Guild Wars 2

You thought mesmers were stronger in PvP than Revenant or Scrapper or Reaper? Seriously?

I get it. I just wanted to say the matchmaker wasn’t necessarily predicting 500-50 at the start. People make it that way. And for very understandable reasons, I think.

What seems to be happening is: many people recognize a likely loss in the first several team fights. So they quit wasting their time and hang out and let the score rocket to 500 so they can get on to the next match and see if they have a winning chance.

So a game that would have been 500 – 295 becomes 500-50. Which makes the matchmaking appear to be much more lopsided than it actually was.

The matcher creates losses, but people who have 300 losses under their belt see the pattern quickly. And stop fighting.

Somehow it seems less humiliating to take the loss on purpose, than to work as hard as you can and STILL get crushed.

Players have implemented their own RESIGN button.

Shadowflare, I was responding to the original posters comment that unranked is not a reasonable alternative to solo queue because you cannot progress your league standing in unranked. But I said I wasn’t anywhere close to being pro in gw2 PvP, so how’s that patting myself on the back?

"

I’ve decided that League standing tells me pretty much nothing I don’t already know about my ability. Or lack of it in my case. I just don’t have the twitch reaction and I don’t know the classes well enough. So I am never going to get really good at this. Even if I got lucky and rode the wave up to a higher Division, I still know how good I am, or am not, at this.

It’s okay, I’m really, really good at other things in life that matter to me. I don’t need to worry about if I am good in gw2 PvP.

I find that I enjoy unranked much more than Ranked at the moment. Isn’t that pretty much SoloQ?

We talk about people’s MMR.

I think you can only capture a range of what it might be, with the middle the highest confidence. YES, there is the R number that captures uncertainty.

With more games the range supposedly gets less fuzzy and more like a solid line at the middle.

I don’t think that can really happen accurately in the current system. Why? Because of the current low player population and the need to put matches together in a reasonable wait time.

I was just describing the actual reality, not suggesting a solution.

I am nearly certain there is no budget for any kind of big rewrite of the matchmaking programming. I could be wrong, but I don’t see evidence of a lot of resources going into PvP. And the names keep changing in the past two years. Justin O’ Dell was pretty good, and talked to us on the forum. Evan seems like a good guy. But I sense he may be the only one on PvP duty.

Ponidus,

There is a HUGE amount of noise in the data.

Yes, Team A won. But why?

There is the mix of classes in the match, and it matters how they work to effect a win, or a loss. Class interactions matter.

There are decisions made by each team member that affect the win or loss. Some can be disastrous based on the situation when it was made.

Yet the MMR is adjusted after a single match. A single case. A single data point.

If you had the matched teams play each other 20 matches in a row, you would not get the same number of wins per team! Each player has a “range of skill” and they do not always play their best. In one set of 20 games Team A might win 15. And another series with the same players, Team A might win 10.

That’s a reason why the World Series is seven games; not one.

Yet the MMRs are all adjusted based on ONE match of the two teams.

So, just as in a double blind controlled case study, you need LOTS of data points. n=100,000 is good. n=10 is useless. Because of the confounding variables and the noise that you need to sift out over many, many cases (i.e. matches).

The actual Accurate MMR values will not emerge from game data very quickly. And I think that is the MMR problem.

There is a low PvP population, and they need to throw together matches in a few minutes. They get bad data from such matches about individual player skill. And then they use that bad data to generate MMR adjustments.

I have been getting them too, & I have a fast drive and a hot network connection.

insanemaniac.2456:

personal score and top stats are too narrowly defined to singly indicate a carrier, and the rating system would have to have intelligence behind it to pick out a carrier from an aggregate of each stat and score.

Could be. Especially in light of the things Yasi talked about

Yasi,

Good observations. And yeah, 4000 matches and I still run right out of base! How do people learn this stuff? Probably from watching twitch, right?

But clearly there’s a problem because people’s MMR is putting them among people that they have to carry. If MMR was accurate, they wouldn’t be with people that had to carry.

Since your MMR is determined from the teams you are placed on, it’s kind of circular.

Crinn.7864:

The first problem with this entire thread is the idea that “carriers” even exist.
The concept of whether you’re “carrying” is entirely based on the context of the match.
A average player on a team of rabbit ranks vs some slightly less than average players is obviously going to be “carrying”
That same player on a team of proleague players is no longer “carrying” and is in fact “getting carried”

The second major issue is what being a “carrier” really means. Carriers are if you boil it down, just players who are better at winning than other players. Which means if you want to find carriers you want a system that can evaluate how good a player is at winning, which leads to the glicko and ELO systems that already exist.

