Guild Wars 2

what we need are those tunneling vehicles like the Dredge have.

-ken

by request, I’m posting estimated probabilities for the upcoming matchups in this thread also. these numbers are based on the currently predicted ratings (which are in turn based on the current scores), but if the actual end-of-match ratings differ significantly, that will throw the numbers off.

the “na1” and “eu1” files have probabilities of getting a particular server as an opponent, while the “na2” and “eu2” files have probabilities of getting specific pairs of servers as opponents.

-ken

here are the estimated probabilities for the new matchups, assuming that predicted ratings are reasonably close to what we will see at the end of the current matches. as always, if predicted ratings turn out to be different from the actual end-of-match ratings, that will throw the numbers off.

the “na1” and “eu1” files have probabilities of getting a particular server as an opponent, while the “na2” and “eu2” files have probabilities of getting specific pairs of servers as opponents.

also, I’ve included the predicted ratings that I use for these calculations. these are the “na” and “eu” files. these values should fairly closely match predicted ratings from other sites at the time I post them.

-ken

here are some links that explain how Glicko-2 works:
http://www.guildwars2guru.com/news/884-the-math-behind-wvw-ratings/
http://www.glicko.net/glicko/glicko2.pdf

-ken

I went back to look at the spreadsheet again and this is a more accurate way to phrase things:

TC needs to get fewer than 7991 points for every 10000 points JQ has, in order to lose rating points to JQ.

TC needs to get fewer than 8266 points for every 10000 points BG has, in order to lose rating points to BG.

I can’t attach the spreadsheet itself because the forum software won’t take Excel files. since it’s a binary file, putting a .txt extension on it probably wouldn’t work.

-ken

by request, I’m posting estimated probabilities for the upcoming matchups in this thread also. these numbers are based on the currently predicted ratings (which are in turn based on the current scores), but if the actual end-of-match ratings differ significantly, that will throw the numbers off.

the “na1” and “eu1” files have probabilities of getting a particular server as an opponent, while the “na2” and “eu2” files have probabilities of getting specific pairs of servers as opponents.

-ken

here are the estimated probabilities for the new matchups, assuming that predicted ratings are reasonably close to what we will see at the end of the current matches. as always, if predicted ratings turn out to be different from the actual end-of-match ratings, that will throw the numbers off.

the “na1” and “eu1” files have probabilities of getting a particular server as an opponent, while the “na2” and “eu2” files have probabilities of getting specific pairs of servers as opponents.

also, I’ve included the predicted ratings that I use for these calculations. these are the “na” and “eu” files. these values should fairly closely match predicted ratings from other sites at the time I post them.

-ken

Glicko does not care whether you win or lose. really, it doesn’t.

what Glicko cares about is, did your server score better or worse than expected? if you do better than expected you get points, if you do worse than expected you lose points.

this means that if you lose, but you lose by less than Glicko predicted, you will gain points. likewise, if you win, but you win by less than Glicko predicted, you will lose points.

I have a spreadsheet somewhere that can be used to calculate how big your score needs to be compared to your opponents in order for you to gain (or lose) points. so for example, in T1 NA, if TC wants to lose points they must ensure that they have less than 4k point or so for every 10k of combined points from JQ and BG. so if JQ and BG scores add up to 100K, TC should have less than 40K points in order to lose points. assuming I’m reading the spreadsheet right, it’s been a long time since I made it.

-ken

there are many cases where the “best” match is still horrible. ideally, servers of similar strength would always be found in groups of 3 so they can be matched together, but unfortunately this is rarely the case.

-ken

FYI, for this week I tried editing my previous post instead of adding a new reply (in the other thread I post in, I always reply so there is a record of all the old numbers from previous weeks).

Pros: OP doesn’t need to change the ‘Answer’ to the new post

Cons: thread doesn’t get bumped, and replacing attachments on an old post is cumbersome and slow

I’m undecided on what I’m going to do going forward, but probably I’ll just add new replies. we’ll see.

