The Purge [PURG] – Ehmry Bay
it’s like karma or skill points. if you spent all yours and need more, just level up some more.
it’s not like trait points where you’re capped at a specific number and choosing points in one thing prevents you from ever being able to put points into something else.
if you are worried about ArenaNet adding new abilities, keep some points in reserve. this also applies to laurels, badges, karma and skill points — always keep some in reserve in case ArenaNet adds new things to spend them on.
-ken
The Purge [PURG] – Ehmry Bay
squad chat isn’t subject to suppression, but it has significant disadvantages:
1. people have to actually join your squad to see it
2. joining a squad prevents people from seeing where other commanders are, which is often useful information (especially when you’re setting up a pincer or a flank, or someone is using their pin to show where enemies are on the map).
-ken
The Purge [PURG] – Ehmry Bay
there is some randomness involved, but the highest-rated server is more likely to get green, and the lowest-rated server is more likely to get red, especially if the ratings are far apart. if the ratings are very close, the chances of a color swap are much higher.
“more likely” is of course not the same thing as “always”, so sometimes you will see color swaps that are unlikely.
for example, look at NA tier 5 (SBI vs BP vs AR). this week, the colors came out “as expected”. SBI has the highest rating and got green, and AR with the lowest rating got red. but last week, these same 3 servers played but got different colors even though they had the same relative rankings. last week, BP got red and AR got blue (SBI was still green, though).
to see how rating differences can affect the probabilities, here are some examples:
on June 14, SBI had a 1473.6749 rating, compared to BP and AR with 1344.205 and 1332.0161. given those ratings, if SBI were to face BP and AR (which they did) they would have a 40.5% chance of being green, a 31.1% chance of being blue and a 28.4% chance of being red.
on June 21, SBI had a 1403.425 rating, compared to BP and AR with 1395.4209 and 1343.9324. these ratings were much closer together than they were on June 14, and as a result the probabilities changed so that SBI had a 36.3% chance of getting green, 32.3% for blue and 31.4% for red.
-ken
The Purge [PURG] – Ehmry Bay
all camp upgrades are “personnel” upgrades. personnel upgrades deduct the full amount of supply instantly, then take time to “build”.
towers and keeps have both “personnel” upgrades and “structural” upgrades. structural upgrades slowly consume supply as they are being built, and include things like mortars, cannons, walls and gates.
in general, the structural upgrades require more supply than can be stockpiled at once, so they have to consume supply over time. you can speed up the build rate by buying the “extra worker” personnel upgrade.
the personnel upgrades deduct all their supply at once. but on the plus side once they are started they cannot be stopped even if the enemy cuts off all your supply. the only way to prevent a personnel upgrade from completing is to take control of the objective (and even then, your enemy won’t get the supply back).
personnel upgrades are a great way to prevent your enemy from getting supply from an objective they’re about to capture. if you can’t prevent them from taking an objective, at the very least you can prevent them from getting any supply from it.
-ken
The Purge [PURG] – Ehmry Bay
I think that’s why people are suggesting a change — in tower defense games you are expected to get stomped eventually, the question is, can you hold out before time runs out?
Think plants vs zombies, and the big zerg servers are the zombies.
In GW2 the matches last a week. can we set up a system that feels like a tower defense game, where you start with a certain amount of territory and you “win” if you can avoid being completely pushed off the scoreboard by the end of the match?
The ideal here is for both sides to have fun. In some of the matchup threads you frequently see players from one server get on and boast that they are the best, because they eventually were able to take a fortified objective from their enemy. Obviously, they had fun doing this. But then you see posts from the other side saying “of course you did, you had us totally outnumbered and we expected it. but we farmed you for bags and badges for hours before you finally kicked us out of there”. Obviously they had fun too.
What we need is a game that’s always fun for both sides, no matter whether you’re gaining ground or losing it.
-ken
The Purge [PURG] – Ehmry Bay
(edited by Snowreap.5174)
The idea of starting with a handicap isn’t a bad one, and it has been suggested before. I’ve even commented in support of it. There’s an interesting thread on it and I suggest taking a look: https://forum-en.gw2archive.eu/forum/wuv/wuv/Weekly-Matchups-Changing-The-Mindset/
The handicap idea does have some disadvantages however (or if not exactly “disadvantages”, then “things it doesn’t fix as well as it could”). for example, if you start the week with a +100K handicap, how much ground can you afford to lose per day? if you are down to +60K by Monday, are you doing good or not? Is there a way to gain the benefits of a handicap system, while still giving people a good measure of their day-to-day progress?
I happen to think that cumulative rating change, as well as today’s rating change, would to a better job of telling someone whether they are on track, or if they need to push harder. But I won’t say that handicaps are a bad idea (especially if they are combined with ‘par’ scores), because both options would be an improvement over what we have now.
Personally, I don’t think the notions of “world vs. world” and “balanced” are compatible with each other, as long as players are freely able to choose what servers to play on. Fundamentally I think that if players can choose where to play, you can only achieve balance by imposing it — artificially constraining who can play (so that it won’t be world vs world but “a small part of your world” vs “a big part of theirs”, i.e. portions of one world must be excluded from participation).
Theoretically, you could get both by modifying the scoring system, but I have yet to think of a scoring system that is fun, not subject to “gaming” and not overly artificial. I think a scoring system that increases the “effort per additional point” as your PPT goes up could be workable, but I haven’t thought of an effective and practical way it could be done. In particular, I don’t know how much harder marginal point increases need to be in order to be both fun and effective — that might be the kind of thing you can only find out in a beta test.
The issue with “overly artificial” scoring, by the way, is that players will recognize it as a sham, and deliberately ignore it, which I think is a problem because all players need to be playing the same game by the same rules for any kind of scoring system to be effective (otherwise you might as well make it a sandbox and not keep score at all — let the players decide for themselves what they consider a “win”).
-ken
The Purge [PURG] – Ehmry Bay
I have never argued that it is not in ArenaNet’s interest to improve the matchup system.
What I have argued is that the current system is in fact an improvement over the previous matchup system, and that we should not return to the old one. I have also argued that WULD would not be a better system than the current one.
And, I have offered what advice I can to help people manage their expectations regarding rating and how it affects matchups under the current system. If you are the type of person who cares deeply about your server’s rating, I will tell you what you must do to make it go up, and what you must avoid to prevent it from going down.
