Guild Wars 2

here are the estimated matchup probababilities for the upcoming matchups. “na1” and “eu1” files show the probabilities of getting any single opponent, “na2” and “eu2” files show the probabilities of getting a particular pair of opponents. the last two files are the estimated rating outcomes for the current matchups, assuming that final scores don’t meaningfully differ from today’s scores.

-ken

here are the estimated matchup probababilities for the upcoming matchups. “na1” and “eu1” files show the probabilities of getting any single opponent, “na2” and “eu2” files show the probabilities of getting a particular pair of opponents. the last two files are the estimated rating outcomes for the current matchups, assuming that final scores don’t meaningfully differ from today’s scores.

-ken

here are the estimated matchup probababilities for the upcoming matchups. “na1” and “eu1” files show the probabilities of getting any single opponent, “na2” and “eu2” files show the probabilities of getting a particular pair of opponents. the last two files are the estimated rating outcomes for the current matchups, assuming that final scores don’t meaningfully differ from today’s scores.

-ken

here are the estimated matchup probababilities for the upcoming matchups. “na1” and “eu1” files show the probabilities of getting any single opponent, “na2” and “eu2” files show the probabilities of getting a particular pair of opponents. the last two files are the estimated rating outcomes for the current matchups, assuming that final scores don’t meaningfully differ from today’s scores.

-ken

here are the estimated matchup probababilities for the upcoming matchups. “na1” and “eu1” files show the probabilities of getting any single opponent, “na2” and “eu2” files show the probabilities of getting a particular pair of opponents. the last two files are the estimated rating outcomes for the current matchups, assuming that final scores don’t meaningfully differ from today’s scores.

-ken

here are the estimated matchup probababilities for the upcoming matchups. “na1” and “eu1” files show the probabilities of getting any single opponent, “na2” and “eu2” files show the probabilities of getting a particular pair of opponents. the last two files are the estimated rating outcomes for the current matchups, assuming that final scores don’t meaningfully differ from today’s scores.

-ken

here are the estimated matchup probababilities for the upcoming matchups. “na1” and “eu1” files show the probabilities of getting any single opponent, “na2” and “eu2” files show the probabilities of getting a particular pair of opponents. the last two files are the estimated rating outcomes for the current matchups, assuming that final scores don’t meaningfully differ from today’s scores.

-ken

here are the estimated matchup probababilities for the upcoming matchups. “na1” and “eu1” files show the probabilities of getting any single opponent, “na2” and “eu2” files show the probabilities of getting a particular pair of opponents. the last two files are the estimated rating outcomes for the current matchups, assuming that final scores don’t meaningfully differ from today’s scores.

-ken

here are the estimated matchup probababilities for the upcoming matchups. “na1” and “eu1” files show the probabilities of getting any single opponent, “na2” and “eu2” files show the probabilities of getting a particular pair of opponents. the last two files are the estimated rating outcomes for the current matchups, assuming that final scores don’t meaningfully differ from today’s scores.

-ken

here are the estimated matchup probababilities for the upcoming matchups. “na1” and “eu1” files show the probabilities of getting any single opponent, “na2” and “eu2” files show the probabilities of getting a particular pair of opponents. the last two files are the estimated rating outcomes for the current matchups, assuming that final scores don’t meaningfully differ from today’s scores.

-ken

here are the estimated matchup probababilities for the upcoming matchups. “na1” and “eu1” files show the probabilities of getting any single opponent, “na2” and “eu2” files show the probabilities of getting a particular pair of opponents. the last two files are the estimated rating outcomes for the current matchups, assuming that final scores don’t meaningfully differ from today’s scores.

-ken

here are the estimated matchup probababilities for the upcoming matchups. “na1” and “eu1” files show the probabilities of getting any single opponent, “na2” and “eu2” files show the probabilities of getting a particular pair of opponents. the last two files are the estimated rating outcomes for the current matchups, assuming that final scores don’t meaningfully differ from today’s scores.

-ken

here are the estimated matchup probababilities for the upcoming matchups. “na1” and “eu1” files show the probabilities of getting any single opponent, “na2” and “eu2” files show the probabilities of getting a particular pair of opponents. the last two files are the estimated rating outcomes for the current matchups, assuming that final scores don’t meaningfully differ from today’s scores.

due to the reset time change, starting next week I will make these posts on Thursdays.

