Guild Wars 2

YBs performance in t2 actually shows that the choice was a good one. Well, the Glicko-matchmaking system just doesn’t react fast enough on the current population changes. But I also think it is weird to intervene with Glicko adjustments instead of just creating matchups that make sense, because when you change something then there are always other servers affected by this decision and if those servers are the “wrong” ones to go down or up then they will of course cry and also ask for intervention.

It really would be easiest if a person would take a look at the results at the end of the week and decide over the new matchups with the population, results and Glicko in mind. Like Bridget Morrigan I understand why they wanted to make it “random” but that just doesn’t work with the current strength differences of the servers and the Glicko system imho. But it would also be acceptable if the matchmaking system just gets adapted to a more reactive system that takes strength differences into account better and has reduced chance to create blowout matches (so for example would put DB into T3, FA into T2, CD in T4 and NSP in T3 at the end of this week if current scores would be final ones).

Rink.6108:

YBs performance in t2 actually shows that the choice was a good one. Well, the Glicko-matchmaking system just doesn’t react fast enough on the current population changes. But I also think it is weird to intervene with Glicko adjustments instead of just creating matchups that make sense, because when you change something then there are always other servers affected by this decision and if those servers are the “wrong” ones to go down or up then they will of course cry and also ask for intervention.

It really would be easiest if a person would take a look at the results at the end of the week and decide over the new matchups with the population, results and Glicko in mind. Like Bridget Morrigan I understand why they wanted to make it “random” but that just doesn’t work with the current strength differences of the servers and the Glicko system imho. But it would also be acceptable if the matchmaking system just gets adapted to a more reactive system that takes strength differences into account better and has reduced chance to create blowout matches (so for example would put DB into T3, FA into T2, CD in T4 and NSP in T3 at the end of this week if current scores would be final ones).

That the problem Glicko should never = population its a ranking system if ppl move then they move that should not make the RANK of a world go up. Playing wvw should be the only thing that makes rank go up.
You can move the world to what ever rank you want but your never going to pin down players and your never going to be able to have ranks keep up even when you “fix” them as they did last reset. Unless you want them to changes the glicko by there own hands (something that is going to rip the hart out of wvw) your always going to have blow outs.

Bridget Morrigan.1752:

I think it’s pretty obvious for anyone watching the scores right now that both the T2 and T3 matchups are closer and more competitive in their current makeup than they would have been if YB and FA had stayed where they were.

Bridget, I finally got some time earlier this evening to read your essay. I think you write very well on the issue. This one quote bothers me though and here’s why.

A server’s performance is always sensitive to multiple factors, the most visible one being a server’s population and coverage at any given time. The other factor not immediately visible is the mix of servers they are together in a match with, what those server’s population and coverage looks like. To me that makes what is “obvious” not so “obvious”. DB lost a lot of population and isn’t linked to another server and can’t bring the numbers the other servers bring. Score gain by the two other servers therefore may not truly be reflective of what those servers are capable of in a true three-way match, and especially the rating gain/loss due to the “lag” of glicko adjustment along with the fact that volatility and deviation were not reset with the last server links.

This is why I never truly trust score as a method measuring server performance and why this round of glicko ratings are hard to rely upon.

Jski.6180:

That the problem Glicko should never = population its a ranking system if ppl move then they move that should not make the RANK of a world go up. Playing wvw should be the only thing that makes rank go up.

The point is the current system did not react fast enough on the population changes that happened, thus creating blowout-matches. So I think the population has to be a factor the arenanet admins have to take into account as well a) for linking and b) for matchmaking. It can play a minor role behind the performance of a server of course, but especially for having a server move up or down a tier it surely would be a good variable to find out how good the server would perform there. For this tier changes the performance of the server in a lower tier is not a good predictor variable because the opponents also are likely to have far less population than in higher tiers.

Rink.6108:

Jski.6180:

That the problem Glicko should never = population its a ranking system if ppl move then they move that should not make the RANK of a world go up. Playing wvw should be the only thing that makes rank go up.

