Guild Wars 2

The original Glicko system was found to be a bit faulty because the tiers became “locked” and there was rarely any movement. So, ANet added a random variable, which is also behaving in a way that might not be intended. Currently, there is a 240 point gap between ranks 12 and 13. (Crystal Desert: 1662, Borlis Pass: 1422) 240 points above Crystal Desert includes 6 servers, and 240 points under Borlis Pass includes 5 servers.

When lower ranked servers face higher ranked servers, they often gain ranking even if they are stomped. And the reverse is true for the higher ranks; they aren’t stomping hard enough to gain rank points.

Statistically, we’re more likely to be matched up against servers that are closer to our ranking, so Borlis will probably fight opponents under them, and Crystal opponents above them. We’ve seen matchups that exceed this 240 point spread already, so it’s possible that we can bridge this chasm in points. If the RNG doesn’t make it happen soon, however, we’ll have a gap in points that can’t be bridged.

240 isnt big enough of a gap to cause real issues in this system. glicko cant handle ~400 point gaps.

I wouldnt be too worried about a 240ish point gap causeing stagnation. Look at SOS’s matchup list since this started. We’ve had TC or one of the T1 servers every time so far. Heck we got SOR with a rateing point gap well over 400 at the time (and still I think) Our rating has risen dramaticly even after 5 weeks of geting steamrolled just cause the RNG keeps throwing us upwards over relitively large pointgaps.

Honestly I didnt mind the 2 strong servers vs us matchups. the 2 stronger servers tended to focus on eachother out of necessity while ocasionaly turning to stomp us to pad their PTT and that let us hold our own at times. The match’s vs 1 strong server we suffered more because that server could stomp both us and whatever poor server got to come along for the ride on coverage alone.

the problem we had in the previous system was “artificial” rating gaps, and that the sizes of these gaps were not true indicators of the differences in server strength. these artificial gaps arose because of the relative rarity of matches played across tier boundaries, and the particular matchup system that only ever allowed rating points to flow up across a tier boundary but never down again.

“real” rating gaps that reflect real gaps in server strength are actually a good thing — in order to get good matchups, you want to avoid having very strong servers facing very weak ones, so it’s important that the ratings be decently accurate measures of server strength (at least, as long as you’re using rating to decide matchups).

under the new matchup system, a rating gap is a good indicator that the servers above the gap are consistently stronger than servers below the gap. the validity of the gap can be periodically verified by having a server above the gap play a server below the gap (and the randomization system helps to ensure that this kind of thing happens from time to time).

all that said, there is definitely a risk that we will see 2 tiers (1-12, and 13-24). if the gap between 12 and 13 gets big enough, randomization won’t allow matches that cross this gap to happen (there would be much less risk if the gap fell on a boundary that wasn’t a multiple of three — the fact that the number of servers on each side of the gap is an exact multiple of three is what makes this situation possible). you will know that this is happening if the number of matchups that cross this boundary get rarer and rarer, eventually stopping altogether. the way the randomization system works, natural tier boundaries will develop slowly, rather than suddenly. and if they occur, they can be “fixed” by adjusting the randomness parameters.

-ken

They need to make it so if you win you gain rating, if 2nd you either gain a little or lose a little(depending on score) and 3rd loses rating, makes thing simple and actually at least a little point to win and won’t get stuck in similar matches/continually widening gaps

Also, something that might assist in fixing the issue would be for Anet to take the 600 rating points they put into the bottom servers out of the top servers thus closing the gap slightly between rank 12 & 13 since we do have an extra 600 rating floating in the NA bracket.

ratings are actually relative. taken individually, a rating doesn’t tell you anything at all. but if you take the ratings for two servers and compare them, the difference between the ratings tells you how much stronger one server is than the other.

only the difference matters.

a server with a rating 200 higher than another server is expected to win by a certain amount. it doesn’t matter whether the ratings are 2200 and 2000, or 400 and 200. as a result, “extra” points in the rating pool don’t make any difference — the extra points will have a tendency to spread themselves out over all the servers, shifting every server up by about the same amount. you could also artificially take rating points out of the pool by subtracting the same number of rating points from every server, and none of the rating or ranking math would change (although there would be some impact on matchups).