If you want to know how the current ratings affect the likelihoods of getting certain matchups, I can help with that.
If you want to know what to do in order to get a better chance of facing a particular opponent (or lower your chance of getting another) I can tell you what rating changes need to happen in order to change those probabilities.
I will tell you what the current system requires you to do in order to achieve your desired outcome, even if doing those things would not be fun, because it’s up to you to decide for yourself how much that outcome is worth to you. If that outcome is so important to you that you are willing to put fun aside to achieve it, that’s your business.
but I do not have the power to implement a system that promotes more fun. Complaining that my posts don’t do so is pointless. You could as easily complain that my posts don’t solve world hunger, or address income inequality.
If you don’t like the current system, you are welcome to suggest a better one. I do not see how complaining about people who are trying to make the best of the current system helps anyone.
-ken
The Purge [PURG] – Ehmry Bay
This post reeks of somebody who considers “Ratings Purity” more important than “Having Fun” in a Game.
The Ratings, the Scoring System, the complete lack of any kind of Balancing or Handicapping Mechanics – these are all serious flaws in WvW.
I love WvW and have never really experienced anything else like it in a game I’ve played but ANet really needs to get their priorities in-line. Ratings and Rankings are perfect for things like sPvP but are statistically meaningless in WvW. There is way too much variation and too many non-controllable factors to produce “meaningful” results.
The focus needs to be on getting people to participate and have fun in great fights – not “maximizing Rating” or “statistical accuracy” or any of that kitten.
it’s not that I consider it important, it’s that I don’t see any reason for people to complain about it if it’s not important to them. if rating isn’t as important as having fun, and you don’t think it’s fun to camp enemy spawns, then don’t do it. but don’t expect the rating system to reward you for choosing fun over points.
this is the unfortunate part of the system. if you think an opponent is too far below you to be fun, and you don’t want to play them again, you need to ensure that your rating goes up. if you don’t spawn camp them and your rating ends up going down, you are going to get more of these mismatches.
likewise, if you think you are hopelessly outclassed, but you’ve been doing your best regardless and gaining rating as a result, then expect to see more of this kind of matchup. if you want to see less, you need to lose by more, so that you lose rating, so that the system will give you this matchup less often.
I think this is a fundamental problem of the rating system. the system as it exists does do one thing fairly well — it gives us reasonable matchups (but not perfect ones) and it gives us some variation in who we play (arguably too much variation sometimes). but the quality of the matchups depends on the quality of the ratings, because the ratings are what the system uses to determine which matchups are “good”.
if you ‘throw’ your ratings then you’re not going to end up getting the appropriate matchups. going easy on lower opponents (and scoring lower than expected) is going to give you lower opponents more often. fighting hard against higher opponents (and scoring better than expected) is going to give you higher opponents more often.
-ken
The Purge [PURG] – Ehmry Bay
server ratings do not measure how well a server could do if they tried really hard. the ratings measure how well a server actually does. why your server does that well, rather than doing worse or better, is immaterial.
if your server doesn’t have the fortitude to do more than win by “just enough” then your rating will reflect that. if you want the kind of high rating that says “we will stomp you into the ground”, then unfortunately you’re going to have to stomp people into the ground to get it.
it’s important to keep in mind that the purpose of the rating is to estimate what your average score will be if you play another server with a different rating. in most cases, score is a good indicator of “strength” but sometimes it measures other things like “pity”.
-ken
The Purge [PURG] – Ehmry Bay
yeah, I forgot about them.
and it looks like some SoS are unhappy about their matchup too. since the randomized matches started they’ve always gotten matched against a higher server — they’ve never had a chance to be top dog under the new system.
-ken
The Purge [PURG] – Ehmry Bay
that’s not entirely true. I don’t think we would have been very happy to get BP+AR again.
the #5 server may be more than we can handle, but I think we have a reasonable shot at #7. it’s looking like it’s going to be another great week. with each passing week, ArenaNet’s new randomized matchup system looks better and better.
I do feel bad for the servers that got screwed this round, though. it looks like all the NA servers got good matchups this week, but I think some of the EU servers are going to be complaining.
-ken
The Purge [PURG] – Ehmry Bay
there are no more tiers. just likely matchups, and unlikely ones.
and ratings, of course. and rankings. those are still there, and still desirable for the bragging rights.
-ken
The Purge [PURG] – Ehmry Bay
have you seen the GW2 Map Tile API?
https://forum-en.gw2archive.eu/forum/community/api/API-Documentation/2250616
-ken
The Purge [PURG] – Ehmry Bay
there’s no reason to suppose that ArenaNet is seeding their random number generator from the time of day. hardware entropy sources are commonly available now, and even built into many processors.
and even if they were, after 2 million trials (or 96 million 32-bit random integers) the time of day makes no discernable difference. when the time comes, ArenaNet is going to run just 1 trial, and whatever numbers come up they’re going to go with, regardless of how likely or unlikely that combination of numbers might have been.
that’s kind of the point of 2 million trials. you get a wide variety of results, and by counting how many of each you can figure out which results are likely and which ones aren’t.
also, the thing about having a 5.1% chance of getting TC+DB is, that means a 94.9% chance of getting something else. getting TC+DB would be like rolling a 20 on a 20-sided die. (or a 1, depending on whether you think TC+DB would be fun or not).
-ken
The Purge [PURG] – Ehmry Bay
in the Potential Matchups thread (https://forum-en.gw2archive.eu/forum/wuv/matchups/Potential-Matchups-6-21) there are some nice graphs posted that help explain why we are seeing so many mismatches in EU.
for the EU servers in the middle, they are getting very wide and flat spans of possible matchup rankings. for the NA servers in the middle, the possible matchup rankings aren’t as wide, and they usually have a clear peak.
what this means is, NA servers will have more of a tendency to get rankings closer to the expectation, while EU servers will get a lot more variation.
this may be because there is a lot more uncertainty about the ‘correctness’ of the EU ratings, as shown by the larger deviations they tend to have. it would be interesting to run simulations to see if EU would see similar spreads if they had lower deviations like the NA servers.
