Guild Wars 2

Zen.8497:

Snowreap.5174:

GW2 doesn’t have new servers coming online, and it currently has no way to get cross-tier matches. So we need another mechanism to supply lower servers that deserve to move up with a source of rating points they can steal.
-ken

That is working backwards artificially boosting servers points above their levels when in reality the server strength remains the same. While everyone want’s diverse matchups nobody needs blowouts. As long the servers population is uneven and scoring system rewards coverage and population there will be diverse but some ugly matches down the road. The post below says it all.

Maladon.5760:

The problem isn’t the first week of matchups. The problem will be a few weeks down the line when higher ranked servers have been bled of ratings points by being matched up with lower ranked servers. At that point it’s very possible for really ugly matches to occur.

I wouldn’t be surprised if they ran some simulations and saw that coming. Vizunah Square vs Vabbi in three weeks! Which is why they have gone back to rework this.

Actinotus.6410:

Zen.8497:

Snowreap.5174:

GW2 doesn’t have new servers coming online, and it currently has no way to get cross-tier matches. So we need another mechanism to supply lower servers that deserve to move up with a source of rating points they can steal.
-ken

That is working backwards artificially boosting servers points above their levels when in reality the server strength remains the same. While everyone want’s diverse matchups nobody needs blowouts. As long the servers population is uneven and scoring system rewards coverage and population there will be diverse but some ugly matches down the road. The post below says it all.

Maladon.5760:

The problem isn’t the first week of matchups. The problem will be a few weeks down the line when higher ranked servers have been bled of ratings points by being matched up with lower ranked servers. At that point it’s very possible for really ugly matches to occur.

I wouldn’t be surprised if they ran some simulations and saw that coming. Vizunah Square vs Vabbi in three weeks! Which is why they have gone back to rework this.

Exactly. The problem is that any rating systems are methods for assessing players/servers strength in games and not to fix them. The root cause is uneven population and scoring system which rewards coverage and population.

Zen.8497:

The problem is that any rating systems are methods for assessing players/servers strength in games and not to fix them. The root cause is uneven population and scoring system which rewards coverage and population.

They could fix this by putting in a means to limit each world to be only able to field the same number of people.

Say for example.. you put in a mandatory 5 min Queue to get into the WvW System. Then the people that entered the queue would be allowed to enter the WvW system in equal numbers with the other servers.

IE: Each team is limited on how many people they can have in the whole of WvW, based on how many people each side has on the whole of WvW.

Once in the WvW System, they would be free to move about any map, so queues for maps and such would go away.

In Example- Lets say JQ had 100 in the system and 100 more hit the Queue, and Blackate had 98 people because 2 just left, and 70 more in queue. Blackgate would get all 70 people into the WvW system on the next 5 min pulse, and JQ would get 68.

I guess they could even put the number up on the WvW panel, so that players know if they have a chance to get into the WvW game at all tonight. If you see that you server is filling the Queue, and the other servers have no one waiting to get in, then you might not get to play tonight.

I guess a kick mechanic would also need to be put in, like if a server had a lot people just leave, then a re-balancing would need to happen at the 5 min pulse, and people would need to opt to leave WvW or run he risk of being RNG kicked.

IE: Lets say, ET has 40 people in the system and FC has 34 because 6 people left, both have no one in the Queue, so a quick “re-balancing – 6 people must leave” would come up for ET, like it does in sPvP, and players could opt to leave, but would see the number count down, so that 10 people don’t jump ship, and even if they did, then FC would need to lose people by the next 5 min pulse, to keep things even.

I suppose a 10% variable could be allowed, so you are not kicking single people when you have hundreds playing.

Do I think people would be happy about that kind of equalizing, I doubt it, but it would balance the forces out. But we all know, players would complain about that “feature” and if not that, they would complain about something else.

Zen.8497:

Exactly. The problem is that any rating systems are methods for assessing players/servers strength in games and not to fix them. The root cause is uneven population and scoring system which rewards coverage and population.

And the particular problem is that the rating system we’re using (Glicko-2) cannot accurately kitten server strength when a server only plays servers in their own tier.

According to the inventor of Glicko-2, the system “works best when tthe number of games in a rating period is moderate to large, say an average of at least 10-15 games”. When you play that many games, it is expected that you will play some players that are close to you in rating, and others who are far.

But in GW2, each rating period has only 2 games (1 game, played against 2 simultaneous opponents, counts as 2 games for Glicko-2), and the 2 opponents are usually the same ones over and over.

As a result, the rating system is only a useful measure of relative strength within a tier, because those are the servers you have actually played against recently. But the ratings are not an accurate measure of server strength across tiers, because cross-tier matches happen so rarely.

The new system doesn’t artificially give points to servers that don’t deserve them — a high-tier server is expected to blow out a low-tier server, and the high-tier server would only lose points if they fail to win by at least the expected amount.

But if a high-rated server is actually overrated (say, because a lot of their players have quit or transferred) there needs to be a way for the points that server got when it was strong to bleed into lower servers as it weakens. Right now, the #1 ranked server can only bleed points to ranks #2 and #3, but the points can never fall any further than that.

Imagine if everybody in tier 1 quit today. The ratings for those servers would then obviously be too high, but there is no possible way for those points to ever come down to a lower tier.

