Guild Wars 2

Oozo.7856:

It should be easy enough for someone who knows the mathematics behind the matchup system to determine the probability of both events happening.

Hint: It’s going to be ridiculously small.

It certainly made me curious that we went to three leagues at around the same time it was announced that Devon would be playing on NSP. And now this?

I’m not a big believer in coincidence with extremely low probabilities involved.

I ran the numbers, and the probability of GoM+NSP+AR is pretty small, around 1.63% (about 1 chance in 60). that’s actually a pretty high probability.

NSP had 168 possible matchups ranging from HoD+DH (most likely, 2.26%) to JQ+SBI (least likely, 0.000001%). GoM+AR was #19 in the list (only 18 other matchups out of 168 were more likely).

the full numbers are attached.

-ken

they have already responded to this multiple times. unfortunately, because the forum search feature is essentially worthless, it’s hard to find these previous responses.

anyway, the answer is that they are aware of it, and they have plans to fix it so that the difficulty of this (and other) wvw achievement is more in-line with the difficulty of other achievements.

as always, they have not made any promises about a specific delivery date.

-ken

maybe it’s just a time of day thing, but I have to agree: YB isn’t very good at open field fighting compared to most other servers we’ve played. this doesn’t mean they never win, just that when they do we almost never feel like we got outplayed — we just got outnumbered.

from what I’ve seen, NSP is much better, although their population disadvantage is very apparent.

-ken

here are the estimated probabilities for the next matchup.

-ken

interestingly, the BP+IoJ+CD matchup is only possible because of the random matchups. the randomness is what allows #11, #12 and #13 to play each other.

without randomness, #11 and #12 would always play #10, and #13 would always play #14 and #15.

-ken

I’ll bring my DJ bot so you can all listen to my awesome playlists of broadway show tunes.

-ken

a video so good, you’ll want to watch it again. twice.

-ken

here are the estimated probabilities for the next matchup.

-ken

since in-game wvw is down, how does forum wvw work? is it like roleplaying?

“I go to the northeast camp, and kill 2 of the veteran guards with fire. Seeing an enemy approach, I back off and put on my wizard robe and hat.”

something like that?

it’s remarkable (hence, I’m remarking upon it) that this thread has surpassed the ‘Server Match up is terrible’ thread in length, and in much less time.

in other news, it looks like ArenaNet figured out a new way today to interfere with GvG. personally, I’d rather let the GvG’ers continue if this is the alternative.

-ken

rank might be a useful measure of a commander’s experience if rank were account-wide, but it’s not. I regularly follow commanders who have alts (some with commander tags, and some without). some of those alts will have very low rank.

there is no substitute for knowing who you’re following and what their playstyle (and command style) is. knowing what rank they are doesn’t help with this unless you know whether the person you’re following has alts (and what ranks their alts are). but if you know this much about them, you already probably know enough to be able to tell whether you should follow them or not.

-ken

here are updated probability estimates for the next round of matches.

-ken

your color is based on your ranking within your matchup. the #1 ranked server gets green, the #3 server gets red.

but your matchup ranking has a random factor added in which makes it hard to predict. statistically speaking, higher-ranked servers have a better chance of ending up green than lower-ranked servers.

if the colors were completely random, you would always have a 33.33% chance of getting any particular color. but your actual chance can vary depending on who you’re matched up against — if you go up against servers with exactly the same rating as you, it will stay at 33.33%, if you play servers with lower ratings your chance of getting green will go up, and if you play servers with higher ratings your chance of getting green will go down. how far up or down it goes depends on how far apart the ratings are.

-ken

arguably, WvW play is the only play that you need to worry about balancing, because WvW is the only game mode that doesn’t allow players from different servers to group together.

PvE has guesting so you can PvE with people from any server. and if you don’t want to guest, PvE supports ‘overflow’ maps.

I’ve heard that sPvP has something similar so that people from different servers can form premades.. and sPvP is instanced so population balance is never a problem.

but in WvW you are limited to playing for your own server only, and there are no overflow maps. since WvW is the only mode where server choice actually matters, then WvW population ought to be a key factor in incentivizing players to move.

incidentally, incenting players to move would be a good idea, but imagine for a moment lines in a movie theater where you had to pay to switch lines, and if you stood in a different line than your friends that meant you either wouldn’t be able to sit with them, or you would have to watch a completely different movie.

-ken

DevonCarver.5370:

To head it off at the pass, there have not been drastically declining WvW numbers, or even marginally declining numbers. WvW continues to be as strong as it has been over the past several months.