The point is the current system did not react fast enough on the population changes that happened, thus creating blowout-matches. So I think the population has to be a factor the arenanet admins have to take into account as well a) for linking and b) for matchmaking. It can play a minor role behind the performance of a server of course, but especially for having a server move up or down a tier it surely would be a good variable to find out how good the server would perform there. For this tier changes the performance of the server in a lower tier is not a good predictor variable because the opponents also are likely to have far less population than in higher tiers.

When your talking about say T1 and T2 its more then population there is a true skill factor going on both player (pugs) and commanders. T3 players so far do not put up real fights in open field. Its kind of boring.

It also might not hurt to remember that YB went into free fall originally due to the way glicko works. It didn’t actually lose nearly as many guilds or pop as DB has – but it got stuck in the cycle of “can’t possibly perform at the level glicko thinks it should, therefore freefall”.

Barring any addition major changes, it was going to bottom out in t4, then move back up, as glicko adjusted. That didn’t happen normally because of the temporary glicko given to CD.

Jski.6180:

When your talking about say T1 and T2 its more then population there is a true skill factor going on both player (pugs) and commanders. T3 players so far do not put up real fights in open field. Its kind of boring.

There is clearly a huge population difference between the tiers. Maybe there is some skill factor as well with some specific commanders, but I doubt that the impact is that big. I don’t think the average pug is worse in lower tiers, just less used to fighting big blobs maybe.
I think the population difference has to be taken into account even if there would be an additional skill factor.

SBI loves to fight open field given the fight-commanders (look for HATE guild for example) and I know some commanders on SoS and CD that love fights as well, even if they are outnumbered. If you play primetime and don’t demoralize the opponent too much, then you should get your fights. If you don’t play primetime, then you will probably outnumber the opponent too much for a fair fight. ^^

Rink.6108:

Jski.6180:

That the problem Glicko should never = population its a ranking system if ppl move then they move that should not make the RANK of a world go up. Playing wvw should be the only thing that makes rank go up.

The point is the current system did not react fast enough on the population changes that happened, thus creating blowout-matches. So I think the population has to be a factor the arenanet admins have to take into account as well a) for linking and b) for matchmaking. It can play a minor role behind the performance of a server of course, but especially for having a server move up or down a tier it surely would be a good variable to find out how good the server would perform there. For this tier changes the performance of the server in a lower tier is not a good predictor variable because the opponents also are likely to have far less population than in higher tiers.

The GW2 glicko is manifestly inadequate in matching servers and in responding to server changes and yes server population could somehow be factored into the matchmaking calculations.

On linkages however anet is manually making decisions and is clearly doing a woeful job. Of course server population at a point in time has to be taken into account when deciding who is linked, however in deciding what servers don’t get a link I would now argue that actual results should play a large role in deciding whether a server gets a link or not. Anet do not seem to understand how hamstrung servers get when they don’t get a link for 2 whole months.

morrolan.9608:

On linkages however anet is manually making decisions and is clearly doing a woeful job. […] in deciding what servers don’t get a link I would now argue that actual results should play a large role in deciding whether a server gets a link or not.

I agree that the results should also be included in the choice of the linking as well of course.
I don’t agree on the “Woeful job” though. Arenanet made the linkings in NA with the feedback from past linkings in mind: T1 got too stacked when combining a full server and smaller servers where people could transfer to. This lead to long queues additionally to small servers complaining about the loss of their identity. so they wanted to make T1 out of servers without a link. Small linkings to T2 and T3 and 2 similarly big servers together in T4 with a small server. It may not be perfect, but at the time the linkings were made, this really seemed like a good, balanced choice and good matchups. Arenanet isn’t to blame about people leaving DB – if that is what you are refering to .

morrolan.9608:

Anet do not seem to understand how hamstrung servers get when they don’t get a link for 2 whole months.

Nah. It’s the players still thinking it is 2013. Ever since the world population calculations were changed, it became clear what the vision was for server size. I don’t think it is an accident that FA remains Very High and has skirted Full status without a single timezone one would consider as stacked ever since the population calculation was changed.

Then there’s the skill lag that occurs in larger fights that I think Anet tried to discourage in the higher tiers with the population calculation. In addition, something not talked about a lot, not everyone plays this game on the best computer available. FPS dropping to slideshow levels is a big disincentive to players to zerg. I wouldn’t be surprised if a survey were taken that showed a relationship between tier/playstyle and hardware/graphics settings.