-ken

here are the estimated probabilities for the new matchups, assuming that predicted ratings are reasonably close to what we will see at the end of the current matches. as always, if predicted ratings turn out to be different from the actual end-of-match ratings, that will throw the numbers off.

the “na1” and “eu1” files have probabilities of getting a particular server as an opponent, while the “na2” and “eu2” files have probabilities of getting specific pairs of servers as opponents.

also, I’ve included the predicted ratings that I use for these calculations. these are the “na” and “eu” files. these values should fairly closely match predicted ratings from other sites at the time I post them.

-ken

by request, I’m posting estimated probabilities for the upcoming matchups in this thread also. these numbers are based on the currently predicted ratings (which are in turn based on the current scores), but if the actual end-of-match ratings differ significantly, that will throw the numbers off.

the “na1” and “eu1” files have probabilities of getting a particular server as an opponent, while the “na2” and “eu2” files have probabilities of getting specific pairs of servers as opponents.

-ken

2014-07-31 edit: updated for the new matchup

here are the estimated probabilities for the new matchups, assuming that predicted ratings are reasonably close to what we will see at the end of the current matches. as always, if predicted ratings turn out to be different from the actual end-of-match ratings, that will throw the numbers off.

the “na1” and “eu1” files have probabilities of getting a particular server as an opponent, while the “na2” and “eu2” files have probabilities of getting specific pairs of servers as opponents.

also, I’ve included the predicted ratings that I use for these calculations. these are the “na” and “eu” files. these values should fairly closely match predicted ratings from other sites at the time I post them.

-ken

of course this is allowed. you cannot backstab a server that thought they were allied with you until you pretend to be allied with them first.

-ken

I just took a look at the matchup history here:
http://mos.millenium.org/servers/history/36

oh look, TC has been green once during the past 7 weeks, blue 3 times and red 3 times. in the past 5 weeks TC has been blue twice.

so OP’s complaint about TC being red 5 times in a row is actually false, and the rest of this thread is therefore a bit pointless.

-ken

if TC has been getting red over and over, that’s just bad luck. they should get red most of the time, blue sometimes, and green rarely.

chance	green				blue				red
42.95%	Jade Quarry		Blackgate			Tarnished Coast
21.76%	Jade Quarry		Tarnished Coast	Blackgate
20.33%	Blackgate			Jade Quarry		Tarnished Coast
5.93%	Blackgate			Tarnished Coast	Jade Quarry
6.29%	Tarnished Coast	Jade Quarry		Blackgate
2.73%	Tarnished Coast	Blackgate			Jade Quarry

-ken

here are the estimated probabilities for the new matchups, assuming that predicted ratings are reasonably close to what we will see at the end of the current matches. as always, if predicted ratings turn out to be different from the actual end-of-match ratings, that will throw the numbers off.

the “na1” and “eu1” files have probabilities of getting a particular server as an opponent, while the “na2” and “eu2” files have probabilities of getting specific pairs of servers as opponents.

also, I’ve included the predicted ratings that I use for these calculations. these are the “na” and “eu” files. these values should fairly closely match predicted ratings from other sites at the time I post them.

-ken

sorry, forgot. too late for EU but here are the NA probabilities.

-ken

here are the estimated probabilities for the new matchups, assuming that predicted ratings are reasonably close to what we will see at the end of the current matches. as always, if predicted ratings turn out to be different from the actual end-of-match ratings, that will throw the numbers off.

the “na1” and “eu1” files have probabilities of getting a particular server as an opponent, while the “na2” and “eu2” files have probabilities of getting specific pairs of servers as opponents.

also, I’ve included the predicted ratings that I use for these calculations. these are the “na” and “eu” files. these values should fairly closely match predicted ratings from other sites at the time I post them.

-ken

I do it the same way, using simulated random matchups (also called Monte Carlo analysis, or the Monte Carlo Method), except that I run 100,000,000 (100 million) matchups for each file. doing that many trials takes 4-5 minutes on my laptop, using a program written in C#.

-ken

Chaba.5410:

Snowreap, please consider adding the predicted ratings used in your analysis. I’m kind of wondering what “reasonably close” means and how much of a difference is needed to throw the numbers significantly off.