artificially closing the gap is a bad idea, unless you want the 12-and-up servers to play the 13-and-under servers more often. but the reason the gap exists at all is because the 12-and-up servers are much stronger than the 13-and-under servers — any server rank 12 or higher could easily beat any server rank 13 or lower. such a matchup wouldn’t be fun for either side, so they should be made to happen less often, not more often.

in the current matchup system, the way you make a matchup happen less often is, you make the ratings farther apart. servers that have closer ratings will play each other more frequently, and servers that have ratings far apart will play each other rarely. if the ratings are far enough apart, some servers will never play each other at all (for example #1 would never play #24 — the ratings are simply too far apart to allow it to happen).

if you really want to reduce the rating gap, you need to reduce the ‘strength’ difference by weakening the higher-ranked servers, or strengthening the lower-ranked servers (or both). you can do this by changing the scoring system so that stronger servers play at some disadvantage (for example, make objectives worth fewer and fewer points the more of them you already hold, or make them harder to hold when you have lots of them by weakening walls/gates), or you can do it by encouraging (or forcing) the players to redistribute themselves more ‘fairly’ across all the servers.

-ken

Merendel.7128:

I wouldnt be too worried about a 240ish point gap causeing stagnation. Look at SOS’s matchup list since this started. We’ve had TC or one of the T1 servers every time so far. Heck we got SOR with a rateing point gap well over 400 at the time (and still I think) Our rating has risen dramaticly even after 5 weeks of geting steamrolled just cause the RNG keeps throwing us upwards over relitively large pointgaps.

Honestly I didnt mind the 2 strong servers vs us matchups. the 2 stronger servers tended to focus on eachother out of necessity while ocasionaly turning to stomp us to pad their PTT and that let us hold our own at times. The match’s vs 1 strong server we suffered more because that server could stomp both us and whatever poor server got to come along for the ride on coverage alone.

We on CD haven’t seen that. What we have seen is that the two stronger servers take and keep our stuff for the ppt that is easier for them. No, not everyone is giving WvW up but I do see fewer and fewer of the more casual players in the maps.

it is time to group the servers into 3 pools (rank 1-rank9, rank10-rank18 , rank19-rank 24) instead of just 2 pools (rank 1-rank12, rank 13-rank24).

randomized match up from the same pool, just like current system……

I understand that my post is a bit watered down. 240 points is not pause for concern just yet. But we’re on the road to repeating the past mistakes of Glicko.

better still, do a complete randomness. you might get SoR fighting ET. wth, a blowout match is a blowout match anyway. no difference if a top rank server thumping middle rank or low rank server. and you’ll get your match up data, anet……

Toxic.4320:

We on CD haven’t seen that. What we have seen is that the two stronger servers take and keep our stuff for the ppt that is easier for them. No, not everyone is giving WvW up but I do see fewer and fewer of the more casual players in the maps.

Aye. Does our garrison on CDBL flip constantly because we are outmatched, or do opposing servers go for it because they know it’s made of paper and easier than capping the stronger server’s garrison? Most likely it’s a combination of both.

I know we’re not the only ones struggling over the past week or two. I often check in on SBI and SoS (our tier-locked opponents when we had tiers). SBI was beat up pretty bad the first few weeks, and SoS can’t seem to catch a break from TC. It’s probably too early for nostalgia, but hope you both are doing well despite the hardships!

Toxic.4320:

Merendel.7128:

I wouldnt be too worried about a 240ish point gap causeing stagnation. Look at SOS’s matchup list since this started. We’ve had TC or one of the T1 servers every time so far. Heck we got SOR with a rateing point gap well over 400 at the time (and still I think) Our rating has risen dramaticly even after 5 weeks of geting steamrolled just cause the RNG keeps throwing us upwards over relitively large pointgaps.

Honestly I didnt mind the 2 strong servers vs us matchups. the 2 stronger servers tended to focus on eachother out of necessity while ocasionaly turning to stomp us to pad their PTT and that let us hold our own at times. The match’s vs 1 strong server we suffered more because that server could stomp both us and whatever poor server got to come along for the ride on coverage alone.

We on CD haven’t seen that. What we have seen is that the two stronger servers take and keep our stuff for the ppt that is easier for them. No, not everyone is giving WvW up but I do see fewer and fewer of the more casual players in the maps.

