates on GW1.

Get off your high horse, it’s not helping you, nor the players, nor the game, nor your job…

Didn’t mean to be disrespectful! I also don’t have access to a horse. I just didn’t want to try describing our entire internal build process to explain why a change won’t get to live as quickly as players may want. The bugs I mentioned for the matchmaker are so minor that they don’t warrant a hot fix. The new matchmaker is not worth risking game stability in a hotfix either. The earliest build to ship these on with adequate testing that doesn’t throw off the rest of the studio will be the off-season build.

There are a whole lot of other teams shipping things here that are part of the flow too, and its a fun dance to make sure we’re all getting things shipped with a comfortable cadence. If PvP was its own standalone product with no shared resources like the games you mention, then shipping changes quickly might be easier.

My plan is to set the respawn time to default (15 seconds) plus the maximum amount of bleed time they could have suffered on disconnect. This should be the most fair for everyone. If you accidentally d/c, then you want to get back in as soon as possible.

If there is respawn time left on reconnect, the player will be killed.

Yep, we’re tracking this.

Cynz.9437:

Evan Lesh.3295:

Found a couple minor bugs in addition to the lack of reshuffling. I’ve also worked on a new version of the matchmaker that will improve roster size and profession matching. We will likely trial it during the off-season.

If I remember, I will update the wiki!

So we have to endure the bugs and flaws of the current system (that certain players abuse) until next season? Lovely… :/

Perhaps you’d understand if you researched a bit on game development. Should be tons of resources if you feel like googling; things like quality assurance, build processes, release management, etc.

Looking into it.

Crinn.7864:

Evan Lesh.3295:

This happens after taking breaks, but doesn’t really make sense for season play. We are looking to change this behavior for season 6.

How long can breaks be before they start mucking with my rating?

I typically don’t do ranked for a couple days then binge queue. Are those couple day breaks really making that huge of a difference?

It gets more volatile the longer you don’t play, so it depends.

This happens after taking breaks, but doesn’t really make sense for season play. We are looking to change this behavior for season 6.

Fifteen forum bugs?

Sins.4782:

Nice. Does this mean profession specific MMRs will actually be used then? Or are the changes just to class stacking/shuffling?

Nah, we can’t use profession MMR without locking them after queuing. It can better handle mirroring comps across teams.

Found a couple minor bugs in addition to the lack of reshuffling. I’ve also worked on a new version of the matchmaker that will improve roster size and profession matching. We will likely trial it during the off-season.

If I remember, I will update the wiki!

FWIW, I’m looking into other possible stacking issues after dissecting the crap out of the substitutions problem.

MOAR DATA: ~3% of matches have substitutes which could* cause the original problem in this thread. We are testing fixes now.

I have good news and bad news.

This example is an extreme case so it is not screwing over everybody. There were multiple people that failed to ready up, which meant multiple rounds of substitutes were needed.

Bad news. Teams are not rebalanced after finding substitutes and it has been that way for years. There are complicated computer-science reasons why, but it is solvable and I will attempt to fix it. In this case since it happened multiple times, it compounded the inaccuracies.

Barry.3147:

I’m definitely going to be looking up the match. I need to see what the ‘real’ before/after ratings are without things like decay and make sure those are the right players.

10 players are collected before making teams.
You are looking at ratings after the match which are going to be further apart than the ratings that were used to make the match.

The two duo queues on one team is annoying, and something could be done to improve that.

And as always (though slightly out of date) the matchmaking process has been documented here:
https://wiki.guildwars2.com/wiki/PvP_Matchmaking_Algorithm

Also all those games are on thief, maybe you could switch up your play a bit and then come back to thief.

Of your 132 post placement matches, you peaked at 1413, bottomed at 1140, and have a win/loss of 63/69. Your longest losing streak is 9 games after a dense set of wins. I’m not seeing anything wrong with this. Perhaps you’re just having an off day.

I understand the hyperbole because you’re likely frustrated, but it doesn’t help when discussing feedback for leagues.

Leaderboard is missing byes, forfeits, and desertions right now.

Someone needs to make a Glicko test program that uses rules from the wiki and post it online…

Preview /15chars

Our amazing UI artist has already made a pass on these and they are much more distinguishable and bright.

jcbroe.4329:

Evan Lesh.3295:

Pyriall.1683:

Evan Lesh.3295:

Pyriall.1683:

If you are using MMR to pair teams, why are 1,4,2,3 even being teamed with 8,9,10’s? That’s why this system continues to frustrate people.

Because it’s a 10 player game. Those 10 players could have the exact same rating in the best case scenario.

Am I misunderstanding what you are saying. 1-10 denotes skill or simply the number of players?

Just ordering of skill. The skill range could be big or small. (hopefully small)

Hey Evan,

Just wanted to take a quick moment to ask that if in the future, you would consider implementing a “hard cap” on initial tier placements (can’t place higher than gold3, plat1, etc).

While it’s true I don’t know your exact algorithm and 10 placement matches is a generally accepted industry standard for a multitude of good reasons, as a player, judging from my own experience and observations as well as some others; the perception is that the placement matches have over or under-inflated peoples placements.

With a cap, it just seems like you could still place people within proper thresholds accurately, but then people wouldn’t be able to play the “place and beat the decay” game, at least, not if they wanted those higher tiers, which would “force” people into participating in more matches than a lot currently have (granted we’re only 2-3 days in).