Carriers are good at making the TEAM win, that they are assigned to. They make enough of a difference somehow to erase the skill difference between the teams, even if their MMR is similar to their teammates.

But aren’t we at the tail end of the INITIAL STARTING PERIOD?

I mean we are 13 days into a 60 day season.

It might be as simple as carriers win all their fights and are the only person on the team who tends to do so at a high degree. They take very high damage on a non-bunker class but never die once. Or something like that.

I am not so sure that it is so decision-oriented that patterns would not emerge from seeing the data from games that had a carrier. Decisions have consequences, and the consequences should show up in the data in some way. And when you see the pattern, it might not make immediately obvious sense to you what was happening in the game that generated that data pattern. But it would be there.

This is what machine-based learning does today for many things as they track what they call “big data”.

Well I know how I would do it if I were in charge. Based on my background computer science, I would back gather hundreds of matches where there was an known carrier. And I would use a machine learning algorithm to find patterns that are not immediately obvious.

Sometimes these patterns don’t make obvious sense, but you will find them always with the carrier and so you would look for people with that sort of pattern.

As a start, maybe a carrier has few deaths, even with lots of damage to himself, and a larger number amount of caps.

This is tough so I would expect a lot of iterations and discussion before we could get to the heart of what a carrier would look like strictly on statistics from a battle.

Mightybird.6034:

Oh no a stranger has a negative opinion about something I drew, MOM MOM! MOM! better post about it on the internet MOM! someone said they didn’t like me!

Nah. It’s not about me. It’s about the idea the slot machine represents and his rejection of that idea.

Windwalker,

Yes, I DID see myself in some of the things he called out in his video. And I, too, have a lot of matches, 4700+, and yet I realize I played most of those the same way, with the same lack of deeper understanding.

So I will watch his next video to learn more.

I think ANet has the right idea in moving people up and out of division faster; they just urgently need a way to identify decent players caught with an artificially low MMR and up their MMR as well and move them further up the divisions as well.

antichecker.1675:

To the topic, the image doesn’t look like a high quality edit, still I think it does reflect the current matchmaking system.

No problem. I did it in Photoshop in about 30 minutes, just as a powerful illustration of the idea. If I was more knowledgeable in Photoshop techniques and shortcuts, I could have done it in 10 minutes.

Sirbeaumerdier,

Good point. Is it also possible that people seeing him on the opposing team start to focus on him, thus creating favorable numeric matchups for his team in other fights across the map?

Question: is there something the matchmaker can learn from combat statistics and personal score that it could use to tweak upward the MMR of the people stuck in low-MMR teams?

They would have to keep this secret to avoid people stacking their stats a certain way.

What kinds of stats in a match identify the carriers from the carried?

So I made this graphic of a “GW2 slot machine”. It illustrates that the matchmaker will quickly throw together a match between two teams who have wildly different playing skill.

As expected, the higher skill team usually wins, and the lower skill team typically loses.

Why a slot machine? Because it is RANDOM as to how close the teams will match, on skill. The matchmaker does not try to make the match close in skill; it mainly wants to just move all the better players up and out of the division.

Yet Helseth presents my slot machine picture, “Now I want you to look at this. Look at this! This is the noobiest thing ever. This is the noobiest thing I have ever seen!”, he shouts.

Because to him, it’s fine to have unequal matchups, you just have to carry your team. And if you cannot carry then you are trash.

Really?

I watched his video and I did learn some things, and I did implement them using my scrapper. And it helped some, but not enough to make my teammates rotate better or win their fights when I was not there with my Scrapper.

I would stare at the score 2 minutes into the game and other than myself and one other person, my team would have 0 points, while the other team was a list of 80 and 100 points for every player.

There is a limit to how much you can carry.

I really want to watch his next, expanded video he promised, to learn more about carrying.

But there seems to be a limit to what you can carry.

Not to mention at the developers conference this week ANet said they were “well aware that there are some frustrating matchups” and they hope to do something to alleviate that in Season 3.

As usual, the truth lies somewhere between the extremes of “everything is fine, it’s you that sucks” and “I am fine, it’s the matchmaking that sucks”.

How do you do it if your teammates lose every fight?

The OP was a very good and provocative question that really gets into understanding the situation and the complaints about sPvP.

Having Diamond Bracket players play against lower bracket players would not be a valid matchup. Brackets are brackets.

It would not be fair to have some Diamond Bracket players win by playing less-challenging, non-Diamond opponents.

How about making people pass a test before being allowed into ladder play, showing they know how to play Conquest maps?

Something like this could be cheated on unless you change the questions daily.

But it would be interesting to have an online test of multiple-choice questions dealing with strategic thinking, rotations, MAP READING, specific map strategies.

—-You could have people from the PvP community submit questions.

