Guild Wars 2

What´s the formula for the rating please? For example, Arborstone leads with 305k points, is there a chance that they move up in top tier and Riverside moves down?

Uder.9187:

What´s the formula for the rating please? For example, Arborstone leads with 305k points, is there a chance that they move up in top tier and Riverside moves down?

I’d be interested in seeing how this works too. It’s nice to just see the hard numbers sometimes.

The algorithms for the ratings are described in the linked Glicko 2 document, are they not?

^ well, yes, but it’s on a really long, really complicated PDF. we’d like a more “for dummies” explanation :P

A “for dummies” would be good, I thought Elo-like systems like Glicko were designed for one vs. one competitions like Chess, not sure how they work in a three-way competition like GW2’s WvWvW.

Thanks for this, Always knew it was based off of a win vs lose and how much of a score difference been trying to nail down my own formula was getting pretty good with a cross the board guess and getting a 3/3 lol.

How do you think we will ever get around a massive rating difference in a 3 x matchup though? I guess there isn’t much to work with given the limited field – but some matchup’s this week two of the servers are similar rated, but the other is way out in front and absolutely dominating the others – problem is, the next “tier” above they are better rated again – not much fun for the dominated servers..

Not sure of how to fix it though..
Maybe limit the max players for each server per map, instead of total for that map (will never get so one sided then?) the undermanned buff is nowhere near enough..

> How do you think we will ever get around a massive rating difference in a 3 x matchup though?

One good way of doing that is to not have the winning team rewarded with things that make them stronger. Guess what GW2 decided to do? rollseyes

WvW is flawed and the rating system, while a good way to balance things on a shallow level, will never result in a system that can reliably and consistently have the week long battles on a balanced nature, because the more a team gets, the EASIER it is for them to get more, and that’s 100% backwards.

not just images ^^

But thanks for telling us how you calculate things.
Any chance to get info´s about pairings?

Are the best servers only that good because they only played against the worst servers?

They mean nothing…we can server transfer and make the last server be the first, but we won’t do that. “We” like winning..so the first server will constantly be first while the last constantly last.

Positions mean nothing while server transfers makes the “random” factor overrule every other indicator by a large variable factor.

And thats´s why i really want additional info on:
– Population
– Percentage 1-10, 2-20 etc

This way you can see if a server suddenly has a lot more players than the week before…

A week? We are talking about days even. That’s how bad it is.

I agree that the current rating system is meaningless while players can freely change to any server they wish, as often as they wish.

The table rankings don’t mean much without 24-hour population coverage per server being taken into account.

It’s too late, all the wannabe-pro player guilds have “moved” to the dominating servers, so they even dominate more. And there is no way to balance.

The only thing I (or Anet) can hope for, is that they get bored with smashing people at their spawn, and guilds decide to move to other servers to try and increase competition, but I shouldn’t get my hopes up..

Well anyways…does anyone have any idea on how we could calculate some kind of “overall rating” ?

We only have the ratings for each week…how could we calculate some kind of permanent rating?

thought about something like
sum of all prior ratings / count of ratings

e.g
5 Weeks played
sum up all 5 ratings / 5 = total Rating?

The rating for the most recent week is the overal rating, not the rating for that week

Biggest problem that I see, is that very soon, the ratings will be such that there is never any movement on the table, and that week in week out, the matchups will be the same, with the same results each and every week.. basically.. stagnation. And when that happens, wvw will become a wasteland..

DaysOfNIght.5928:

The rating for the most recent week is the overal rating, not the rating for that week

Are you sure?

That means… we only have last week and this week to see which server improved his rating and which dont?
If a server wins 10 times in a row and improves his rating by far, and looses 2 matches afterwards …does it still have a high rating? We could only see that he lost rating but not that he gained very much prior.

Thought the ratings are new once a week…so that every week servers have a chance to get into a higher bracket etc.

The starting ratings were only set 2 weeks ago (based on the number of wins during the 24 hours matches period) thats why there are only 2 weeks of stats, week 3 should be out after the reset later.

Have a read on how the Glicko rating system works.

So I keep hearing this question come up, so I’ll be the one to ask. Is there any plans or talks about declaring a winner in Wv3? I know I’ve read before that sPvP will have a gold flag and a title for someone that comes in first in the sPvP season, but what about Wv3? I figure so, because there is a ranking system, but would like to get an official answer on this.

Don’t think the problem is in how you calculate the rating, the problem is how the score’s are gained.

The major flaw is giving the score for objectives when there is NO PvP, so this is the thing that need’s to get back on drawing board.

What’s happening on majority of servers now:

You have strong server’s that during prime time ( according to your own graphs you can see what prime time is ) steamroll’s both enemies as they really should not be in the same bracket’s with them.