-ken

here are the estimated matchup probababilities for the upcoming matchups. “na1” and “eu1” files show the probabilities of getting any single opponent, “na2” and “eu2” files show the probabilities of getting a particular pair of opponents. the last two files are the estimated rating outcomes for the current matchups, assuming that final scores don’t meaningfully differ from today’s scores.

-ken

here are the estimated matchup probababilities for the upcoming matchups. “na1” and “eu1” files show the probabilities of getting any single opponent, “na2” and “eu2” files show the probabilities of getting a particular pair of opponents. the last two files are the estimated rating outcomes for the current matchups, assuming that final scores don’t meaningfully differ from today’s scores.

-ken

here are the estimated matchup probababilities for the upcoming matchups. “na1” and “eu1” files show the probabilities of getting any single opponent, “na2” and “eu2” files show the probabilities of getting a particular pair of opponents. the last two files are the estimated rating outcomes for the current matchups, assuming that final scores don’t meaningfully differ from today’s scores.

-ken

here are the estimated matchup probababilities for the upcoming matchups. “na1” and “eu1” files show the probabilities of getting any single opponent, “na2” and “eu2” files show the probabilities of getting a particular pair of opponents. the last two files are the estimated rating outcomes for the current matchups, assuming that final scores don’t meaningfully differ from today’s scores.

-ken

They don’t apply this reset, don’t we get the server pairs this reset?
They also said Glicko will be reseted to 1500 for every server/pair of server, so every match should be equally likely.

Anet hasn’t said they won’t apply. Also they said only volatility and deviation are being reset, not the rating itself. Most likely scenario is match-making will occur as currently done then the partial reset will take place.

In the release notes: “Since we’re significantly shaking up existing team membership, we’ll be resetting matchmaking data.”
I am not aware of any other source talking about it.

It’s a partial reset — volatility and deviation only, according to Tyler on reddit. It’s not a total reset of everything. I asked this question, and that was the answer I received. I could probably dig up the source, but it’ll just tell you what I did :P

I haven’t played in ages and I don’t really keep up with the forums either so I wasn’t aware of the server linking until today.

Yes, last week’s numbers were no good for 2 reasons: server linking (which affects the number of groups being randomized) and the deviation/volatility reset (which affects the amount of randomization applied).

This week we know the new deviation/volatility values, and we know the server groupings/linkings, but I need to figure out how scoring is done in the new system. I suspect that it might be enough to simply drop all the ‘guest’ servers from the data and run the calculations for the ‘host’ servers only so I’ll see about doing that this week and we’ll see what we get.

-ken

OK, here are the probabilities for the next matchups (assuming the method I’m using is still valid, which I don’t know for sure yet). If you are in a pairing as a ‘guest’ world then look up the ‘host’ you’re paired with to see who you are likely to play against.

The key files to look at this week are the “na” and “eu” files. If I am accurately predicting new rankings and ratings then everything else is probably good too. If the new rankings and ratings are seriously out of whack then the probabilities are not to be trusted.

Well, it looks like my predictions for rating and deviation are not accurate, which means the probabilities aren’t accurate either.

I don’t know if I can fix this; there may not be enough information available to me to identify the issue. As an example, look at NSP:

I predicted their rating would go up from 1447.1173 to 1467.7891 and their deviation would go down from 260.0455 to 214.2828. But the GW2 leaderboards say that NSP actually went to 1490.1838 / 188.2237.

This means that Glicko-2 is zeroing in on what it thinks the “true” ratings are much faster than I am (the rating is moving faster, and the deviation is shrinking faster). The speed at which Glicko-2 does this is determined by the Volatility value from the leaderboards and an internal constant called Tau. In 2012 Habib Loew stated that ArenaNet was using a value of 0.6 for Tau. It is possible that ArenaNet is using a different Tau value now, but without knowing what it is I can’t do the calculations.

Until this is resolved I am going to stop posting probabilities. It might be possible to figure things out by trial and error but the time I have available to work on this is very limited.

-ken

NSP/SF and DH/FC got the rawest deals with this pair beta. According to the updated leaderboards NSP shouldn’t even be a host this next week and DH is in the same situition NSP was two weeks ago, only DH is now listed as Full. Unless Anet reworks the pairs, if we get another T4 match with the current pairings, no matter who gets green is guaranteed glicko gain.

HazyDaisy.4107:

NSP/SF and DH/FC got the rawest deals with this pair beta. According to the updated leaderboards NSP shouldn’t even be a host this next week and DH is in the same situition NSP was two weeks ago, only DH is now listed as Full. Unless Anet reworks the pairs, if we get another T4 match with the current pairings, no matter who gets green is guaranteed glicko gain.