-ken
The Purge [PURG] – Ehmry Bay
current matchup probabilities:
9.976% EB vs BP
8.316% EB vs AR
32.816% BP vs AR
1.910% EB vs BP vs AR
Most likely opponents for EB:
28.524% Yak’s Bend
28.313% Kaineng
28.195% Crystal Desert
22.300% Maguuma
20.426% Sea of Sorrows
Most likely opponents for BP:
33.926% Stormbluff Isle
32.816% Anvil Rock
20.474% Darkhaven
19.044% Isle of Janthir
12.806% Gate of Madness
Most likely opponents for AR:
32.816% Borlis Pass
31.989% Stormbluff Isle
21.530% Darkhaven
20.281% Isle of Janthir
15.098% Gate of Madness
-ken
The Purge [PURG] – Ehmry Bay
these new graphs are very illuminating. they explain why we are seeing more complaints about lopsided matchups from EU servers, as compared to NA servers.
in NA, the likely rankings for each server are concentrated in smaller spans. in EU, the middle servers have very wide spreads of possible rankings, so they are more likely to find themselves all over the place in potential matchups.
-ken
The Purge [PURG] – Ehmry Bay
Thanks to SBI for zoom hacking at bay. Way to stay classy.
screenshot, please. I’m trying to identify the spots that can be hit without hacks (just using windowed mode) versus the ones that definitely require zoom hacks.
-ken
Or better yet email exploits@arena.net if you feel that it is a exploit. Perhaps if enough people do this they will either fix it eventually or come out and say it is not a exploit.
Or people will get sick of this sort of garbage and move to another game. Its complete fail on Anets part to not police their own game. It took them months to fix the wall jumping hacks.
the problem here is that some of the things that people claim to be zoom hacks are really just people using windowed mode, or multiple monitors. there are certain places where a very wide aspect ratio can let you see (and target) extra stuff. but there are other places where it doesn’t, and if you don’t know the difference it leads to useless reports that will do nothing.
when I see a screenshot or a video showing alleged zoom hacking, I try to go there myself and see if I can get the same result just using windowed mode. the major WvW guilds on EBay have said they’re against hacks, and it’s important for our own internal policing efforts to be aware of which spots involve hacks, and which ones don’t.
-ken
The Purge [PURG] – Ehmry Bay
I originally posted this in another thread, but it’s on-topic here and I think a lot of people didn’t see it in that other thread, so here it is again (it was actually two posts but I’ve combined them here into one):
________________
I’m not sure why there are so many complaints about the new matchups. they look pretty reasonable to me this week. not perfect, of course, but you will rarely get a ‘perfect’ week when there’s any randomness involved. and we definitely need some randomness — the old system had none and we had a lot of problems as a result.
Here are the ranks of the servers in each NA matchup this week:
T1: 1 2 3
T2: 5 7 11
T3: 4 6 8
T4: 9 10 12
T5: 13 14 15
T6: 16 17 20
T7: 18 19 22
T8: 21 23 24
look at that. 8 out of the top 9 servers are in the top 3 matchups. 9 out of the bottom 9 servers are in the bottom 3 matchups, and the middle servers tend to be in the middle matchups.
really the only NA servers that should be complaining are BG, SoR and JQ who are getting exactly the same matchup that the old stagnation-prone system kept giving them over and over (although I wouldn’t be suprised if they actually preferred the stagnation since for them, any match outside of tier 1 might be considered a demotion), and CD (who does have a very difficult match this week against TC and SoS).
and here are the EU matchups:
T1: 1 4 7
T2: 3 5 6
T3: 2 10 11
T4: 8 15 16
T5: 9 17 18
T6: 12 13 14
T7: 19 22 25
T8: 20 21 23
T9: 24 26 27
EU does have a little more room to complain. only 7 out of the top 9 servers got into one of the top 3 matchups. 9 out of the bottom 9 are in the bottom 3 matchups. the middle servers generally are in the middle.
on the surface, Vizunah Square vs Jade Sea vs Desolation looks a bit lopsided, but this is a T1 vs T2 vs T3 match, which is exactly the kind of match you’d expect to see every other week if we were using the Winner-Up-Loser-Down system that so many people are claiming would be better. so there’s little point complaining about that one.
Riverside and Abaddon’s Mouth will have their hands full this week playing Elona Reach (although if RS and AM are anything like Ehmry Bay they’ll be welcoming this opportunity rather than complaining about it). Drakkar Lake and Far Shiverpeaks will similarly have their hands full playing Piken Square this week, just like Fort Ranik and Underworld will versus Augury Rock. these three are all somewhat lopsided matches, but in all cases it’s 1 strong server versus 2 weaker ones, which is a much better outcome than 2 strong servers versus 1 weak opponent (although I can understand why people on FoW and Vabbi might actually prefer such a match, if only to see how it would play out for them).
I do happen to think that the randomness needs to be adjusted down just a bit, but really these matches don’t look that bad. there is some variety in who plays whom, but for the most part the higher servers are playing the higher servers and the lower servers are playing the lower servers.
I’m not certain what server you’re on, but if it’s one of FSP, AM or FR then the matchup system is working properly for you. It has put you up against a very strong opponent to see how you do, and you are losing rating against them. When your rating goes down, the likelihood of facing those opponents again goes down as well.
The exception is if you’re on AM. in the ER/RS/AM matchup, the big winner is RS who is proving that they are every bit as good as AM and that they shouldn’t have a lower rating than AM. and even though ER is winning (as expected), they aren’t winning by as much as expected so their rating is actually coming down. as a result, AM (and RS) are actually more likely to get this matchup again. if you’re on AM and you don’t want to face ER again you need to lose by more.
This is something the old system never did — it gives lower servers the opportunity more often to prove that they deserve to move up. but that same opportunity to prove that you deserve to move up can also prove that you deserve to stay exactly where you are.
If server strengths never changed once established, the old system could have worked. But server strengths do change over time, and in order to determine when a server is over-rated or under-rated it needs to play a lopsided match now and then, just to make sure that the actual outcomes match the expected outcomes.
It is bad luck to get 3 matches in a row where you’re severely outclassed. I do hope that you have better luck next matchup. on Ehmry Bay we got the worst possible roll for the first week of randomized matchups (we got the exact matchup that everyone was certain they did not want to get), but the next week we got a much better roll that matched us against very high opponents (we lost that match, but we gained rating and we had a lot of fun testing ourselves against a high-rated server). so I do understand how heartbreaking bad luck on the matchups can be.