I’ve seen you guys hold your own in some of your matches although you’ve had your blowouts as well. when you got paired up with us and TC we both got crushed because TC could field covrage comperable to both of us togeather. on the other hand we(SoS) did better in relitive terms in our TC/FA/SoS match in part because 1 TC couldnt compleatly blow out FA to pad the PpT even more but because the needed more of their population to counter FA leaveing fewer to stomp on us.

I’m not saying the more powerful servers wont target the weak ones to pad the PPT, they will and would be foolish not to. However they are less likely to leave significant numbers in your BL once they cap all if they need them to counter the other powerful server. Then agian you could be unlucky enough for the other two servers to have a proxy war on your BL as they fight over your land.

I am mystified by the formula that results in 5th and 6th playing 12th (Crystal Desert). I get that randomisation happens of course, but throw us a bone soon Mr Glicko sir, a lot of frustrated folk due to being matched at higher tiers for weeks on end. The match-up below us currently consists of 3 servers that are higher than us.

My brain hurts now.

5th and 6th playing 12th could be very likely, if the ratings are all close together in that area of the rankings, because matchups are made on the basis of how close the ratings are, rather than strictly on the basis of ranking (which is how the old system did it).

in fact, the ratings are within a few hundred of each other, which isn’t exactly “close” but it’s not far either. as a result this kind of matchup isn’t especially likely (at the start of the match, CD had a 9.6% chance to get DB and a 16.6% chance to get FA — getting both of them had a very small likelihood and you simply got unlucky this week).

I ran some calculations and CD only had a 1.019% chance to draw DB and FA in a tier 3 match, 0.106% to get them in tier 2, and 0.056% to get them in tier 4. by comparison, “likely” matchups include things like T4:YB+EB (5.874%), T4:KN+EB (5.383%), T4:KN+YB (5.316%). the most likely T3 matchup for you was T3:Mag+SoS (2.283%).

-ken

240 points should be the beginning of concern. We know now that the widest gap tends to be between 400-500 depending on the other variables.

Despite some wide swings, the majority of matches are within 300 points with the occasional outlier.

Additionally, since 12 is such a solid number for this system, it becomes increasingly unlikely that BP and SBI will get matched up with any server above #12 since all it takes is for them to get a single low roll to move up.

Likewise, CD with it’s last two matches has had artificial gains and continues to pull away.

This system is pulling servers towards their correct ranking, but it’s also creating an uncross-able bridge between #12 and #13.

The only screwball in the mix is Kaineng, which appears to be falling apart. Again, however, all they need is a ridiculously high roll to make the gap wider.

As long as the game mechanics favor zerginess then the wvw will continue to lose players.

Remove AOE cap

Remove downed in wvw

Fix the reward system. It currently favors zerglings with zero attention to skill or tactics.

The above will allow small guilds/servers to fight back by rewarding them for tactics and punish stupid zergs. Don’t worry, the casuals already left and never contributed much aside from the friday ‘guild wvw night’ !

Keep the consumables because Anet needs to make money off of it

Increase the rewards for running in small groups/zergs by increasing the badge rewards or titles etc.

one of the things to keep in mind is, the random factor is used only to set up a matchup ranking. the matches are then made by picking groups of 3 servers, starting at the top and working your way down. when the matchups are actually made, the sizes of the rating gaps are irrelevant.

this means there is no “hard cap” on how far apart ratings must be to make a matchup impossible. for example, if there were only 6 servers in NA, with ratings of 9990, 9980, 9970, 9960, 9950 and 100, the server with 100 rating points would get blown out every single week, no matter how far apart the ratings get. 3 out of the top 5 would get matched together, and the 2 left over would play the bottom server, and there is no amount of “rating gap” that would prevent it.

the only time a rating gap would prevent certain matchups from occurring is if the gap falls somewhere so there is an exact multiple of three servers above the gap, and an exact multiple of three servers below the gap. so if we had 12 servers with ratings 9990, 9980, 9970, 9960, 9950, 9940, 150, 140, 130, 120, 110, and 100, then the top 6 would always play each other, and the bottom 6 would always play each other, but there would never be a matchup with a bottom 6 server playing a top 6 server.