I am guessing that the numbers you are looking at are totals across all servers. I am willing to believe that large numbers of players are leaving some servers in order to play WvW on other servers (indeed, half of the still-active WvWers in my own guild have done this), but I find it hard to believe that the majority of players who perceive fewer players on their own server (myself included) are imagining things.

-ken

new week, new numbers.

-ken

there is a random factor involved, but it’s not totally random.

if the color choice was completely random, AR would have a 33.3% chance of getting green each week. but with ratings the way they are now, the current ‘random factor’ matchup algorithm gives AR only a 20.4% chance of getting green. also if you’re interested, the chance of getting blue is 28.7% and the chance of getting red is 50.9%.

if AR hasn’t been getting green, it’s just bad luck with the random numbers, not because the system doesn’t allow it.

-ken

oh look it’s Thursday already. time to post new numbers.

-ken

how to build 8 golems in one minute, in 5 easy steps:

1. get 40 people together, each of them with 10 supply. (400 supply total)
2. portal to a borderland map, and build 4 golems.
3. stack on Commander Siegerazer and wait for 1 minute countdown.
4. after countdown, all stacked players get 10 supply (400 supply total)
5. build 4 more golems

if you have 2 commanders in your home citadel, you can do this twice (be careful not to start both countdowns at once).

if your 40 players don’t have supply to start with, detour to some other borderland first, and start Siegerazer there to get everyone filled up. the portal back to your ‘target’ map and proceed to step 2.

-ken

well, our local support team was able to get me up and running again, so here are the estimated matchup probabilities for the next round.

-ken

my laptop died yesterday (CPU fan quit) so I may not be able to post matchup probabilities this week (last week I simply forgot, if you’re wondering).

if you’d like to be able to calculate these yourself, and you have a C# compiler (Visual C# Express 2005 or newer, or Visual Studio 2005 or newer) the source code for the program I use is posted in the API Development forum.

-ken

a lot of the time, it’s actually not hyperbole.

the problem is, players only play as hard as they have to to get the rewards they want. when you have a population advantage, the quality of gameplay from the average player goes down. against lower-population opponents, you simply don’t have to try as hard to win. if you have timezone coverage sufficient to carry you, the natural tendency is to let them.

EBay is just as prone to this problem as any other server — playing against BP and AR for so many weeks before matchup randomization went in was bad for us — we played worse because we didn’t have to work to win. likewise, playing against KN on their way up the ranks was great for us, because it showed us how much work we needed to do to improve.

we’ve seen what happens ourselves when it’s too easy, so we recognize it when we see it in others (or at least, when we see something that looks vaguely like it). I think part of the problem is that when we face servers with more population (or better coverage) their better players take a break, so we end up seeing proportionally a lot more of their scrubs and fairweathers. I know that I play a lot less when it’s an easy match for us (although I’m not claiming to be one of our better players).

I wouldn’t guess that JQ (or any T1 server) suffers from this problem, because they mostly just play each other and I imagine the competition for the #1 ranking is fierce. but in the lower tiers where servers regularly find themselves playing against much lower numbers, I would expect to see a lot more of this “lazy” style of play.

-ken

it’s rare, but you do find them occasionally. only in EBG, of course. I’ve never seen them in any of the borderlands maps.

but their frequency of occurrence seems to have dropped somewhat. I used to reliably find one or two nodes per day, but lately it seems like I only find one node every day or two.

-ken

A truly massively-multiplayer-capable backend can’t be patched together based on an upgraded engine originally designed for smaller fights. You have to build the engine with that kind of traffic in mind from the very start, and the switching infrastructure that supports it needs to be included as part of the design.

You don’t get meaningful performance increases by simply upgrading servers, unless you are a completely fail organization that let your servers get overloaded to begin with while you stood by doing nothing (and whatever ArenaNet’s failings may be, I’m confident that this at least isn’t one of them). The only reason you upgrade servers is to reduce the risks and/or costs of operating/maintaining them.

the current engine is either up to the task (in which case donations shouldn’t be necessary) or it is not (in which case donations would be useless).

-ken

DevonCarver.5370:

skeet.1390:

So do 4 leagues in NA. This eliminates the enormous population difference from server 1 to server 12 and 13 to 24.

If you really think the server that is ranked 12th at the start of this stands any chance against the top servers you are delusional.