-ken
The Purge [PURG] – Ehmry Bay
Thanks to SBI for zoom hacking at bay. Way to stay classy.
screenshot, please. I’m trying to identify the spots that can be hit without hacks (just using windowed mode) versus the ones that definitely require zoom hacks.
-ken
The Purge [PURG] – Ehmry Bay
So, I was able to run some ranking simulations myself. Yesterday, I was having trouble figuring out how to format the data so that I could easily import it into Excel to generate graphs, but today after a good night’s sleep, the solution was self-evident. I’ve attached the numeric results (I’m not going to make images to post; it’s too much work). these are based on 10 million trials each for EU and NA.
I’ve noticed is that I’m not getting nearly as many ‘notches’ in the data. In most cases the rank probabilities monotonically rise to a peak, then decline. The exceptions are:
EU Fort Ranik [FR] has a dip spanning ranks 17 through 19
EU Underworld has a dip spanning ranks 17 through 19
EU Gunnar’s Hold has a dip spanning ranks 17 through 19
EU Ruins of Surmia has a dip at rank 18
NA Sea of Sorrows has a dip at rank 8
NA Maguuma has a dip spanning ranks 8 and 9
NA Darkhaven has a dip spanning ranks 17 and 18
NA Isle of Janthir has a dip spanning ranks 17 and 18
In particular, I don’t show any dip at all for Kaineng at rank 10. Ehmry Bay is another example where my graph shows a significant difference from yours.
I don’t think our results should differ by this much.
-ken
Attachments:
The Purge [PURG] – Ehmry Bay
there is still some disparity in the numbers I’m trying to figure out.
also, I’ve noticed in many of the graphs a ‘notch’ where a specific central ranking is low but the ranks to both sides are high. the rank just to the right is often especially high. for example, on the graph for Kaineng, I’m trying to understand why rank 10 is so unlikely for them compared to rank 9 or (especially) rank 11. right now I suspect that this notch is an artifact caused by the fact that rankings aren’t independent — the ranking you get depends a great deal on what rankings other servers get, and the rankings that other servers get aren’t evenly distributed; they are skewed based on rating differences. but I won’t know for sure until I’m able to run these kinds of simulations myself, and see if I get the same results.
-ken
The Purge [PURG] – Ehmry Bay
there’s another reason why ArenaNet might not want to ban exploiters right away.
if you ban a couple of exploiters as soon as you see them using a new exploit, two things happen: those few exploiters get banned, and people who are considering using that new exploit will know not to use it because it’s easily detected.
when cause and effect are close together, people can figure out pretty easily which exploit is unsafe to use.
but if you let the exploiters continue for a while, they’ll start to think it’s safe and more exploiters will use it (both alone and in combination with other exploits). then when you hand out the bans, you hit a lot more targets, and the people using multiple exploits can’t be sure exactly which one got them banned. that makes people mistrust all of them.
when cause and effect are far apart, nobody is sure exactly which exploit is the one that they got banned for, so all of them become suspect.
-ken
The Purge [PURG] – Ehmry Bay
only 2 million rolls? I’ll do some runs with 100 million rolls and add my data to that thread.
-ken
That statistical difference between the two is insignificant.
Or were you actually being serious?
it was tongue in cheek. but I really did do runs with 100 million rolls (actually it was 600 million: I did 3 runs of 100 million each to get single-opponent, paired-opponent and color-sensitive probabilities, for both NA and EU).
2 million rolls appears to produce results with about one tenth of a percent accuracy (I don’t happen to know the confidence interval). I’m guessing that the extra 98 million rolls adds a couple of digits of additional accuracy.
since it only takes me a few minutes to do 100 million, and it gives slightly more accurate results, I went ahead and did it. the program I use to do the calculations is actually designed to handle runs of 10 billion trials or more (run overnight of course — that many trials can take hours to finish) so I consider runs between 1 and 100 million to be pretty small.
-ken
The Purge [PURG] – Ehmry Bay
It’s complicated.
You’re right, FA has a closer rating to you than TC does, and if you calculate the probabilities of getting a particular server as one of your two opponents, you will get FA more often than you get TC:
Sea of Sorrows
5.395263 Sanctum of Rall
5.851026 Blackgate
9.915445 Jade Quarry
17.306499 Tarnished Coast
19.826024 Dragonbrand
21.769310 Fort Aspenwood
23.766748 Maguuma
22.275408 Kaineng
22.277815 Yak’s Bend
21.650310 Crystal Desert
20.571471 Ehmry Bay
3.945356 Stormbluff Isle
3.635472 Borlis Pass
1.617123 Anvil Rock
but when you calculate the probabilities of getting particular pairs of servers, the numbers don’t always fall the way you expect. I only posted the top 3 matches above, but there are lots more:
Sea of Sorrows
5.151287 Tarnished Coast Dragonbrand
3.946896 Maguuma Kaineng
3.938119 Maguuma Yak’s Bend
3.789833 Yak’s Bend Ehmry Bay
3.786308 Kaineng Yak’s Bend
3.766360 Maguuma Crystal Desert
3.750176 Crystal Desert Ehmry Bay
3.743293 Kaineng Ehmry Bay
3.741479 Yak’s Bend Crystal Desert
3.711017 Kaineng Crystal Desert
3.484831 Maguuma Ehmry Bay
3.074847 Tarnished Coast Fort Aspenwood
2.677353 Jade Quarry Dragonbrand
2.475916 Fort Aspenwood Yak’s Bend
2.473737 Dragonbrand Fort Aspenwood
2.470704 Fort Aspenwood Kaineng
2.354762 Fort Aspenwood Crystal Desert
2.298803 Fort Aspenwood Maguuma
2.172299 Fort Aspenwood Ehmry Bay
2.032138 Jade Quarry Fort Aspenwood
2.011202 Jade Quarry Tarnished Coast
1.611680 Dragonbrand Maguuma
1.586965 Tarnished Coast Maguuma
1.383278 Dragonbrand Kaineng
1.379293 Blackgate Dragonbrand
1.374626 Dragonbrand Yak’s Bend
1.308281 Dragonbrand Crystal Desert
1.249175 Sanctum of Rall Dragonbrand
1.150110 Dragonbrand Ehmry Bay
1.077918 Blackgate Fort Aspenwood
(and there are well over 100 more after this 30, the full list is posted in that other thread).