+10
The linked servers work well together, but the population balance has a lot to be desired. Green has had the advantage (overwhelmingly) in that category since the linking. Starting to believe “Outnumbered” is a base skill. xD

I suppose there is another possibility. If the final matchup scores were very different (in terms of percentages) from what they were on Thursday, it is possible that would have caused the predicted ratings to be different from the actual ratings.

But it looks like every server in the NA list has the wrong rating (except Blackgate, which is off by only about 1 point). It seemed unlikely to me that all the matchups would have seen this kind of last-day change. But I suppose it’s possible that there has been a surge of activity/interest in WvW right down to the last day due to the recent changes.

-ken

Ok, I’ve been watching the ratings and scores for a few weeks, and I have concluded that there’s nothing wrong with my math.

For one matchup only (April 22 – April 29) it looks like ArenaNet used a rating calculation very very different from the standard Glicko-2. I was not able to figure out where the numbers came from, even after making adjustments to the Glicko-2 Tau parameter. It’s entirely possible that the ArenaNet calculated ratings are outright wrong, based on bad data or buggy code.

But wrong or not, (or more to the point, explainable or not) ArenaNet’s decision is final so those are the ratings we ended up with on April 29. And after that, rating changes appear to track closely to what would be expected based on the scores, using the old Glicko-2 algorithm we’ve had for years.

So it would not have been possible for me to predict probabilities for the April 29 matchups, but it looks like I can do so for any matchups before or after. So I’m resuming the posting of probabilities.

With that, let the cut-and-paste resume:

here are the estimated matchup probababilities for the upcoming matchups. “na1” and “eu1” files show the probabilities of getting any single opponent, “na2” and “eu2” files show the probabilities of getting a particular pair of opponents. the last two files are the estimated rating outcomes for the current matchups, assuming that final scores don’t meaningfully differ from today’s scores.

p.s. Sorry about the late post, my normal plan is to post these numbers on Thursdays.

here are the estimated matchup probababilities for the upcoming matchups. “na1” and “eu1” files show the probabilities of getting any single opponent, “na2” and “eu2” files show the probabilities of getting a particular pair of opponents. the last two files are the estimated rating outcomes for the current matchups, assuming that final scores don’t meaningfully differ from today’s scores.

here are the estimated matchup probababilities for the upcoming matchups. “na1” and “eu1” files show the probabilities of getting any single opponent, “na2” and “eu2” files show the probabilities of getting a particular pair of opponents. the last two files are the estimated rating outcomes for the current matchups, assuming that final scores don’t meaningfully differ from today’s scores.

here are the estimated matchup probababilities for the upcoming matchups. “na1” and “eu1” files show the probabilities of getting any single opponent, “na2” and “eu2” files show the probabilities of getting a particular pair of opponents. the last two files are the estimated rating outcomes for the current matchups, assuming that final scores don’t meaningfully differ from today’s scores.

here are the estimated matchup probababilities for the upcoming matchups. “na1” and “eu1” files show the probabilities of getting any single opponent, “na2” and “eu2” files show the probabilities of getting a particular pair of opponents. the last two files are the estimated rating outcomes for the current matchups, assuming that final scores don’t meaningfully differ from today’s scores.

here are the estimated matchup probababilities for the upcoming matchups. “na1” and “eu1” files show the probabilities of getting any single opponent, “na2” and “eu2” files show the probabilities of getting a particular pair of opponents. the last two files are the estimated rating outcomes for the current matchups, assuming that final scores don’t meaningfully differ from today’s scores.

no numbers last week because I forgot. even if I had remembered the numbers would have been worthless due to the world linking changes.

I see that matchup stagnation / locked tiers are back in NA with a vengeance:

The following servers are now 100% likely to play each other and nobody else:
Crystal Desert, Kaineng, Eredon Terrace, Borlis Pass
Sorrow’s Furnace, Ferguson’s Crossing, Gate of Madness
Darkhaven, Ehmry Bay

the rating gap between CD and SBI is over 300 points, so it’s possible this situation will persist for a long time.

they did say their deviation and something else was changing to make the matchups less stagnant. I never saw any actual hard numbers though (and doubt it is enough to cover a 300 pt gap).

