Relink algorithm?

[Evenge.4067]

Why does it seem so hard to place teams in correct tiers to start relinks? Last week, in Tier 3 NA, we were against a red team that was so overwhelmingly stacked and dominant. Why weren't they put in Tier 1 to start the relink?

The running joke in WvW is that a dartboard is used to randomly set relink matches. Seems no use of any previous month's numbers. No algorithm? It can't be that hard, yet it's so obviously broken to regular WvW players. 

Many of us love this game mode, but it is so poorly managed. Constantly frustrating. 

 

[Bizgurk.5639]

We know that there's an algorithm creating the teams, but do we have any information about how each team is placed into its starting tier? I don't remember seeing any kind of information so far. All we know is that is was done manually before the world restructuring. Perhaps that's still the case?

[Dean Calaway.9718]

16 hours ago, Evenge.4067 said:

The running joke in WvW is that a dartboard is used to randomly set relink matches.

That's correct. The worse part is they can't even hit the board most times.

16 hours ago, Evenge.4067 said:

It can't be that hard

In all seriousness it is really hard, that's a job Anet couldn't pay me enough to do.

You're literally trying to balance up to over 50 v 50 v 50 players per map on matches that run 24/7...

The only way you'll ever get any resemblance of a shoulder to shoulder match is if you start buffing the side with fewer players, and I mean perma boons, barriers all that kinda stuff.

Alternatively you can dynamically cap the maps to the lowest side +3, this way you wouldn't end up with 60 v 4 v 11 maps and you'd disensentivise the gang up tactic of always trying to get on the side with the most players.

[disForm.2837]

We have squat in terms of information, just some bits here and there from streams and posts. 

The *intent* was using hours played, timezones and open tags to balance teams. We have no clue which one, if any of those criteria were implemented. 

They dropped some interesting bit about using Alliance averages hours  for the balance. Very little else was provided. 

Balance WvW is not an easy problem. But after months and months of implementation, multiple beta tests, weeks of live action, they still seem incapable of actually balancing WvW. At this point, a company with the size of Anet should hire an specialist to crunch the numbers. 

[urd.8306]

You know something has to be off when you look at this on a sunday (1,5 days after reset)
 

https://imgur.com/aQ4xmQ2.jpg

[TheGrimm.5624]

3 hours ago, disForm.2837 said:

We have squat in terms of information, just some bits here and there from streams and posts. 

The *intent* was using hours played, timezones and open tags to balance teams. We have no clue which one, if any of those criteria were implemented. 

They dropped some interesting bit about using Alliance averages hours  for the balance. Very little else was provided. 

Balance WvW is not an easy problem. But after months and months of implementation, multiple beta tests, weeks of live action, they still seem incapable of actually balancing WvW. At this point, a company with the size of Anet should hire an specialist to crunch the numbers. 

What you mention is more sorting logic. The OP was talking about initial placement which has been an issue since server linking started way before the WR. It also was a potential issue while testing single week WR tests since if the newly formed servers were badly placed the matches could be all the more out of balance. 

[Dinas Dragonbane.2978]

The server I am on now started in t4 na and throughout that match had a total kill count of the other two combined. At least we have time to get to t1 for the final week! We also could in actuality be a t3 server this restructure though, we will see next week as this week we are already going strong again.

[Chaba.5410]

When server links were introduced, it was stated that servers would keep their Glicko rating while the volatility and deviation was reset.

https://forum-en.gw2archive.eu/forum/game/wuv/World-Linking-Beta

In 2017, Anet introduced 1-up, 1-down.

Here's from patch notes:
"If there is a change in which worlds are being used for matchmaking purposes (such as when world links change), matchmaking falls back on the usual Glicko rating sort (plus a potential deviation-based shuffle). Even while 1 up / 1 down is active, Glicko data will continue to be updated as it has been when each week’s round ends. This way it is always current when it needs to be relied upon."

https://wiki.guildwars2.com/wiki/Game_updates/2017-06-20

Over time with all the relinks that have been done and some of the linked servers becoming hosts then becoming links again, the Glicko ratings on servers got close to each other as has been seen on the leaderboards page.  So we've had relink resets in the past where BG, an NA T1 server, rolled a T4 match.