The single most likely matchup for you is TC+DB, but notice that the top 30 doesn’t list TC very often (only 4 times, TC+DB, TC+FA, JQ+TC and TC+Mag). meanwhile, FA shows up a lot (9 times: TC+FA, FA+YB, DB+FA, FA+KN, FA+CD, FA+Mag, FA+EB, JQ+FA and BG+FA). individually, each of the FA matchups is less likely than TC+DB, but because there are more of them you’re actually more likely to get FA than you are to get TC (and the likelihood of getting both FA and TC is 3.07%).
-ken
The Purge [PURG] – Ehmry Bay
I was never able to get my calculated ratings to exactly match mos.millenium.org or yours either, but I assume it’s due to rounding errors (but when I was off by a lot, I knew something was wrong).
I think a big part of the problem is that ArenaNet doesn’t publish exact ratings and deviations; the numbers they publish at https://leaderboards.guildwars2.com/en/eu/wvw are all rounded to 4 decimal places and I think that rounding accounts for the differences we see. when I run trial calculations using adjusted values with 5 or 6 decimal places (all of which round to the same value ArenaNet publishes) I get very different outputs, so clearly those lost decimal places make a difference.
-ken
The Purge [PURG] – Ehmry Bay
Arborstone has a deviation that’s +20 over WSR’s, but their rating is +34 higher so that cancels out. Arborstone can’t get a match rating any lower than 981.401, which is still higher than 950.226.
-ken
The Purge [PURG] – Ehmry Bay
I’ll just explain the process.
I used numbers from millenium with my single match calculator (which shows the new deviation/volatility scores) to get the new ratings and deviation for the next week.
oh, http://xerol.org/gw2/what-if.html is yours? that’s an excellent tool and I used it to debug my own rating calculator (the detailed breakdown really helped me a lot). since your calculated ratings match mine (and we both match mos.millenium.org) I’m pretty sure we’re all doing that part right.
ArenaNet posted the new matchup algorithm here:
https://www.guildwars2.com/en/news/big-changes-coming-to-wvw-matchups/
In that post they say that the ‘random factor’ is 40.0 + 1.0 * deviation. those parameters (40.0, 1.0) are what I’m using since I haven’t seen a post from them saying they’ve changed them.
I plugged in the values you used (0.0, 1.0) into my program and I got very different results, so I think that probably explains some of the difference, but not all of it.
Fundamentally, the idea of using bit fields sounds fine to me and I don’t see any problem with that approach. however, I suspect you may have another minor error hiding somewhere in there. the situation with Vabbi and FoW in EU is a good example.
Vabbi and FoW both have very low ratings compared to the other EU servers. they are so low, in fact, that I don’t think there’s any possible way for Vabbi and FoW to avoid being in the same matchup together. In order for Vabbi to avoid facing FoW, FoW needs a really high roll and at least 2 servers above FoW need a really low roll. if 2 servers can come in ‘under’ FoW, then FoW will play in EU T8 and Vabbi will play those 2 other servers in EU T9.
If FoW rolls a +1, their match rating will be 690.284 + 1.0 x (40 + 1.0 × 219.942) = 950.226.
If Blacktide rolls a -1, their match rating will be 1104.519 – 1.0 x (40 + 1.0 × 231.990) = 832.529
If Whiteside Ridge rolls a -1, their match rating will be 1190.501 – 1.0 x (40 + 1.0 × 184.883) = 965.618
and there’s the problem. it’s possible for Blacktide to roll below FoW, but it’s not possible for Whiteside Ridge. 965.618 > 950.226. even using (0.0, 1.0) to determine matchups as you were doing before would yield:
FoW: 690.284 + 219.942 = 910.226
WsR: 1190.501 – 184.883 = 925.618
and 925.618 > 910.226, so again WsR cannot come in ‘under’ FoW.
what this tells me is that both Vabbi and FoW must always play in tier 9 — there is no way for either of them to ever get a tier 8 match. for FoW to get a tier 8 match, it must roll high enough, and 2 other servers must roll low enough, that those other 2 end up in tier 9 pushing FoW up to tier 8. with the current ratings, it’s possible for 1 server to undercut FoW, but it’s not possible for 2 servers to do so. and if FoW can’t get into tier 8, there’s certainly no way Vabbi can either.
since your results show FoW playing in tier 8 sometimes, I think you have an error hiding in there somewhere. I think your basic structure and reasoning are sound so it should simply be a matter of finding the error and fixing it — there should be no need to start from scratch (unless you want to).
incidentally, the program I use doesn’t use much memory, I simply keep a three-dimensional array with dimensions (24,24,24) for NA and (27,27,27) for EU and each array element (i,j,k) is a 64-bit integer that tallies up the number of matches where servers i, j and k played each other. this means that the number of runs is limited only by how long I’m willing to wait — 10 billion runs (or more) are reasonable if I don’t mind waiting overnight (2 million NA runs takes about 10 seconds). the penalty of doing it this way is that I don’t track rankings (Mag+DB+Kaineng playing in tier 3 would be recorded exactly the same way as Mag+DB+Kaineng playing in tier 4, making my method tier agnostic) so there’s no way for me to produce the graphs you can.
-ken
The Purge [PURG] – Ehmry Bay
(edited by Snowreap.5174)
I wasn’t kidding when I said I’d worry about this. Just to check, I did 4 runs of 2 million rolls each, and checked the top 3 matchups for SoS after each run. here are the results:
Sea of Sorrows
5.152700 Tarnished Coast Dragonbrand
3.954150 Maguuma Yak’s Bend
3.943900 Maguuma Kaineng
Sea of Sorrows
5.168800 Tarnished Coast Dragonbrand
3.955300 Maguuma Kaineng
3.924700 Maguuma Yak’s Bend
Sea of Sorrows
5.144200 Tarnished Coast Dragonbrand
3.947200 Maguuma Yak’s Bend
3.934800 Maguuma Kaineng
Sea of Sorrows
5.149500 Tarnished Coast Dragonbrand
3.933400 Maguuma Yak’s Bend
3.924150 Maguuma Kaineng
if my code is wrong, it is at least consistently wrong (but I don’t think it’s wrong).