So whats the result, the server’s that are loosign badly and i mean badly do not have any fun in WvW so after few days that start to give up on WvW. The server that is winning have really boring time as there is no challenge and owning majority of map again during prime time when most people are online is just meh …

So why does this happen ?
Because in most cases you will have few yes few .. check graphs again players running from 2:00am on and getting objectives and keep them during night to gain score for their server.

I dont think when you guys designed WvW you meant to award same amount of points if there is 0 player’s on map.

Balance wise the off time capper’s are screwing the majority of other players as they move them up the bracket where they will just loose badly.

So rather then checking and rechecking how to do rating .. repair the scoring.

Thanks a lot for this. Nice to see the math behind it.
In general I really like how you guys at Anet state a lot of things your working on with great amount of detail what exactly you’re doing and why, which almost no other mmo company ever provides to their playerbase.

Just for fun…..can someone tell me how to calculate the rating for NA vs EU?

ThisWeek
EU has an average Rating of 1500.814
NA has an average rating of 1499,304

thats by summing up the ratings divided through servercount.
Does that make sense?

It´s just for fun as we talked about that recently

IKShadow.4582:

Don’t think the problem is in how you calculate the rating, the problem is how the score’s are gained.

The major flaw is giving the score for objectives when there is NO PvP, so this is the thing that need’s to get back on drawing board.

What’s happening on majority of servers now:

You have strong server’s that during prime time ( according to your own graphs you can see what prime time is ) steamroll’s both enemies as they really should not be in the same bracket’s with them.

So whats the result, the server’s that are loosign badly and i mean badly do not have any fun in WvW so after few days that start to give up on WvW. The server that is winning have really boring time as there is no challenge and owning majority of map again during prime time when most people are online is just meh …

So why does this happen ?
Because in most cases you will have few yes few .. check graphs again players running from 2:00am on and getting objectives and keep them during night to gain score for their server.

I dont think when you guys designed WvW you meant to award same amount of points if there is 0 player’s on map.

Balance wise the off time capper’s are screwing the majority of other players as they move them up the bracket where they will just loose badly.

So rather then checking and rechecking how to do rating .. repair the scoring.

so true, right now our “cap time” on Seafarer’s Rest (fighting against Blacktide and Baruch Bay) is during prime time (7pm-1am) when we mosty have rating 400+ points and ONLY reason why BT is loosing only by ~45k points is because they have a lot of players playing ~7am who actualy play PvD (player versus door)
AN you shoud reconsider changes in points sytem because its no fun when we fight near BT and BB starting points

Oh well Anvil Rock did take a beating finally ahha!

Hey Devs. Your formula for sAGlicko score seems a bit arbitrary. It is clear that you want something that smoothly interpolates 0 to 1 with a punishment (reward) for very low (high) scores, but using the sine function seems to me a bad choice. A better choice would be to use a cumulative distribution function for a symmetric Beta distribution , not only does it give you more control (by adjusting the shape parameter), but it actually makes sense conceptually: If you assume that sA score and sB score follow a Gamma distribution with the same shape parameter (this is just saying that you want the computation to treat each team equally), then this formula would give you the probability that sA/(sB+sA) is better than random.

To me this makes a lot of sense because you would like to use sAGlicko score to measure how much better sA is than sB during that match.

Now that being said, it is still an assumption that over all shards scores follow a Gamma distribution, but this is not very difficult to determine once you get more data.

In fact, once you determine a good model for shards’ scores, you simply can produce the proper formula by computing the cdf of the random variable X/(X+X) where X is the random variable for scores

Sorry if this is a bit technical… I do get excited about maths
example plot:
http://www.wolframalpha.com/input/?i=cdf+beta+distribution+2%2C2

The biggest problem with the system is it provides a higher ranked server an easier time to stay higher ranked and a harder time for others to move up.

Now this is fine in an already large timeframe and datapool, but not now where we are in our system.

As an example and no offense at all to this server, but lets take SBI who has been in the top tier since the beginning. The first 2 weeks I believe they retained second and this week they are in 3rd, so really lost all 3 weeks, yet there is a chance, with this system, that even after the reset, they can retain that top tier, because of the tier favoritism in the system.
Now again I’m not picking on that server. They could drop to 4 and completely dominate their matchup, but the system calculations assume that they will do exactly that when it may not be the case at all.
The data from the 24hr matchups should have been used solely to determine the 1st week’s matchups and then should have been purged. that would have also allieviated the system as well
24hr match up vs 7 day matchup I think we can all agree are a totally different animal

Hmm…

mangarrage.1062:

As an example and no offense at all to this server, but lets take SBI who has been in the top tier since the beginning. The first 2 weeks I believe they retained second and this week they are in 3rd, so really lost all 3 weeks, yet there is a chance, with this system, that even after the reset, they can retain that top tier, because of the tier favoritism in the system.
Now again I’m not picking on that server. They could drop to 4 and completely dominate their matchup, but the system calculations assume that they will do exactly that when it may not be the case at all.