I never said they did. However, I’d argue that they do in fact stand a chance against the 6th ranked server and the 5th ranked server. We are not trying to only ever match you up against your closest comparable server. That was what was happening with the system before and it was bad. If you are saying you don’t ever even want the chance to try and show your skill against a higher ranked server, then I don’t know what to say.

under the current matchup randomization system, #12 already has the opportunity to test themselves against #5 and #6. the server I’m on is ranked #11 this week, and we’ve gotten the chance under this system to play #6 and #5 (and #4 too) already. we’ve gotten to play #13, #14 and #15 too.

I can say with some certainty that the #12 server is going to have no chance to win against #5 and #6. or even #8 and #9, for that matter. unless they somehow get a lot of transfers, of course, but that seems unlikely — anyone contemplating a transfer is much better off transferring to #1, #2, #13 or #14 instead.

the only ‘good’ thing about being rank #11 and #12 I can think of is the likelihood that they will find themselves in the #13 and #14 spots for the second season, since it’s likely that ‘stacked’ #13 and #14 servers from season one will find themselves in the Gold league for season two.

-ken

Here are the estimated matchup probabilities for tomorrow, based on today’s scores.

-ken

sorry about the lateness this week (and forgetting last week altogether). it’s summertime and between work and vacation my mind has been on other things.

here are the estimated probabilities for the new matchups.

-ken

it will be a massive jumping puzzle, and when you get to the top there will be fixed siege units that can bombard hills, bay and garrison, which you can power up with gems. also there will be a trading post NPC there so you can buy more gems in case you run out.

-ken

all these posts about tower flipping, more zerging and less defending are the reason that giving points for objective captures is a bad idea.

giving points for upgrading objectives (and only for upgrading objectives) is still a good idea though. although if ArenaNet does this I hope they change the upgrade pricing so the burden of winning a match doesn’t fall on the few players who pay for the upgrades.

-ken

the matchup system isn’t the problem, the problem is server imbalance.

once you have server imbalance, all a matchup system can do is control which servers get bad matches. under the old system, the same servers got bad matches over and over. in the new system, bad matches are more spread out.

but somebody will always get a bad match unless a solution is found for server imbalance.

-ken

I think awarding points for capturing an objective is probably a bad idea. two servers could easily conspire to make a third server lose by trading objectives back and forth between them, racking up a lot of points.

But I think awarding points for completing upgrades is a good idea. If the only source of points were from upgrades, the path to victory would be:

1. capture objectives
2. hold objectives until all upgrades are complete
3. allow enemies to take objectives
4. immediately (before your enemies can build upgrades) go to step 1.

it’s still possible to ‘game’ this system, but it takes so much longer that a lot fewer people would have the patience for it. a very strong server would have two options: (a) get a lead then attempt to hold everything, preventing opponents from scoring any points, or (b) allow your opponents to take objectives in hopes that you can take them back and score more points. option (a) would be very boring for everybody. (b) would be more fun, but letting your opponents take stuff only so you can take it back is kind of lame.

and the biggest flaw here is that the last server to go to sleep has a big advantage, because they can capture everything and let their overnight crew upgrade it all getting essentially uncontested points. but the current system has that flaw too so at least this wouldn’t be any worse.

-ken

Clockradio’s example is an excellent reason why 1-up-1-down is a bad system.

suppose that JQ and TC are almost equal strength, and the best possible opponents for each other. in 1U1D (also known as WULD — winner up loser down) JQ and TC would never get to play each other. during the weeks that JQ is in T1, TC will be in T2. and the weeks that TC is in T1, JQ will be in T2.

the other reason that WULD is a bad system is because some servers will alternate every week between winning by a lot and losing by a lot — they will never get a competitive matchup where the outcome has any doubt.

if you want to know the details of how the random rolls contribute to the matchups, you can find the details here:
https://www.guildwars2.com/en/news/big-changes-coming-to-wvw-matchups/

-ken

for a better example, go back farther in the history, before FA and KN were in T2 (when IoJ, ET and HoD were all top-tier servers).

-ken

LoBo.1724:

Am I missing something completely vital here?

Why isn’t the top WvWvW rated, high population, super organized WvWvW guild, servers matched against each other?

It’s the issue of a handful of servers floating around every week totally wrecking the fun for the majority.
Let the top servers duke it out and let the other servers be at peace.

Wouldn’t that solve all problems?
I’m eager to hear the cop-out!

the simple answer is, WvW matches are done in groups of 3 servers, but there are 4 or 5 ‘top tier’ servers. so 3 of the 5 get to play each other, and they generally have a good matchup. the 1 or 2 ‘top tier’ servers that are left over get to play somebody else, and those matches are almost always blowouts.

under the old system, it was always the same server that got blown out (#6). under the new system, it varies, so a lot more people get to experience what the #6 server had to endure week after week after week.

the old matchup system only worked well when servers of similar strength could be found in groups that were exact multiples of three. otherwise you got blowouts. the new matchup system can’t (and doesn’t) fix this basic problem, all it does is spread the pain out so that specific servers don’t get stuck being beaten every week in an endless series of blowouts.