2 million rolls looks like it should be enough for accurate probability estimates down to about a tenth of a percent.
-ken
The Purge [PURG] – Ehmry Bay
ugh, I hate when results don’t match. 2 million rolls is certainly enough to get in the right ballpark; all an extra 98 million rolls should do is give you a couple more digits of accuracy. our numbers differ by enough that I don’t think the number of rolls can explain it.
maybe his results are wrong, or maybe mine are.
I don’t think my results are wrong.
but what if they are?
I’m going to be worried about this all day.
-ken
The Purge [PURG] – Ehmry Bay
Xerol, my numbers are very different from yours. would you mind posting the code you used to calculate base ratings and the code you used to calculate matchup ratings?
I’m thinking that either your code or my code (or possibly both) are wrong and I’d like to track down why so that we both produce similar results.
attached are the base ratings I used for my matchup calculations; they are predicted ratings based on live scores from a couple of hours ago.
-ken
Attachments:
The Purge [PURG] – Ehmry Bay
(edited by Snowreap.5174)
moar data! I simulated 100 million matchups for NA and EU based on the anticipated ratings that would result from the current scores.
for each region there are 3 files. #1 shows the probability of getting a particular server as one of your two opponents. #2 shows the probability of getting any particular pair of opponents (with different color assignments considered the same matchup). #3 shows the probability of getting any particular matchup (with different color assignments considered different matchups).
if you’d like to run these numbers yourself for larger or smaller sample sizes, the software used has been posted here:
https://forum-en.gw2archive.eu/forum/community/api/Simple-C-Example-Rating-Calculation/2154273
-ken
The Purge [PURG] – Ehmry Bay
only 2 million rolls? I’ll do some runs with 100 million rolls and add my data to that thread.
-ken
The Purge [PURG] – Ehmry Bay
Enlighten us how glicko works here.
http://www.guildwars2guru.com/news/884-the-math-behind-wvw-ratings/
http://www.glicko.net/glicko/glicko2.pdf
also, here’s the post that explains the new matchup system:
https://www.guildwars2.com/en/news/big-changes-coming-to-wvw-matchups/
-ken
The Purge [PURG] – Ehmry Bay
This isn’t true at all, and I demonstrated this earlier today. I can upload some screen shots of it right now if you want?
I totally stand corrected then. when I tried this myself the first time I never made the window any wider than it would have been with 3 widescreen monitors side-by-side (a 48:9 aspect ratio). and sized that way, you can’t usually see anything especially useful (a little bit extra, but not enough to explain some of the crazy things I’ve seen).
but sized down to the equivalent of 5, 7 or 9 monitors side-by-side, you can. once you get about 7 monitors wide, there’s pretty much nothing you can’t see, in certain spots. so unfortunately this doesn’t always have to be a hack. but there are certain spots you can’t see over no matter how wide the window (for example, you cannot shoot into the courtyard of the SW borderland tower from just north of the balcony behind the lord room without zoom hacks, no matter how wide your window).
so I still encourage people to try this themselves, and if they see behavior they can’t reproduce themselves without hacks, to report it.
-ken
The Purge [PURG] – Ehmry Bay
(edited by Snowreap.5174)
Can you get the higher zoom without hacking? For example, if you play on 3 monitors, do you get that level of zoom by default?
you can see exactly what would happen by pressing Alt-Enter to switch your game from fullscreen to windowed mode. then resize your window so it’s really wide but not very tall, as if you had 3 monitors. or heck, 5 or 7 monitors if that’s what you want.
if you do that, you will discover that it actually doesn’t help that much when you’re close to a structure. it widens your field of view so you can see more stuff at once without moving your camera, but it doesn’t actually make previously-invisible things visible.
in particular, making the window really really wide won’t let you see over a wall to let you shoot an arrow cart at something inside. doing that requires you to zoom your camera out far enough that the camera is well above the wall, which is what allows you to see over the top of it.
in order to see behind a wall, you need hacks / exploits — those can do things regular panoramic views can’t, and those are the things that need to be vigorously reported to arenanet any time you see them.
try the windowed mode thing yourself so you know what is possible that way, and what’s not possible. and anything that’s not possible, REPORT.
(the exception to seeing behind walls is if there’s any kind of hole that you can see through, such as the gap above a gate. don’t confuse people targeting through the gap above a gate as hackers).
-ken
The Purge [PURG] – Ehmry Bay
I should roll a thief then. I would be excellent at it.
-ken
The Purge [PURG] – Ehmry Bay
I think the answer to the question is, “yes, they do, but it shouldn’t require quite so much effort from the players as it does now.”
-ken
The Purge [PURG] – Ehmry Bay
I think Divinity’s Reach was originally intended to have a Canthan sector. now there’s just a big hole there.
-ken
The Purge [PURG] – Ehmry Bay
if Piken Square is overrated, or if your server is underrated, now is your chance to steal some points from a higher-rated server.
if Piken Square has the correct rating, and your server also has the correct rating, then a match like this is necessary from time to time to confirm that the size of the ratings gap between your two servers is still correct.
you shouldn’t have to go against them very often, but you should expect to fight them occasionally. the probability of PS/FSP/DL getting matched together is less than 1%.
-ken
The Purge [PURG] – Ehmry Bay
so I ran some numbers on June 13 to estimate probabilities of certain matchups, but I don’t happen to recall if I ever posted them. Here are the results for SoS:
Sea of Sorrows
7.190480 Blackgate
7.687542 Sanctum of Rall
11.719310 Jade Quarry
20.438576 Tarnished Coast
20.573732 Dragonbrand
21.865941 Fort Aspenwood
21.317536 Maguuma
21.405964 Kaineng
21.095310 Yak’s Bend
20.123159 Crystal Desert
15.505194 Ehmry Bay
7.253758 Stormbluff Isle
1.986013 Borlis Pass
1.327344 Anvil Rock
0.291216 Isle of Janthir
0.110536 Darkhaven
0.049415 Gate of Madness
0.034056 Sorrow’s Furnace
0.020476 Northern Shiverpeaks
0.002532 Henge of Denravi
0.001907 Devona’s Rest
0.000003 Ferguson’s Crossing
0.000000 Eredon Terrace
there are 7 servers that have around a 20% chance for SoS to get matched up with (the percentages all add up to 200% because every match each server gets matched up with two others). this week SoS got matched against two of them — this particular outcome was a fairly likely one.