Things from the recent “basics for new players” thread would be asked.

You would not be allowed to queue unless you passed the test.

I bet the quality of play would skyrocket!

I think people who come from PvE or soloing mobs in MMOs fail to grasp team PvP concepts.

Sirbeaumerdier,

You make an important point. The MMR is a TEAM rating. It is your rating of how well teams you are on tend to do. But since the team makeups vary so much, it could take 1,000 matches of all kinds of teammates and compositions before it could know what your “team play MMR” really is. Not to mention if you play different classes as well!

I think the MMR is not very precise in such a fluid environment with so many kinds of teammates.

People want to know their MMR. In truth it is something like “somewhere between low and a little above average”, or “a 60% chance that is it rather high, but 40% chance you got carried”. Something like that.

It is certainly NOT 1252, or 957. It’s a range, not a precise number. Maybe in 1,000 matches it might be more like an actual, precise number.

It’s like, what’s your Raid group rating?

Great thread. Thanks.

I see people posting that they win and then “get stuck” in a division where they are always placed a a team that loses.

I sounds like when you arrive at your “proper” division, you become the “conveyor belt” to push higher MMRs on towards their own “final destination” division.

As such, your team will lose constantly until there are no more “higher MMRs” to ram through your division. And eventually you will have 50/50 chance matches within your division.

If anyone other than Legendary Division is still playing Ranked, that is.

Pyriall,

Every blowout win for you team is a blowout loss for the opposing team.

The current matchmaker tries to move higher MMRs out of a division quickly on a “win streak conveyor belt.” Which is unfortunately a blowout conveyor belt for the lower MMR team.

Honestly now, (I mean, honestly) in how many of your wins in Amber through Emerald did you win by 350 points or more? Probably most.

Even an Abjured member got “stuck” in Amber and posted a rant YouTube video, eventually removing it.

Pyriall,

When three of your teammates start by running to cap home as a group…explain that as your personal problem.

When they then as a group go take down the beast, dungeon style, explain that as a personal problem.

When they die and return to the game be running into an outnumbered fight and dying again, explain that.

And how to you “carry” 3 people doing that, when you against a team that is playing properly? I mean specifically, not in general terms. “Talk to them”? They aren’t listening.

You don’t have time to explain such basics to them.

People say, “but Unranked doesn’t tell me how good I am”.

And Ranked accurately does?

I’d like to see the link as well. What happened?

My queue times are short. The games are close.

It’s just more fun. I suspect many of the people who love PvP and did not win the slot machine in Ranked Amber or Emerald just went to unranked.

There are not ever going to be all that many players in the diamond or Legendary ranks. The GW2 PvP player base is not large enough.

And with the guaranteed loss blowout problem in amber/emerald, people seem to be leaving ranked.

Pull the handle, watch ’em spin.

They should just award the points and not have you play these matches.

Remember, your MMR actually measures the win/loss of the TEAMS you were assigned to.

It only changes based on what the TEAM you are assigned to does, and that includes all the teammates you got that went AFK or ran useless builds and died in seconds every single fight. The ones who streamed one by one into 3 enemies and died instantly. Over and over.

All of their unskilled, clueless actions change your MMR.

Well, there are certain things you watch in the first two minutes that are so amateurish, that you just know that your assigned team members are not going to know what to do further down the line. And then when they start confirming that, people quit trying.

When experienced people observe their assigned team members do something really stupid at the beginning, and all die in 2 minutes, and see the opponent score start to skyrocket, they pull the plug in a few minutes. They stand around. They go through the motions.
Why go through the motions.

Mightybird.6034:

Haha, yes exactly. There is a definite random factor feel when soloq. The difference in MMR between teams is always so great there is never any reason to actually play.
Score will end up ~500-100 either way.

To make it worse I can almost predict how it will turn out within the first 20 seconds or so of a match (depending on map.) Just by my team’s initial behavior and where they head.

I can tell quickly as well.

Unranked
People work towards objectives
Matches are generally fair
No need to play more than you want to
Less is riding on each win or loss

Ranked
People work towards objectives
Matches are generally not fair at lower ranks
Must play daily for hours to progress fast enough
If you play for 500 hours, you might get a backpack “skin”

Thanks, I just had fun making it. Wasn’t trying to beat the horse.

Pull the handle…

They could make a slot machine screen where you “pull the handle” to join the queue.

Then you’ll see two rows of five spinning wheels. Whee!!

They settle, and you see the MMR of each person and their class, and the expected win.

BINGO!!!

You get a pip or lose a pip.

As I said a few days ago:

“Why play the match?”…

https://forum-en.gw2archive.eu/forum/game/pvp/Don-t-have-them-play-just-award-the-match/first#post6011358