-ken

welcome to an EB vs. BP matchup, SBI! while you’ve played each of us before, this will be the first time many of you get to see us together. you’re in for a treat.

please keep in mind that EB+BP matchup threads are expected to run to 15+ pages of trash talking and trolling (we got tired of getting threads locked so we just make one big thread now) but since a lot of our most frequent posters have gotten banned from the forums we may need some extra help to reach that goal.

note that we do not use GIF / meme spam to artificially lengthen our threads. that is too easy, and the kind of thing you would find in a tier 1 matchup thread.

-ken

Estimated probabilities for the next matchup (August 2).

-ken

it’s too much load on the servers, mostly because every player can move around individually. combine that with lag and you end up with problems of people warping past you that you thought you were blocking (because you saw them get to the bottleneck after you, but they saw you get there just after they were already through) and everybody will be frustrated.

stationary blocks work fine. they don’t move and aren’t affected by lag (except that lag can cause them to appear later than you intended).

a “block” skill that had a windup time (to counteract lag) and that rooted you the entire time (so that your obstruction doesn’t move around) would probably work fine though. I’d definitely support that.

-ken

JakobGW.5730:

Ignore the score. Ignore which server took points when you were offline. Pick an objective and try to accomplish that objective competitively. Attempt clever siege placements. Even troll your enemies.

If you truly cared about the score, you’d be on one of the T1 stacked servers, but you’re not. So ignore the score and enjoy the fights. Trust me, you won’t find a higher server morale than when most of the folks who fight along side you have this kind of attitude.

you’re preaching to the choir — the people with this attitude are already playing WvW.

we’re trying to recruit new players to try WvW, players who are currently just doing PvE. some of these players do not have the right attitude, and they will quit as soon as things get difficult. we don’t mind losing them.

but some of them will stick around through thick and thin, and those are especially the ones we want. the trick is getting them to try WvW, and stick with it long enough to see for themselves why it’s the best thing this game has to offer. this is easier to do when we’re doing well.

it’s hard to get people to “try out” WvW when we’re severely outnumbered because that makes it difficult to see the fun wins among all the losses. when you’re new to WvW it’s often hard to see the subtleties — what looks like a resounding defeat to a newbie might be recognized by WvW veterans as a successful delaying action, with a small number of us delaying a much larger enemy force before wiping.

a newbie would never understand why I am so happy to waste silver buying a supply camp upgrade (and draining the supply to 0) just moments before an enemy zerg rolls in to take it.

anyway, recruitment is our problem, and there is no reason anyone else should have to worry about it.

-ken

I am on a middle-tier server with a community that I like. I would not encourage our guild to move, even if enticed with gems, etc.

now, if our whole server was considering moving to the same destination (or at least the WvW guilds on our server) then I would of course consider moving our guild to stay with them.

-ken

Terissimo.6852:

I hate new match up system

I understand why. under the old matchup system, you mostly played Vabbi and FoW (you played them 8 out of the last 9 weeks the old system was in place) and you came in first place every time you played them, generally by a very wide margin.

now with the new matchup system, you have only played Vabbi and FoW 3 times out of the last 9 weeks, and those were the only matches where you came in first place. and when you get beat by another server as badly as you have been beating Vabbi and FoW, you don’t like it.

so, under the old matchup system, you played 9 weeks with the following results:
#1: 8 matches
#2: 0 matches
#3: 1 match

under the new system, you played 9 weeks with the following results:
#1: 3 matches
#2: 2 matches
#3: 4 matches

I completely understand why you would hate the new matchup system. Apparently the new system makes it about equally likely for you to come in first place, second place or third place, depending on who you get matched up against. Arguably, these results have been much more “fair”.

But this has not been nearly as much fun as winning pretty much every single week, which is the outcome you had under the old system. You don’t want fair games, you want fun games, and that means you should get to beat other servers by a large margin, but you should never have to lose to other servers by a large margin.

In other words, you should be able to dish it out, but you should never have to take it.

If you could build an online game that allowed this to happen, I’m sure you would make a lot of money. everybody would want to play a competitive game where they get to win almost all the time.

-ken

this sounds like one of the rare cases where players moving to a higher rank server actually helps the game.