meanwhile, here are the numbers for CD:
Crystal Desert
0.853664 Blackgate
1.039510 Sanctum of Rall
2.563388 Jade Quarry
9.905005 Tarnished Coast
9.924684 Dragonbrand
14.637574 Fort Aspenwood
20.123159 Sea of Sorrows
26.308808 Maguuma
27.015308 Kaineng
27.605955 Yak’s Bend
25.252554 Ehmry Bay
13.658515 Stormbluff Isle
7.240854 Borlis Pass
6.338689 Anvil Rock
3.294241 Isle of Janthir
1.915625 Darkhaven
1.057261 Gate of Madness
0.743427 Sorrow’s Furnace
0.433900 Northern Shiverpeaks
0.050544 Henge of Denravi
0.037258 Devona’s Rest
0.000072 Ferguson’s Crossing
0.000005 Eredon Terrace
CD had 4 servers with 25% or higher matchup probabilities. they got none of them, and instead got matched with a server with a 20% likelihood, and one with a 10% likelihood. definitely an unlucky matchup, but unlucky matchups are expected from time to time.
what is not expected is an unlucky matchup week after week after week.
you do raise an interesting point, however: if a server goes up against a stronger server and loses, would it make sense to have some kind of mechanism in place to ‘cap’ outlier matches so that they don’t play that server again (or any server rated higher) for some minimum cooldown period? in other words: periodically a server is given a chance to prove that they deserve to play in the big leagues. if they fail to prove this, they are temporarily sent back to the minors and don’t get another chance until some minimum waiting period elapses.
if there were 2 servers in each match, this kind of thing might be obviously a good idea. but with 3 servers in each match, I’m not sure, because that gives the third server the opportunity to play kingmaker. let’s suppose that CD does badly against TC, but mostly it’s because SoS keeps ruining every good plan CD comes up with. can CD argue that they should get another shot at TC, but this time without SoS there to interfere? would anybody on CD actually feel that way?
it doesn’t even matter whether SoS actually does anything to interfere or not. as long as anybody on CD feels like they did, would they think it fair that they get locked out of trying again?
at least with random numbers people can just blame bad luck.
-ken
The Purge [PURG] – Ehmry Bay
walls and siege are not intended to allow 1 defender to hold off a zerg indefinitely. they should be able to hold off a zerg long enough for you to get some defenders in. if you can get 10-15 people inside before a wall or gate goes down, you should be able to hold out for a long time, assuming you had a fair amount of siege pre-built in places where it can’t easily be aoe’d down from outside.
the primary purpose of a lone player guarding a tower isn’t actually to defend or upgrade it — your job is to watch for the enemy and call for reinforcements with a minimum of delay. when the enemy zerg is big enough, you can’t afford to wait 30 seconds for swords to appear on the map (assuming people are paying attention to the map to begin with).
the actual defense is the responsibility of your whole team, and paying for upgrades is too. your teammates should be swinging by as they roll around the map to help pay to start the next upgrade or drop another blueprint before they’re off to the next thing. and they need to be ready to come back at an instant’s notice.
enemy zergs are opportunities for your side to farm bags, and lots of them. the trick is figuring out where they are and getting the word out.
-ken
The Purge [PURG] – Ehmry Bay
and yet, if ArenaNet looks at their API traffic I’m sure they’ll see that people are querying that often anyway.
if the API included timestamps you could get away with querying once a minute (or less), if all you wanted to know is when something becomes capturable again.
in that case, the only reason to query more frequently would be to more rapidly report the fact that the objective has been flipped (because knowing that it has just flipped can often give you a good idea of where your enemies are right this instant — knowing where they were 20 seconds or a minute ago is usually a lot less useful).
arguably, if people are going to query that frequently anyway, then making the responses bigger to report more accurate countdowns would do nothing except increase API traffic. but it seems reasonable to suppose that there are people out there who would query less frequently if the API included timestamps, so why not let them do so?
-ken
The Purge [PURG] – Ehmry Bay
http://mos.millenium.org/na/matchups
I’m also beginning to think the 3 colors now don’t mean anything. Its not the server ranks that decide the color. It seems rather random. You get grouped with 2 other servers and then the game decides which color you are randomly?
I think you are correct. The first week of the new system AR, BP and EBay ended up together and although EBay was clearly ranked higher, BP was green.
yes. while there is a higher probability that the highest-ranked server in a matchup will get green, there is enough randomness that you certainly can’t rely on it. this is a good thing for people who have had trouble with map completion because their server never ends up being green.
I haven’t run numbers recently, but here’s an example from a run I did back on May 28. on that date, the probability of an SBI+BP+AR matchup for 5/31 was about 4.6125%. here are the various likelihoods for each color combination:
1.2096% g:SBI b:BP r:AR
1.0658% g:SBI b:AR r:BP
0.9192% g:BP b:SBI r:AR
0.5142% g:BP b:AR r:SBI
0.5723% g:AR b:SBI r:BP
0.3310% g:AR b:BP r:SBI
note that SBI being green was more likely due to their higher rating, but that there was still plenty of opportunity for them to get blue or red instead.
-ken
The Purge [PURG] – Ehmry Bay
loot bags are the great equalizer. when one side does too well, a bunch of their players have to back out of the fight to salvage stuff or do whatever to free up bag space, then run around and pick up all the bags they missed, before they can rejoin the fight.
if enough players have to back out (or waypoint away to find an npc to sell at) it can turn the tide of a battle. there are times I have suspected that our enemies are trying to do exactly this: to so overwhelm us with loot bags that we have to abandon our position to waypoint back to our citadel and sell our loot and buy more salvage kits.