-ken

Actinotus.6410:

Maladon.5760:

The problem isn’t the first week of matchups. The problem will be a few weeks down the line when higher ranked servers have been bled of ratings points by being matched up with lower ranked servers. At that point it’s very possible for really ugly matches to occur.

I wouldn’t be surprised if they ran some simulations and saw that coming. Vizunah Square vs Vabbi in three weeks! Which is why they have gone back to rework this.

I don’t really believe that, especially I think the example is at least as bad as ANet ad-example, but should be something like that be the problem the only way out I see is shorter matches, they hurt less (as they are faster over) and they produce more ratings in a shorter time, so an adequate adjustment happens faster.

If the players would learn how to play to minimize the lopsidedness of the matches then things would be much better.

I am hoping that this new system causes people to stop “playing for second” to avoid going down a tier. Because essentially tiers no longer exist. Your next match is not soley determined by your rating.

So there is no reason for S2 to focus S3 to avoid dropping and vice versa. They can both do what they should – focus S1.

Thobek.1730:

Vincent Von Demetri.1450:

JonPeters.5630:

timidobserver.7925:

I have a couple of serious questions.
1. Did the person that came up with the math for this graduate from high school.
2. To prove it, please upload a copy of your high school diploma.

How about just a copy of their resume. I don’t see his high school experience on there. Hopefully the other stuff can count as relevant experience for your assessment.

http://www.glicko.net/cv.html

While the Glicko system is a staple of chess rankings and many other ranking systems, I think it is worth saying that the implementation of a tool is more important than the tool itself. This is where the famous phrase “The right tool for the job” comes in. For example, the screwdriver and hammer are both time tested tools, both are proven to be successful, but each has a very specific application. If I was trying to torque a screw, I wouldn’t want to use a hammer. Get the picture Jon? Someone’s qualifications don’t indicate the success of a system and it’s implementation. I’m by no means saying that the Glicko system isn’t the right application, because, well, I don’t know whether it is or isn’t. It doesn’t take a statistician however to say that the current ranking system is entirely broken… the biggest problem is that the servers are not all on equal footing. Why not create a system that encourages players to spread out to various servers… for example, why not control the maximum amount of players in WvW for one server by applying a cap based on the standard deviation of the lowest server? This would allow for some level of control, but won’t entirely prevent one team from outmatching another (by more than a standard deviation).

Even though this was the creator of the glicko system, are they using it for GW2? Do you not think Mark Glickman could of possibly create a system specifically for WvW.

Glickman does not work for ANET. They are using his Glicko rating system, which was not specifically developed for GW2.

Using such a precise system is silly when WvW is so unbalanced because of WvW timezone populations.

Just keep it simple.

There’s cross tier population groups forming now, and I think it might be warranted to separate by population at this point.

tier 1/2 are distinct from all servers right now in terms of coverage/population

tier 3/4 are fairly similar

tier 5/6 are similar minus the green in t5

tier 7/8 are fairly stable

Instead of having 3 server clusters, I’d like to see a system that does a 6 server cluster with rotating matches between. Promotions and demotions should be done manually. It’s pretty clear when a server needs to move.

The 6 server clusters themselves would have much more mobility than the current 3 cluster, maybe even doing the winner up loser down, but restricted to their own group.

Dayra.7405:

fishergrip.4082:

This change will break WvW worse than the proposed reset of a few months ago.

Definitely not.
The reset would have put a uniform distribution over match-likelihood for several weeks, where in the second round a match First vs Winner of T8/9 was nearly sure.

This one puts a bell curve centered around you capabilities over the match-likelihood.
And a match spanning over more than 10 places is impossible.

Yes, the reset would be been insanely broken in the short-term, but matchups would have gone back to normal after a month or two. Under this new system matchups are permanently broken.

Snowreap.5174:

Mystery.7392:

What will happen with Vabbi [27th]? They are like 600 points behind the 25th rank.

so, based on the numbers, Vabbi and Fissure of Woe are almost guaranteed to be matched with each other. the third server in the matchup is more of a tossup, but Blacktide, Arborstone and Whiteside Ridge are likely. Ring of Fire and Dzagonur are less likely.

-ken

Hey u were right about 1 thing Vabbi and FoW were matched together, the 3rd server is Ring of Fire though.

Personally i don’t know about NA but EU’s match up’s are very messed up in my opinion. I dont really understand this system but i do hope that it stabilizes in a few weeks, It ain’t fun atm.

I’ve just come from RoF, playing against Vabbi and FoW. There’s really no fun in trouncing someone that hard. It’s not because we were better players. It’s the servers themselves. Someone made the suggestion, Kajabi or someone, I forget, that if servers are so underpopulated they get grouped together, or put with better servers. To hell with the rewards. People play the game for a fair challenge, not for some virtual reward. We currently control the map, +695 points, about 200k to 15k to 7k or so, and I wish we didn’t, for everyone’s sakes. If you really want people to learn new tactics, group 2 servers up on each colour, so it’s 2v2v2. Then you can mix 2 good (high pop)servers with 2 bad (low pop) servers, or just randomly. It would even things up a lot. Hell, even have weeks when you have 3v3v3, just for a laugh. People would enjoy that, and actually learn some new tactics instead of being swarmed by superior numbers. People might make some friends on other servers as well. I don’t know what issues this system was supposed to combat, but at the moment it does look sucky.