No it doesn’t, it just ranks all the players in linear order based on their Glicko score. It is possible to drop many ranks. Just because you haven’t seen it in the top tiers doesn’t mean it can’t happen.
Example: SoS goes -2 ranks, IoJ goes +2.

Moreover, there is nothing in the system that makes it easier or harder for higher or lower ranked teams to move up or down.

I think what you are feeling is that you would like the rounds to be shorter. But this does not affect the main point of your post which was that this system makes it easy for higher ranked teams to stay high.

So does this reflect who we’ll be matched up with next week?

I ask because I have no interest in facing the never ending zerg from Miller’s again this week. Enough of that nonsense.

Cool. Nice to see, even though I dont do a lot of WvW, that the server I’m on (Far Shiverpeaks) is ranked 2nd in the EU

that simply means that the pairings are
rankings 1-3, 4-6, 7-9 and so on?

This current cycle should end 10/5 (today) right? Or do the matches reset on Saturday now?

afaik is cycle end 23:59 on Friday

I’m a bit confused. Is this saying that the system has no memory? A number one server could be number 10 and last can move to first?

I’m explaining oddly.

Every week (two weeks when it’s working like that), the system looks at the scores on all the servers, puts them through this formula and spits out ranks. Using those ranks, the servers are trebled (see what I did there?) up in groups of three. 1-3 are together, 4-6, 15-18, etc. The system doesn’t care (has no memory) of the previous matchups, it’s based purely on *this* match?

So what’s to stop a high ranked server from taking a dive one match to completely dominate a lower ranked server? Not saying that the fine players of GW2 would stoop to such nefarious deeds, but that would be possible within the system, yes?

HabibLoew.2590:

DeeJayTC: Yes, that’s how we do the matchups.

Sir William PD: The cycle ends at 23:59 UTC today (Friday)

Ah that pesky old time zone thing!

Thanks for the clarification.

Yet again 2 in the freaking morning for EU servers, so the night kap krewes can come in and take over.

Is there a reason why NA and EU servers have to be re-matched at the same time?

I’m a bit confused. Is this saying that the system has no memory? A number one server could be number 10 and last can move to first?

I’m explaining oddly.

Every week (two weeks when it’s working like that), the system looks at the scores on all the servers, puts them through this formula and spits out ranks. Using those ranks, the servers are trebled (see what I did there?) up in groups of three. 1-3 are together, 4-6, 15-18, etc. The system doesn’t care (has no memory) of the previous matchups, it’s based purely on this match?

So what’s to stop a high ranked server from taking a dive one match to completely dominate a lower ranked server? Not saying that the fine players of GW2 would stoop to such nefarious deeds, but that would be possible within the system, yes?

The current match purely determines the change in the rating. The aggregate of all the previous matches is what the current rating is based on.

There is a natural limiting factor in how much servers can move up or down due to the fact that they are always going to be matched up servers at most 2 ranks away from them.

Even as we are seeing in some matches where a server is completely dominating, they are still only moving up usually 1 tier or at most two. You’ll never see a last to first or a first to last because the match ups aren’t set up that way.

For all the non-believers, Desolation just dropped 3 ranks, which is immense considering they had a big lead over the realm just under them. So yes, you can drop very fast from the rankings.

I have a question: Why did Gandara drop in rating? We won our week this time (and rather convincingly, too).

where are the new ratings??

The bottom of the first post!

Also, I noticed that the picture suggests Gandara would be matched up against the Underworld and Abaddon’s Mouth, when in fact we’re matched up against Baruch Bay (SP) and Fort Ranik (FR).

Cayafas.8290:

The bottom of the first post!

Also, I noticed that the picture suggests Gandara would be matched up against the Underworld and Abaddon’s Mouth, when in fact we’re matched up against Baruch Bay (SP) and Fort Ranik (FR).

That are the ratings from last week….

Cayafas.8290:

The bottom of the first post!

Also, I noticed that the picture suggests Gandara would be matched up against the Underworld and Abaddon’s Mouth, when in fact we’re matched up against Baruch Bay (SP) and Fort Ranik (FR).

No, the ratings have not been updated yet. They will be in a few minutes if its like last week.

They are? That’s rather confusing under a header that says “Update 10/5”

Thanks!

The ratings haven’t been updated at all, despite saying 10/5.