-ken
The Purge [PURG] – Ehmry Bay
I disagree that a mismatch cannot be fun, or productive. look at NA week 23 (http://mos.millenium.org/na/matchups/history/77) when Ehmry Bay (EB) got to play Fort Aspenwood (FA). under the old “tier” system, that would have been a T2 server versus a T5 server.
sure, FA won the match. but EB played hard and we had a lot of fun. we even managed to get the #1 PPT a few times during the match (though of course we didn’t have the coverage to hold that PPT very long). more importantly, we got a big rating increase from that match even though we lost.
this is very much a game about coverage. to put it simply, at some point during the day most of your players are going to be asleep or at work. at that point, you are vulnerable if your opponent can get more players on.
but even if you have no night coverage at all, if you can fill your queues at prime time then you can be sure you won’t be outnumbered at that time, because at worst the enemy will have queues also (and at best they might be underpopulated at that time).
and if you don’t have the numbers to queue all maps, then just try to queue up one or two maps (like your home borderland, and eternal battlegrounds). some matches you just have to write off all the enemy borderlands as a loss, and concentrate on holding your own stuff while you’re on (or taking it back), and on not worrying about what happens while you sleep.
-ken
The Purge [PURG] – Ehmry Bay
here’s a made-up example to help explain one of the matchup problems. let’s suppose there are 6 servers, with the following ratings:
Server A 2400
Server B 2375
Server C 2350
Server D 2325
Server E 1025
Server F 1000
Servers A, B, C and D are all similar strength and having any of them play any other would be a good match. Likewise, E and F are similar strength and could easily play each other.
but in GW2, WvW matches are always played in groups of 3. so, no matter which 3 out of the top 4 servers you choose for one match, the one left over is going to end up playing servers E and F. it doesn’t matter how small the random factor — even a random factor of zero (the old matchup system) will not prevent E and F from getting a bad match. this is the problem FoW and Vabbi are currently having. no matter who you pick to play against them, that server is going to beat both of them. there are people who believe that FoW and Vabbi would actually have more fun if the randomness were increased high enough that, for example, FoW might face two stronger servers that mostly fight each other, giving FoW more opportunities to capture an undefended objective.
The random factor needs to be big enough to ensure that matchups have enough variation so that all servers within a rating ‘group’ have opportunities to play each other, and so that the ratings can accurately measure the size of the rating gaps between groups of servers.
If the random amount is zero, then every tier becomes its own ‘rating group’ and while the ratings can be an accurate measure of strength within that tier, they aren’t a good measure of strength across tiers (in particular, it is possible to have situations where the #1 server in Tier 5 is actually much stronger than the #3 server in Tier 4, but the rating gakitten o high that they never get the chance to prove it).
If you make the random amount small, you will have have opportunities for servers within a rating group to play each other, but they will never play anyone outside their group. this helps the situation when you have adjacent tiers with very similar ratings, but it doesn’t help servers that are trying to cross larger gaps.
To give servers the opportunity to cross those gaps, the matches need to happen sometimes. Let’s suppose that a whole bunch of players transfer from server D to server F (so many that server F becomes the strongest server in the game). in order for F to be able to prove their new strength, they need lots of opportunities to play stronger servers and prove that they can do well. if the randomness is small, the ratings might settle out this way:
Server A 2400
Server B 2375
Server C 2350
Server D 1325
Server E 1225
Server F 1800
but if the randomness isn’t enough to allow F to play against A, B or C they will never get the chance to prove that they actually deserve a rating of 2500.
this is a made-up example, of course, but the underlying point is the same: there are ‘natural’ rating gaps between groups of servers, and the only way for a server to cross these gaps if they become stronger or weaker is to play servers on the other side of the gap sometimes.
-ken
The Purge [PURG] – Ehmry Bay
(edited by Snowreap.5174)
I am not saying just because it has worked for us there are not teething problems with the new system. I DO think too wide a range of servers are meeting and they should chop off the outer extremities of the bell curve, limiting matchups to +/- 250 glicko. We have been lucky that none of our matchups have been too extreme. But I am liking the variety and I do think this system can offer a better way than the old, very static one.
random offsets are limited to deviation plus 40. I think there has been a persistent rumor that the offsets are plus or minus 400, but that’s not true.
last week the EU matchups were plus or minus 276 maximum (for Arborstone, the server with the highest deviation) down to plus or minus 212 minimum (for AG, the server with the lowest deviation). the mean was plus or minus 226.
16 servers had offsets in the 10-point range from 217 through 227. 5 were below 217, 3 fell between 227 and 250 and only 3 EU servers had random offsets higher than 250: Arborstone, Fissure of Woe and Vabbi.
-ken
The Purge [PURG] – Ehmry Bay
The primary goal of the new matchup system is to make the ratings more accurate, by ensuring that they are both valid measures of relative server strength within a single tier, and also valid across tiers. To put it another way, the old system ensured that your rating was accurate when compared to the servers you were already playing. But the ratings were not accurate measures of strength against servers you weren’t already playing. So we had cases of weak servers with high ratings who couldn’t move down, and strong servers with low ratings who couldn’t move up. Those servers might have been perfect matches for each other, but the old system never let us find out.
The reason the ratings need to be accurate is not to make matches balanced — this is not actually possible to do given the current population imbalances. The reason the ratings need to be accurate is, we want to have some matchup variability. We don’t want to be stuck playing the same servers week after week, we want to play somebody harder than usual or easier than usual once in a while, and we want to play more of our neighbors more often (so if you’re the #3 server you don’t want to face #1 and #2 over and over, you also want to play against #4 and #5 sometimes).
in order to determine how often these unusual matches should take place, the ratings need to be accurate. if all the servers had the same rating, you would see #1 playing #24 as often as they play #2. having accurate ratings allows the system to make matchups that are reasonably good, with reasonable variation, while avoiding horrible matchups that cannot possibly be fun for anybody.
yes, those servers that were in ‘perfect’ tiers where every matchup was balanced (but there was no variation) will actually see more unbalanced matchups — in some cases imbalance is the price you pay to get variation. but for most servers, the benefits of match variability will far outweigh the costs of imbalance because variability is what’s going to allow servers to move up and down the ranks more freely as they get stronger or weaker.
that is the big problem the old system had that the new system is trying to fix — it is a system that allows most of the matches to be reasonably good (but not perfect) while still allowing mismatches that provide an opportunity for an over-rated or under-rated server to prove that they deserve to move.
to do this, we need to be more flexible about what we consider “good” matches. under the old system, “good” matches were essentially “the best possible” matches. under the new system, “good” matches may sometimes have to be merely “not horrible”, because one of the things the new system does is spread the pain of bad matchups across all the servers, rather than just a few that were stuck perpetually in bad matchups every single week under the old system.
-ken
The Purge [PURG] – Ehmry Bay