Vabbi and FoW need a third server of similar strength to play against, that they actually have a chance of winning against. if people leaving helps to make WSR that third server, it will make EU tier 9 a better place.

-ken

people who are looking for a new server to play on really need to specify what time zone they play in.

you may not need to move to a mid-tier server — depending on what time zone you play in, TC, DB, FA or SoS may work out just fine for you.

as a guess I’d say SoS may be your best option.

-ken

Jess.6710:

Ebay stay off of our God kitten teamspeak. Its lame and makes your whole server look like an kitten . Thank you.

who hates EBay enough to want to make our whole server look bad? probably not somebody on EBay. we all love our server; it’s the very best one.

maybe it’s someone on a server we’ve played in the past, who would have good reason to hate us. or maybe it’s somebody on SBI who wants your attention focused not on them.

or maybe it’s a random Internet sociopath who just wants to give you a hard time and doesn’t actually care what game you’re playing or who you’re playing it against, only that they have the power to affect you emotionally, because their real life is so sad and hopeless that they have no power over anything.

-ken

don’t expect to get full WvW map completion in one week. this week you’re the red team, so you should focus on getting all the map completion objectives on the red map (your home borderlands) and the red corner of EBG.

next week, hopefully you’ll get a different color so you can focus on those.

some weeks you go up against an opponent that’s strong enough to come in and take all your stuff so you can’t get map objectives on your own home borderlands or your corner of EBG. there are 3 options here: (1) get your objectives as soon as the match resets on Friday, before they have time to take your stuff; (2) wait for another week when you get a better matchup; (3) watch and wait for opportunities when your team takes stuff back (even if they only take it back temporarily).

-ken

you will have to post the times of day (or night) you play, because otherwise it may be very hard to find the right server for you.

the servers with thriving playerbases may also favor zerg play during the times of day you’re on. so maybe you’d want a server that’s thriving slightly less.

or, the server with the best small warfare play may only do that during a time of day when you can’t play.

in the meantime, stick it out where you are. AR may be dead right now but they’re still adjusting to the loss of [ALS] and other guilds. once they settle into their new ranking, people who are interested in the kind of gameplay you want may start playing again. AR has always had a reputation for favoring small group play, but a lot of AR are taking time off from WvW until their ranking stabilizes again.

-ken

as requested, here are estimated matchup probabilities using the assumed-current (1.0,40.0) parameters, except using the sum of 2 rolls between -0.5 and +0.5, rather than a single roll from -1.0 to +1.0.

the use of 2 rolls is a ‘cheap’ way to simulate a gaussian distribution of random values, which makes random values near zero more likely than values near +1.0 or -1.0.

-ken

It might be worth considering having different randomness parameters for EU and NA. the ratings distributions are different, so using the same (1.0,40.0) parameters for both regions is probably not optimal.

Here are some sample numbers showing the effect different random parameters would have on the matchup probabilities.

The first 2 files show matchup probabilities for NA and EU using predicted ratings for today (based on yesterday’s scores) and using the current (1.0,40.0) parameters that I assume ArenaNet is still using (because they haven’t said otherwise). These are exactly the same files I posted yesterday in the “Who would you like to face next week?” matchup thread, if you’re wondering.

The next 2 files show matchup probabilities using 50% less randomness — parameters of (0.5,20.0).

The last 2 files show matchup probabilities using 50% more randomness — parameters of (1.5,60.0).

-ken

Cliff.

not that I am in any way superstitious.

-ken

I think ArenaNet had a pretty clear idea of how the community feels about this system by page 7 or 8.

Unfortunately, there’s no easy fix for this. the problem isn’t how the matchups are chosen — the problem is that server populations vary so widely. And the players make it worse every week by leaving low-rank servers to transfer to higher rank servers.

Supposedly ArenaNet is working on something big for WvW but I have no idea what it is. Hopefully it’s a complete revamp to the scoring system, because I think the current scoring system contributes to the problem — players look at the scores, see that their server is way behind, decide that they cannot possibly catch up, and with no hope of winning they then decide to not play at all (the players who only play when they think their team can do well are often called “fair-weather players”). This makes things worse for the ones who are left playing, because it leaves them badly outnumbered. So naturally they choose to transfer to servers with less “fair-weathering”.

The problem is caused by the players, not by ArenaNet. Since the players haven’t done anything about it except make it worse, we must rely on ArenaNet to solve the problem for us. I have no idea how (or if) they’ll manage it.

-ken

Estimated probabilities for the next matchup (July 26).

-ken