I’ve included predicted ratings in my most recent post, and I also went back and added them to previous posts going back to May 28, so that you can see how week-to-week rating changes cause the probabilities to shift.

it’s important to keep in mind that the most important thing affecting the probabilities is the relative difference in ratings — the size of the gap between your rating and the ratings of the servers around you is more important than the exact number.

-ken

here are the estimated probabilities for the new matchups, assuming that predicted ratings are reasonably close to what we will see at the end of the current matches. as always, if predicted ratings turn out to be different from the actual end-of-match ratings, that will throw the numbers off.

the “na1” and “eu1” files have probabilities of getting a particular server as an opponent, while the “na2” and “eu2” files have probabilities of getting specific pairs of servers as opponents.

also, I’ve included the predicted ratings that I use for these calculations. these are the “na” and “eu” files. these values should fairly closely match predicted ratings from other sites at the time I post them.

-ken

yes. during Tournament play the matchups are very predicatable in the sense there is no randomness involved.

but during off-season (normal) play there is some randomness involved. this is not new, it has been this way for a long time.

-ken

agrho.2579:

I am proposing to change the way colors are assigned in WvW.
Right now according to ranking, 1st, 4th, 7th, 10th, 13th, … worlds are green; 2nd 5th, 8th, 11th, 14th, … worlds are blue; 3rd, 6th, 9th, 12th, 15th, … worlds are red.

Right now in NA, the following ranked worlds are green: 2, 4, 7, 11, 12, 15, 18, 21. The following are blue: 1, 5, 8, 13, 14, 17, 22, 23. The following are red: 3, 6, 9, 10, 16, 19, 20, 24.

In EU, the following ranked worlds are green: 2, 5, 7, 10, 12, 16, 18, 22, 25. The following are blue: 3, 4, 6, 11, 15, 17, 20, 23, 26. The following are red: 1, 8, 9, 13, 14, 19, 21, 24, 27.

The underlying assumption behind OP’s idea is actually not true.

-ken

here are the estimated probabilities for the new matchups, assuming that predicted ratings are reasonably close to what we will see at the end of the current matches. as always, if predicted ratings turn out to be different from the actual end-of-match ratings, that will throw the numbers off.

the “na1” and “eu1” files have probabilities of getting a particular server as an opponent, while the “na2” and “eu2” files have probabilities of getting specific pairs of servers as opponents.

-ken

“rank” is based on server rating, and nothing else. all the servers are sorted by rating, and the one with the highest rating is rank 1, the next one is rank 2, and so on until you get to the lowest rated server at rank 27.

“tier” is a concept that no longer has any special meaning. before we had randomized matchups, tiers were used to describe triplets of servers — “tier 1” was simply the group of servers ranked 1, 2 and 3. “tier 2” was the servers ranked 4, 5 and 6, and so on.

with the randomized matchup system we have now, there are no tiers, there are just matchups. “tier” numbers just refer to “matchup” numbers. server ratings still matter, but a random value is added or subtracted to shuffle things just a bit (not too much), then the servers are re-sorted and grouped into threes. the new “tier” number just refers to whether you were in the first grouping, the second one, the third, etc.

last week, your server (Blacktide) was ranked #22, and you were playing against servers ranked #21 (Miller’s Sound) and #19 (Dzagonur). you did well against them, and as a result you gained a rank and are now rank 21, while they each fell a rank and are now #22 and #20.

the important outcome of last week’s match is that you gained a rank, and moved up to rank 21.

who you are playing against this week has nothing to do with your rank or rating, it’s simply the outcome of the randomized matchup system. last week you were #22, playing against #21 and #19. this week you are #21, playing against #23 and #26. it doesn’t mean that Arenanet thinks you’re doing worse and need easier opponents, you just got lucky (or unlucky, depending on how you look at it) with your matchup this week.

for the people who still care about tiers, you actually went up a tier (because ‘classic’ tier 7 is the servers ranked #19, #20 and #21). last week you were in tier 8 (because ‘classic’ tier 8 is the servers ranked #22, #23 and #24).

in the current system, ‘modern’ tiers just represent which randomization group you are in for your matchup. the fact that you were in group 7 last week and in group 8 this week means nothing.