This would help the system gather more statistics and place people more accurately and with a higher indication of their actual skill tier than the way it appears currently, not to mention there being a larger pool of players to create matches from since people would still have the incentive to play for more than just beating MMR decay (I hear reports from friends in higher tiers of games being scarce, and I assume they’d also be less evenly matched since the longer the matchmaker works, the more the rating threshold opens up to try to create a match).

Granted I didn’t really need to explain it all, but I did so in case you do actually decide to respond and do one of those “separate perception from reality” things that you do.

We talked about capping placement and opted to have less restrictions. One thing I would consider is lowering the default rating deviation so people don’t fly around as much and soft reset has more meaning.

Ah yeah, spose I can expose that now since it’s not a secret anymore.

Interpret Interrupt.3824:

Evan Lesh.3295:

Honest John.4673:

I have the feeling Evan has a lot of job security right about now. This is a good thing, he’s like the anti-Gaile Gray

I’m excited to finally have MMR out in the open. We can get some great community feedback like what professions dominate which skill ratings to better handle player perception. We have metrics like this internally, but its super tedious to look up the skill rating of every player that gives balance feedback on the forums.

Now if only skill raiting showed up when you posted on the pvp forums, you could weed out the actual threads pretty quickly

It’s not that things need weeding. It’s just good that players have rating available as a data point when making suggestions and giving feedback. Something like:
“I have sub 1k rating and hate playing against X”
“Well I have over 2k rating and X is terrible”
This is way more helpful to us!

Honest John.4673:

Malediktus.3740:

Evan Lesh.3295:

Soft reset: (Previous Rating + 1200) * 0.5
However, everyone’s deviation got reset, so your still move up and down a lot from this new rating.

Why even include the old rating? Since higher MMR was detrimental (higher queue times, harder to earn pips) a lot of people did not care if MMR was low.

So that the esl players would have less of a chance to match up against people who bought the game last month. It would have been much more chaotic with that approach.

Yes. We’ve done a full reset before and it was awful. Keeping everyone’s relative position means matchmaking isn’t completely random at the start of the season.

That’s some interesting maths. We hadn’t considered queue time when balancing the legendary pips as it was just kind of icing on the cake. Will see what queue times become after a week or two.

Ensign.2189:

Polismassa.6740:

2) Your rating is not a pure measure of your skill MMR in a team based game is, and always will be a measure of your ability to win matches.

Just to build upon this:

There are a lot of different facets to ‘ability to win matches’. Players can be good at winning duels or winning team fights or picking winning fights or fighting on the right points or rotating between points or playing good builds for the meta or swapping to the right build for the match-up or filling in effectively to balance your team composition or playing hard every game or not tilting or encouraging teammates and keeping people focused.

Or a whole lot of things I didn’t mention in that run-on sentence above.

The difference between a pro-level player and an experienced but middling journeyman (say, gold level) player isn’t that the pro is a bit better at all of these things. It’s almost always the case that the pro is good at almost all of these things, while your middling, experienced player is good (sometimes even pro-level!) at some of these things, and just a dumpster fire at others.

Good but not great players include people that can beat everyone they face in a 1v1 but are constantly engaged in bad fights (trying to kill a guardian on a fully capped point with their thief), people who can’t win a duel for the life of them but pick teamfights they can contribute to well, people who know a single trick (condi mesmer that always goes far!) that sometimes dominates and sometimes feeds, people who often play well but tilt and throw as soon as something goes badly (thus losing games that could be recovered and won).

The really insidious part of this? If you are one of those players, you’re probably really focused on the parts of your game that you are good at, and blind to those that are holding you back. You’re also blind to the good things your teammates are doing (that you are not good at), while their flaws (in areas that are your strengths) are on full display.

Which is why you think your teammate (with the same MMR and similar match win performance) is terrible, and why that teammate thinks that you are terrible. It’s why there’s a pretty big range of play (generally the 75th to the 95th percentile, roughly) that is just an utterly miserable rage-fast, a bunch of skilled but fundamentally flawed players picking at each other’s faults.

If you want to get better, don’t look at that ‘scrub’ with the same rating as you in disbelief. Figure out what he’s doing that you aren’t – because, odds are, that’s what’s holding you back.

slow clap

Soft reset: (Previous Rating + 1200) * 0.5
However, everyone’s deviation got reset, so your still move up and down a lot from this new rating.

Honest John.4673:

I have the feeling Evan has a lot of job security right about now. This is a good thing, he’s like the anti-Gaile Gray

I’m excited to finally have MMR out in the open. We can get some great community feedback like what professions dominate which skill ratings to better handle player perception. We have metrics like this internally, but its super tedious to look up the skill rating of every player that gives balance feedback on the forums.

The matchmaker just prefers to put doodoo heads with doodoo heads. It doesn’t change the skill level of doodoo heads you’re playing against.

It’s almost like DH has a low skill floor or something.

If you disconnect from a game, the result will count as a desertion and be calculated as a loss. Getting dishonor in general (like from missing ready-up) doesn’t do anything like subtract from your MMR.

Get 4 of the final tier pip chests.

Pyriall.1683:

Evan Lesh.3295:

Pyriall.1683:

If you are using MMR to pair teams, why are 1,4,2,3 even being teamed with 8,9,10’s? That’s why this system continues to frustrate people.

Because it’s a 10 player game. Those 10 players could have the exact same rating in the best case scenario.

Am I misunderstanding what you are saying. 1-10 denotes skill or simply the number of players?

Just ordering of skill. The skill range could be big or small. (hopefully small)