-ken

insanemaniac.2456:

240 isnt big enough of a gap to cause real issues in this system. glicko cant handle ~400 point gaps.

But 240 gap can already mean that one server with 240 better ranking can pwn the others heavily and get.

Edit: Look at EU: http://mos.millenium.org/matchups
Baruch Bay not even 200 differene to the others in their matchup and almost 2 times the amount of score points.

“Heavyly pwning” for me is 1.5-2 times the amount of score points than the others…
This only shows that the free transfers are needed more than ever. Lower servers need to get more population to balance it and make it more fun.

Luthan.5236:

Edit: Look at EU: http://mos.millenium.org/matchups
Baruch Bay not even 200 differene to the others in their matchup and almost 2 times the amount of score points.

this is exactly what the ratings would predict (or pretty close, at any rate). based on the rating difference, BB is expected to get 1.75 times the score of RS, and 1.64 times the score of AR. the fact that they are getting very close to the expected score is why mos.millenium.org shows that their rating is changing by only a small amount (+3.5 rating points). meanwhile, other servers are getting much larger rating changes because their actual score differs by a lot from what the ratings would predict.

if 1.5 times is your limit for “acceptable” matchups, then the target difference in rating should be about 125 rating points. any matchup where the servers differ by more than 125 points, you should expect the highest server to win by more than 1.5 times the score.

-ken

I think the top servers are overrated and the bottom servers are underrated against each other. The AR/BP/EB matchup was a gateway. Everyone underneath it was accurately rated against each other but not against anyone above. Both BP and AR are bleeding tons of rating this week because it’s the first week they’re fightin anyone from below (I know AR lost some folk and aren’t as strong as they were but I think they would still have bled a lot of rating). I think a lot of the gaps will close as more servers fight people outside what their tiers used to be.

XII.9401:

As long as the game mechanics favor zerginess then the wvw will continue to lose players.

Remove AOE cap

Remove downed in wvw

Fix the reward system. It currently favors zerglings with zero attention to skill or tactics.

The above will allow small guilds/servers to fight back by rewarding them for tactics and punish stupid zergs. Don’t worry, the casuals already left and never contributed much aside from the friday ‘guild wvw night’ !

Keep the consumables because Anet needs to make money off of it

Increase the rewards for running in small groups/zergs by increasing the badge rewards or titles etc.

Its already been stated its a technical limitation that caps AOE. Basically the way it was written that it would cause a ton of additional calculations which would lag down WvW.

As someone who runs in a Smaller group (typically no larger than 10). I tend to agree there needs to be some way to favor smaller group play, and ways to break up the Zerg, but every method that has been thought of goes against the ideology of the way the game was designed to make it easier to group.

Personally I think the way rewards are calculated needs to be changed and divided amongst the people who tagged it, and also have Players being worth more than just 60. Have them worth a Base value then modified by Player level, then modified by Wxp level. And have that Value spread across all that tag it.

Basically allow the zergs to mass kill people and kill tons of people at a lower value, and allow smaller groups to kill at a slower rate but have them worth more (because its not divided up). I’m told Dark Age used a similar system, that made both Zerging and 8 Man groups beneficial.

Snowreap.5174:

if 1.5 times is your limit for “acceptable” matchups, then the target difference in rating should be about 125 rating points. any matchup where the servers differ by more than 125 points, you should expect the highest server to win by more than 1.5 times the score.

That’s exactly the problem… for example with servers like Vabbi at EU that are around 400… then the next one 600.. and then a huge gap to 1000 where the differences are small. In such cases these 2 lowest servers always get pwned. And this needs some incentive for players to move there(for start free transfers would be good).

What if they created a “24×7” league for the top 6 servers from NA and EU. Do they have good enough coverage around the clock for the matches to be fun?

It was fun for us when we fought T2 servers, but it was probably dull for their non-NA players who had nobody to fight.

Chris.3290:

Likewise, CD with it’s last two matches has had artificial gains and continues to pull away.

wat?

http://mos.millenium.org/servers/history/34

Haha, I was going to post the same thing Waf. Something about CD that causes Chris to lose his ability to comprehend facts.