-ken

here are the estimated probabilities for the new matchups, assuming that predicted ratings are reasonably close to what we will see at the end of the current matches. as always, if predicted ratings turn out to be different from the actual end-of-match ratings, that will throw the numbers off.

the “na1” and “eu1” files have probabilities of getting a particular server as an opponent, while the “na2” and “eu2” files have probabilities of getting specific pairs of servers as opponents.

-ken

here are the estimated probabilities for the new matchups, assuming that predicted ratings are reasonably close to what we will see at the end of the current matches. as always, if predicted ratings turn out to be different from the actual end-of-match ratings, that will throw the numbers off.

the “na1” and “eu1” files have probabilities of getting a particular server as an opponent, while the “na2” and “eu2” files have probabilities of getting specific pairs of servers as opponents.

-ken

going by rank, it’s #12 vs #14 vs #16. I don’t really see the problem.

-ken

the meta-achievement was to complete 10 out of 19, I think. progress towards this goal is measured by the achievement at the very top of the panel.

-ken

now that season 2 is ending, it’s my understanding that we will be returning to the randomized matchup system.

under the randomized system, colors are not completely random, so that the highest-rated server in a matchup is more likely to be green, while the lowest-rated server in a matchup is more likely to be red. but there is still quite a bit of randomness (and therefore luck) involved.

for example, SFR has about a 36% chance to get green or blue next week (slightly higher for green), but only a 28% chance to get red.

-ken

last season, they said that ties would be ties (no use of rating as tiebreakers).

they have not said they would do it differently this season, so I think it’s reasonably safe to assume that it will be done the same way again.

-ken

for this coming week, SFR looks like it will have about a 35.9% chance of getting green, 35.5% chance of getting blue, and a 28.5% chance of getting red.

-ken

it’s the last week of season 2 tournament play, which means we should be going back to the randomized matchup system this Friday.

here are the estimated probabilities for Friday’s matchups, assuming that predicted ratings from today are reasonably close to what we will see on Friday. as always, if predicted ratings turn out to be different from the actual end-of-match ratings, that will throw the numbers off.

the “na1” and “eu1” files have probabilities of getting a particular server as an opponent, while the “na2” and “eu2” files have probabilities of getting specific pairs of servers as opponents.

-ken

I think what we would see in your example is:

1st – GoM
2nd – SF/DR
4th – FC/EB/DH
7th – AR
8th – KAI
9th – ET

note that nobody is in 4th, 5th or 6th places. if you come in 7th place, you get the 7th place rewards, even if all the servers 1-6 tied for first.

-ken

it’s my understanding that WvW season 2 was originally going to be only 7 weeks, but then it got extended to 9 weeks.

the new PvE content has probably been planned for a long time, and ArenaNet probably doesn’t want to push it out 2 weeks because that would push other things out as well, which would push other things, and before you know it we’re celebrating Halloween in November.

-ken

if you are getting emails from ArenaNet informing you that somebody is trying to change your account password, or recover access to your account, all this means is that ArenaNet knows your new email address and is sending you an email to confirm that the request is legitimate.

this is as it should be.

it does mean that somebody is actively trying to get access to your account, though. be very very careful not to click on anything in these emails that looks even remotely like it might “allow” the request.

-ken

it’s done by points. 1-22-44-66-8-9 means:

1 server is in first place. they are ranked #1.

2 servers are tied for second place. they are both ranked #2.

2 servers are tied for fourth place. they are both ranked #4.

2 servers are tied for sixth place. they are both ranked #6.

1 server is in eighth place. they are ranked #8.

1 server is in ninth place. they are ranked #9.

in case of a tie at the end of the season, all servers tied for the same place would presumably get the same reward. so if 2 servers are tied for second, both servers would get the second-place rewards. we haven’t gotten specific confirmation from ArenaNet that this is how it will be but it seems sensible.