Chaba.5410:

Bridget Morrigan.1752:

I think it’s pretty obvious for anyone watching the scores right now that both the T2 and T3 matchups are closer and more competitive in their current makeup than they would have been if YB and FA had stayed where they were.

Bridget, I finally got some time earlier this evening to read your essay. I think you write very well on the issue. This one quote bothers me though and here’s why.

A server’s performance is always sensitive to multiple factors, the most visible one being a server’s population and coverage at any given time. The other factor not immediately visible is the mix of servers they are together in a match with, what those server’s population and coverage looks like. To me that makes what is “obvious” not so “obvious”. DB lost a lot of population and isn’t linked to another server and can’t bring the numbers the other servers bring. Score gain by the two other servers therefore may not truly be reflective of what those servers are capable of in a true three-way match, and especially the rating gain/loss due to the “lag” of glicko adjustment along with the fact that volatility and deviation were not reset with the last server links.

This is why I never truly trust score as a method measuring server performance and why this round of glicko ratings are hard to rely upon.

Yeah, at the time I wrote it, YB was in second and the score between YB and JQ close, and I didn’t anticipate the huge YB dominance of ppt in the last 24 hours or so. I didn’t realize DB had been gutted so thoroughly. FA is still running a lot closer match, however, than YB probably would have in the same matchup, though.

But I think the point that the statement was addressing still stands, which is that both T2 and T3 matchups needed intervention of some kind, because the Glicko-driven scoring system would never keep up with the population changes of either YB or DB. If anything, the fact that YB is starting to pull ahead in T2 tends to indicate that the Glicko-driven system was even worse than I originally thought, and the more appropriate matchups would have been JQ/YB/FA and DB/SBI/XX.

In other words, that DB is still inappropriately matched just goes to show that tinkering with the Glicko just causes a cascading effect of other problems, because Glicko is still too unresponsive to the real-time population realities of the game, and it’s carrying too much outdated data with it. Fixing it by hand is not good; not fixing it by hand is also not good. So I’m still in the camp of thinking a complete overhaul of the way in which WvW does ratings and matchups is the only way to get out of this cycle. It needs a total wipe.

Would using last week’s player activity as a rating be better for matchups?

Bridget Morrigan.1752:

In other words, that DB is still inappropriately matched just goes to show that tinkering with the Glicko just causes a cascading effect of other problems, because Glicko is still too unresponsive to the real-time population realities of the game, and it’s carrying too much outdated data with it. Fixing it by hand is not good; not fixing it by hand is also not good. So I’m still in the camp of thinking a complete overhaul of the way in which WvW does ratings and matchups is the only way to get out of this cycle. It needs a total wipe.

That cascading effect is exactly why we needed the partial reset with new server links and why T4 had needed 2-month frozen ratings to be merged/adjusted somehow in order to avoid that artificially created 300 glicko ratings gap.

Here’s more possible tools with regards to “tinkering” of the glicko matchmaking that constitutes systemic changes rather than individual one-offs:

This post from three years ago for example identifies correctly how population and performance are intimately related and suggests solutions to automatic adjustment of performance prediction based on a fluidly changing population over the course of the week long match.

This post illustrates how this fluctuating population over the week affects the predictions, seen as an error which perhaps is only evidence of a symptom. “That 0.05 difference is significant over the course of the week where the overall proportion shrinks as the scores go up.” You see this when you saw that YB was in 2nd place and close in score to JQ and now they are not.

The issue of servers “belonging” to tiers counter-intuitively exacerbates this slow glicko movement. When a server only fights against two or three other servers, their rating becomes inaccurate because they never get compared to a larger sample size.

Bridget Morrigan.1752:

Rating—On each new linkup, ANet should reset Glicko scores, not to completely even, but to a range that first factors in population and participation of both/all servers in each new link pairing, then assigning a score to that composite.