For some reason even though the numerical glicko gap is equal (hell it’s been smaller) to the gap we here in CD faced vs TC and DB for the last couple weeks (SoS and their opponents for another example, EBAY also) for some reason there hasn’t been a “random” matchup of the 13th rated server with the 12th ranked server. Once we “randomly” get one of these matchups then I think you’ll see that server gain massive points. It’s just to hard to believe that after two months of close battles with SBI that we are at a 300 glicko difference.

I’ve posted estimated matchup probabilities for the next set of matches here:
https://forum-en.gw2archive.eu/forum/wuv/matchups/Who-would-you-like-to-face-next-week/2354130

If you look at the NA probabilities, you can see a very clear break between Kaineng and Borlis Pass — servers above the break are relatively unlikely to play servers below the break (and vice-versa).

There is, however, still a small probability of such a matchup happening. It looks like any of the bottom 5 servers in the top half have a small chance to get any of the top 3 servers in the bottom half (and vice-versa) — about equivalent to rolling a “1” on a 20-sided die.

So we haven’t reached the point of total tier stagnation yet. It’s something to keep an eye on, but I’m not terribly worried about it for now.

-ken

Tier stagnation needs to happen.

Look at all the NA servers that are above BP. At one point in time they were the bandwagon server of the week.

The servers in the lower half of the tiers, do not have the population to be competitive with the upper tier servers.

This would be a great time for Anet to break up the NA servers into 2 distinct tiers.

Kodiack.2783:

Tier stagnation needs to happen.

Look at all the NA servers that are above BP. At one point in time they were the bandwagon server of the week.

The servers in the lower half of the tiers, do not have the population to be competitive with the upper tier servers.

This would be a great time for Anet to break up the NA servers into 2 distinct tiers.

Not really, tier stagnation made lots of crappy match ups that should have been broken up a long time ago but weren’t going anywhere. AR/BP/EB was one of them. There’s still a lot of ranking left to be figured out.

Well I’m getting very worried about this rating gap. SBI and BP are going to get screwed by it.

There is no way that SoS should be closer in rating to TC than SBI is to CD. There’s no way SoS should be closer to SoR than SBI and BP are to SoS. Etc… Etc… Etc…

There’s no way the chances of SBI playing servers above and below them should look like this (from Snowreap’s analysis):

https://forum-en.gw2archive.eu/forum/wuv/matchups/Who-would-you-like-to-face-next-week/2354130

Now I’m willing to give it a little more time but unless we or BP get matched up with servers higher than us, we will continue to lose rating. Anet needs to do some manual intervention. These ratings are not becoming more “accurate”.

Kodiack.2783:

Tier stagnation needs to happen.

Look at all the NA servers that are above BP. At one point in time they were the bandwagon server of the week.

The servers in the lower half of the tiers, do not have the population to be competitive with the upper tier servers.

This would be a great time for Anet to break up the NA servers into 2 distinct tiers.

waaaaaaait a second.. ive seen this post before… back when SF was stuck in t8 and all of t7 was too proud of themselves to accept that the whining about t8 was legit.

they were wrong.

you are wrong.

Johje Holan.4607:

Well I’m getting very worried about this rating gap. SBI and BP are going to get screwed by it.

There is no way that SoS should be closer in rating to TC than SBI is to CD. There’s no way SoS should be closer to SoR than SBI and BP are to SoS. Etc… Etc… Etc…

There’s no way the chances of SBI playing servers above and below them should look like this (from Snowreap’s analysis):

https://forum-en.gw2archive.eu/forum/wuv/matchups/Who-would-you-like-to-face-next-week/2354130

Now I’m willing to give it a little more time but unless we or BP get matched up with servers higher than us, we will continue to lose rating. Anet needs to do some manual intervention. These ratings are not becoming more “accurate”.

I agree 100%. Odds of BP and CDI getting matchups that will allow us to gain rating are low. The gap has a huge potential of widening enough to be uncrossable.

insanemaniac’s response to tier stagnation is spot on. The entire point of this system is so we don’t have to ask the question, “What would happen is X fought Y?” again. If someone wants to fight to the top, that’s their prerogative. Though we could come up with a lot of amusing things to call those who bandwagon to rank 13 to stomp the 13-24 bracket.

As ken said, we’re not yet at the point where this is an issue. While the gap widened just a bit this week, one matchup can undo all of that.

Thread is now invalid. Good day.