-ken

bob.8632:

Yes ken, the path to continuous improvement lies in the practice of shrugging ones shoulders and just chalking it up as a loss….Every job/company has bean counters, the problem here is: they acknowledge there is a problem, they claim they will fix it. They have fixed some of it, this is probably one of those areas where instead of trying to wax philosophical, you should probably just move on to another thread that concerns you friend.

there are many paths to continuous improvement, and not every path is an improvement in every way. sometimes improvements in one area make things worse in other areas, and sometimes pursuing an improvement that is regarded as “more important” means abandoning improvements in “less important” areas.

as you’ve pointed out, they have fixed some of it, which is evidence that they made an effort — they did not simply shrug their shoulders. it is unfair for you to characterize abandoning the problem now as “just chalking it up as a loss” and walking away.

-ken

tiebreakers are necessary during the season because every server needs to be put into a specific matchup every week.

but for the end of season rewards, why does it matter? if 2 servers are tied for first place, give them both the first place rewards, then give the third place rewards to the #3 server. since no server came in second place, nobody would get the second place rewards in this example.

-ken

at this point, even if every remaining person who was still affected by this quit the game, it would probably cost them less money in terms of lost gem sales and lost reputation than it would cost them to actually fix it (including the opportunity cost of time spent fixing this means time not spent on something else).

this is simply something you have to accept in online gaming. sometimes, your problem simply won’t get fixed, not because you don’t deserve it but simply because time taken to work on this is time not spent on more important issues.

-ken

typically the green team is the strongest team, the team that can roll in and take other team’s stuff if they want to.

your best option is to get your server to go and take the objectives you need for map completion.

failing that, wait until the tournament is over and we are back to having randomized matchups. then there will be some color variance, unless your server’s rating is so high that it’s guaranteed to always get green (in which case your server is strong enough to just take the objectives you need, see above).

-ken

you can’t even report hackers anymore now that they’ve figured out how to become invisible.

-ken

good point.

also, in order to get some of the ascended jewelry items, you will need laurels. if you don’t want to PvE to get them, make it a point to do the daily achievements every day (and the monthlies too, if you can manage it). I have over 400 laurels banked at this point from consistently doing dailies and monthlies (I don’t do dungeons or fractals).

-ken

if you have a lot of karma banked from when you played before (or you have a lot of WvW badges from achievements) you can also buy exotic gear that’s good for WvW.

the specific stats you would want depend on your class, your personal playstyle, and also to some extent your server’s playstyle (some servers favor zergs with Berserker gear, others favor Soldier’s / Invader’s instead).

in general, I’d say you can’t go wrong starting with Soldier/Invader (power toughness vitality) then make adjustments from there. you can buy this gear in your home borderland using badges, or you can buy it for many of your equipment slots (but not all, iirc) in Orr using karma (but to do this you need to have particular temples unlocked).

-ken

the numbers that are posted in this thread:
https://forum-en.gw2archive.eu/forum/wuv/wuv/Matchup-wishlist

are calculated using a program that was posted a long time ago in the GW2 API forum.

I will post the current version of that program (containing updated randomization parameters) at some point.

right now we are in the middle of tournament play, so there’s no randomization involved. as a result, I haven’t been posting updates to the wishlist thread. when season 2 is over I will likely resume posting probabilities.

-ken

why can’t we all get along? this is why we can’t have nice things.

:D

-ken

if we were smart, we would have rigged all the matches so that everyone ends the season with the same number of points, so we’re all tied for first place. then everybody would get the first place rewards.

-ken

24 leagues! (27 in EU)!

first prize for everyone!

-ken

the thing about hacks is, they can never be entirely eliminated. you put in security measures, they figure out a way to bypass them. you put in countermeasures or mitigations to block the bypass, and they figure out a new bypass or a way to counter your countermeasures.

it’s not a question of writing “secure code”. there is no such thing as “secure code” when your code is executing on a hacker’s own computer, which they can control and inspect. encryption doesn’t work if your code needs to be decrypted before execution (the hacker just waits until the code is already decrypted and running, then extracts it out of memory).

when your software runs on my computer, it becomes exploitable even if it has no security flaws at all, because I can always add my own code that inserts exploits into a program that previously had none.

the fact that you are letting me run your code on my cpu is itself the most fundamental security flaw, but there’s not much we can do about that using today’s technology, so hacks will continue to be a problem for now.

-ken