Matchups—I think if the ratings can be dialed in and adjusted more effectively before the beginning of the linkup period, then a lot of the problems with matchups will be ameliorated because matchups will be predicted more accurately, and there won’t be as much carryover and antiquated data from old linkup periods. Matchups should be done on a system that weighs both Glicko and transfers/population, and I think this can and should be automated, so that any matchup would calculated based on Glicko + n(+/- transfers on/off), where n is a value to be determined by ANet according to what makes sense in their data and where they set the middle point of their initial Glicko range, but kept constant, ofc.

Great posts! Very sensible and level headed.

I think practically everyone agrees with your point on resetting the rating on each new linkup. The question is to what value to reset it. I don’t know if simulations can be run on the past linkings to determine a good value. There also are suggestions that have been made in the past that could be looked at. I do think that the actual rating needs to be adjusted and not just the volatility and deviation.

Changing the matchup method definitely needs to happen – and now, not “soon”. Perhaps an adjustment like you suggest taking transfers into account could be done, I don’t know how hard that would be to program. But it is clear that Anet can adjust ratings so I wouldn’t think it would be too hard. Or perhaps player activity could be used – but that can be gamed.

I don’t understand the problem here?
Is it that people don’t believe YB deserves to be moved?

Since the new pairing system’s purpose is to promote “good matches”, there should be no argument here, right?

Obviously, the old ladder-style system is not compatible with this new philosophy and we as players need to understand that we gain NOTHING for winning a match apart from two more lootbags (or four if we get the double-reset bug).

kefro.9312:

Until someone cries again and ANET do some random stuff making wvw even more of a joke.
Shame ANET and shame on yaksbend, kdr 0,6 running away from fights to go backcaping for ppt, making their way to tiers without earning it only because someone cried and asked help from ANET, probably destroying other server and affecting tier 2 server with cancerous ppt mindless pvdoors “players” that belong to dead tiers

I agree 1000%

Svarty.8019:

I don’t understand the problem here?
Is it that people don’t believe YB deserves to be moved?

Since the new pairing system’s purpose is to promote “good matches”, there should be no argument here, right?

Obviously, the old ladder-style system is not compatible with this new philosophy and we as players need to understand that we gain NOTHING for winning a match apart from two more lootbags (or four if we get the double-reset bug).

The problem is YB stole FA’s spot and now FA is upset.

Svarty.8019:

I don’t understand the problem here?
Is it that people don’t believe YB deserves to be moved?

Since the new pairing system’s purpose is to promote “good matches”, there should be no argument here, right?

Obviously, the old ladder-style system is not compatible with this new philosophy and we as players need to understand that we gain NOTHING for winning a match apart from two more lootbags (or four if we get the double-reset bug).

Aside from kefro, I don’t think there is anyone else who is upset with the fact that FA ended in T3 but with how the issue was actually handled. The big deal here is that a manual intervention happened where it wasn’t entirely necessary (there wasn’t a glicko wall impossible to break), when similar issues were never adressed before and even more serious issues took longer to be adressed (The CD/SF/DH T4 matchup and the SF/FC/ET T8 matchup in the past).

People will say I am biased when I say I don’t mind the change because I am on YB. But tbh, seeing the current T2 matchup, I would rather have stayed in T3 and only go up when DB had finally dropped. I don’t care about my server rank like others seen to do. All I want is to be able to log and find people to fight. Somenthing that is somewhat hard when you are fighting 2 servers that rely purely on asians to make their score.

Dhemize.8649:

Svarty.8019:

I don’t understand the problem here?
Is it that people don’t believe YB deserves to be moved?

Since the new pairing system’s purpose is to promote “good matches”, there should be no argument here, right?

Obviously, the old ladder-style system is not compatible with this new philosophy and we as players need to understand that we gain NOTHING for winning a match apart from two more lootbags (or four if we get the double-reset bug).

The problem is YB stole FA’s spot and now FA is upset.

The real problem is FA went from T1 (all be it a very low % roll) to T3 in one go due to doable reset. That a major drop to the point of crashing levels. Its like going from fighting some of the better fighting groups to fighting ppl who seem to be all up levels. FA did not want to be in T1 so they did not play as such but to have such a massive drop so fast just to have YB jump up because the dev did what YB asked them to do is a bit too much. Its like they had made another YB glicko world in order to save YB and YB alone.