I wish for Blobbing week of Desolation – Far Shiverpeaks – Piken!!, even though piken will ppt like crazyyyyy

Would like to have 1 week away from gandara, even though theyre fun to face

Why just not use website
http://wvwmatchups.azurewebsites.net/#Desolation
?

here are the estimated matchup probababilities for the upcoming matchups. “na1” and “eu1” files show the probabilities of getting any single opponent, “na2” and “eu2” files show the probabilities of getting a particular pair of opponents. the last two files are the estimated rating outcomes for the current matchups, assuming that final scores don’t meaningfully differ from today’s scores.

here are the estimated matchup probababilities for the upcoming matchups. “na1” and “eu1” files show the probabilities of getting any single opponent, “na2” and “eu2” files show the probabilities of getting a particular pair of opponents. the last two files are the estimated rating outcomes for the current matchups, assuming that final scores don’t meaningfully differ from today’s scores.

kandle.2367:

HazyDaisy.4107:

NSP/SF and DH/FC got the rawest deals with this pair beta. According to the updated leaderboards NSP shouldn’t even be a host this next week and DH is in the same situition NSP was two weeks ago, only DH is now listed as Full. Unless Anet reworks the pairs, if we get another T4 match with the current pairings, no matter who gets green is guaranteed glicko gain.

+10
The linked servers work well together, but the population balance has a lot to be desired. Green has had the advantage (overwhelmingly) in that category since the linking. Starting to believe “Outnumbered” is a base skill. xD

Population balance is impossible to achieve with the linking system, any stacked server is balanced against any other stacked servers that is what link tries, but being based on population is to volatile, guild changes server, people quit etc….

(EDIT) NSP will not be a host… how can a empty server be in a host position?
Stack all empty servers in one??

here are the estimated matchup probababilities for the upcoming matchups. “na1” and “eu1” files show the probabilities of getting any single opponent, “na2” and “eu2” files show the probabilities of getting a particular pair of opponents. the last two files are the estimated rating outcomes for the current matchups, assuming that final scores don’t meaningfully differ from today’s scores.

There is a possibility that ArenaNet will manually tweak things starting July 29 to make certain impossible matchups possible:

https://forum-en.gw2archive.eu/forum/game/wuv/Glicko-making-it-impossible-for-CD-to-move-T3/6244179

But as a side effect of this, calculated matchup probabilities will be wrong unless the calculations take those same adjustments into account. If this can’t be done I may have to stop posting numbers.

We’ll see how things go.

-ken

I haven’t seen any confirmation from ArenaNet that they are definitely going to be adjusting the matchmaking, or what the adjustments will be. Despite the silence, I suspect they are going to do it anyway (advance communication not being one of their strengths) but just in case they don’t here are the numbers assuming no adjustment.

Since it’s unlikely that the EU servers will get any adjustments (they have no locked tiers) these numbers are probably accurate for them.

Here are the probabilities for NA, assuming the following matchmaking adjustments:

CD +185
SF +130
DH +100

Note that ArenaNet said they “might” make these adjustments, and that they were based on July 14 ratings. Since today’s ratings are different they might choose to make different adjustments for the upcoming matchups, in which case these numbers won’t be valid.

Aeolus.3615:

(EDIT) NSP will not be a host… how can a empty server be in a host position?
Stack all empty servers in one??

Well, Darkhaven is host again and it is clearly a empty server. And knowing Arena Net’s sense of humor, we’ll be Host AGAIN and our next pairing will be the dying Yak’s Bend.

Lets hope Anet does more than +185. Still 70% chance of the same old match.

Johje Holan.4607:

Lets hope Anet does more than +185. Still 70% chance of the same old match.

Looking at how they described on their old post and the current position, 30% is MORE than what we would get by the time he did his post. Back then 185 would make CD be 80 rating behind the lowest T3, while 185 now will make CD be 60 rating behind the lowest T3. The chance was even lower before.

They are adding 200 to each server pairing in T4.

Snowreap.5174:

I haven’t seen any confirmation from ArenaNet that they are definitely going to be adjusting the matchmaking, or what the adjustments will be.

There is a post today:

https://forum-en.gw2archive.eu/forum/game/wuv/Glicko-Rating-Manual-Adjustments-7-29/first#post6263343

Here are the estimated matchup probababilities for the upcoming EU matchups. I will wait a while before trying to calculate probabilities for NA matchups, in case ArenaNet posts an update regarding NA matchmaking rating adjustments. As always, the “eu1” file lists the likelihoods of getting a particular server as one of your opponents, and the “eu2” file lists the likelihoods of getting particular pairs of servers as your opponents. The “eu” file contains the estimated ratings used to generate the probabilities.