That and FA has been major rivaled with YB and TC for some time to so this is a major slap in the faces and gives the feeling Anet is agned FA over all though i think we have a lot of ppl on FA that Anet do not like.

If you want to get down to it CD and SBI should not be in T3 if FA / YB are going to simply drop down to T3 every now and then. I think they need to team them up with more worlds or take glicko away from these 2 worlds after this match.

Jeknar.6184:

I don’t care about my server rank like others seen to do. All I want is to be able to log and find people to fight.

FA wants GOOD ppl to fight running though zergs that put up no fight is not fun. That why FA wanted to stay in T2 that is where all the gvg fight guilds are and FA was and still is a major world for this.

Jski.6180:

FA wants GOOD ppl to fight running though zergs that put up no fight is not fun. That why FA wanted to stay in T2 that is where all the gvg fight guilds are and FA was and still is a major world for this.

GvG guilds in T2… What is this? 2014? Pretty much the last remaining fight guilds are on T1 right now. No one in their right mind would want to be stuck with JQ and DB.

YB didn’t ask either – OP was from other T3 server

Jski.6180:

The real problem is FA went from T1 (all be it a very low % roll) to T3 in one go due to doable reset.

Hrm no I don’t think that’s the real problem. As glicko rating differences shrink between the top and the bottom, a server in the middle like FA is going to enjoy more variation in match-ups. That is something that can be expected based on the current system. The real problem is player agency moved out of the game and onto this forum. Don’t like your match-up? Just quit playing and ask to have the other server’s rating adjusted here outside of the established system of rules.

Jski.6180:

FA wants GOOD ppl to fight running though zergs that put up no fight is not fun. That why FA wanted to stay in T2 that is where all the gvg fight guilds are and FA was and still is a major world for this.

Where are these fight guilds you speak of? They must not have been on last night much, or I wouldn’t have gotten so many bags…

Jeknar.6184:

Jski.6180:

FA wants GOOD ppl to fight running though zergs that put up no fight is not fun. That why FA wanted to stay in T2 that is where all the gvg fight guilds are and FA was and still is a major world for this.

GvG guilds in T2… What is this? 2014? Pretty much the last remaining fight guilds are on T1 right now. No one in their right mind would want to be stuck with JQ and DB.

Thats rather ironic coming from someone on YB.

Chaba.5410:

morrolan.9608:

Anet do not seem to understand how hamstrung servers get when they don’t get a link for 2 whole months.

Nah. It’s the players still thinking it is 2013. Ever since the world population calculations were changed, it became clear what the vision was for server size.

Yeah I guess but then linkages distort that.

Jeknar.6184:

Jski.6180:

FA wants GOOD ppl to fight running though zergs that put up no fight is not fun. That why FA wanted to stay in T2 that is where all the gvg fight guilds are and FA was and still is a major world for this.

GvG guilds in T2… What is this? 2014? Pretty much the last remaining fight guilds are on T1 right now. No one in their right mind would want to be stuck with JQ and DB.

GvG guilds………there is no competition today. The great ones are gone.

morrolan.9608:

Jeknar.6184:

Jski.6180:

FA wants GOOD ppl to fight running though zergs that put up no fight is not fun. That why FA wanted to stay in T2 that is where all the gvg fight guilds are and FA was and still is a major world for this.

GvG guilds in T2… What is this? 2014? Pretty much the last remaining fight guilds are on T1 right now. No one in their right mind would want to be stuck with JQ and DB.

Thats rather ironic coming from someone on YB.

hes fc, but the real issue is we play at different times of the day where #s are no where close for hours…. You all don’t get the OCX/SEA and we don’t get the NA (maybe EU either)

The only thing that should have happened is for DB to have been manually moved into T3. T2 should’ve been JQ/FA/YB this week. If that’d happened this thread would’ve topped out at 2 pages.

The fact is that YB, right now, is too strong for T3 and below. DB, right now, is too weak for T2. Glicko is not the right system for calculating server strength with linkages taking place, but it’s the system we have.

For all that feel betrayed by ANet for somehow “breaking the rules”. Well with the retooling WvW has had the rules are obsolete.