Since then, there has been no word that I recall of any change to that system.  And now with WR, there's also been nothing stated.  I don't think changing how glicko is used in the initial matchmaking was ever part of the scope of the work they were doing for WR.

So we are left with two potential states:
1 - When teams are formed, the glicko rating of the underlying server is still used.
2 - Something new is being used that has never been communicated to us.

How, might you ask, then did the NA T5 and EU T6 teams get match made if it's still Glicko?  Anet manually set those team's ratings.  They've manually adjusted ratings before on servers, especially in the year leading up to the 1-up, 1-down change.  They used to have a post where they'd announce the manual adjustments on the old forum.  Maybe they manually set all the team ratings today and let the deviation-based randomizer take over?  To me, this is the most likely scenario.

Hope this info helps.

Edited August 5 by Chaba.5410

[RisingDawn.5796]

The system is currently too slow after new team creations, just like after new World linkings, which also had problems controlling transfers.

You have mirror of Lyssa like servers, this time bava nisos from the team creations, which takes weeks from tier 4 not tier 6, to steamroll most of the other servers up to T1.

https://ibb.co/fqnrDK7

Edited August 5 by RisingDawn.5796

[Tula.6021]

In EU we have that power imbalance at reset too. An obvious t1 team got set up in t4 wiping the floor with the other two parties, now in t3 and doing the same there.

While in t1 it was more chill than in t4 . If not for Kazo it'd even have been quite boring in t1.

What's that reconstruction good for if not adjusting the tiers for reset better?

Edited August 5 by Tula.6021

[Dawdler.8521]

2 hours ago, Tula.6021 said:

What's that reconstruction good for if not adjusting the tiers for reset better?

Of course tiers have to be random on a shuffle that’s how matchmaking from the beginning works. Random is the only way that is fair.

You don’t just assume a team is going to win, they are going to have to go through the steps up or down like anyone else. Just saying they are the winners from the start would be ridiculous in any game mode.

Imagine if you started a PvE raid and when porting in the final lootbox is there and now you can exit again. Because you would have beaten the boss anyway, you’re always the winner.

Not to mention if you define a team like “oh yeah this is totally going to be T1 top dog so let’s place them in T1!” then you by the same definition have to specify the T5 bottom for the shuffle.

Very fun to open the matchmaking page and see you’re already predefined as the looser when the shuffle have barely begun. That will surely make people care more.

[Cyninja.2954]

This is an issue which becomes more difficult to solve and my guess is there is currently no aspect in the algorithm which pre sorts shards into tiers and the algorithm merely fills a predefined shard amount to create a predefined amount of tiers (based around expected activity)..

This was already an issue with linking, especially since it would often result in bandwagon to a T4 or T5 server pair which had an empty or near empty link.

Possible solutions could include a post shard creation valuation of created shard based around parameters outside of play time. This comes rather close to valuing guilds/alliances, which would help in spreading those more equally to different shards.

The other approach here would be to set reshuffles back at 8 weeks or even 12 weeks, to allow for better 1up1down sorting. This obviously should only be done once the initial sorting algorithm has been improved (for which the 4 week cycles are good since they allow for more data and adjusting of the algorithm).

Edited August 5 by Cyninja.2954

[Bizgurk.5639]

1 hour ago, Cyninja.2954 said:

Possible solutions could include a post shard creation valuation of created shard based around parameters outside of play time. This comes rather close to valuing guilds/alliances, which would help in spreading those more equally to different shards.

I was thinking about the same issue. The one up one down system is too slow with bringing each team to the tiers they belong to. There needs to be something else that adjusts the teams that were created by the algorithm and then places them in their respective tier after. K/D could be such a parameter, but the D can be influenced by purposely running into the enemy's spawn. Just taking the kill count instead seems better in that case. However, since kills grant only a small portion of the total warscore, teams like my current one tend to go down instead of up, which feels counterintuitive. If the tiers were sorted depending on total kill count, then the warscore for kills needs to be increased further, else teams that focus solely on killing will go down all the way.

Another parameter could be something like total objective captured and/or defended, which gives some sort of impression how successful a team fares without taking in the kill count.

There's another suggestion of one of my guildies. Since teams are balanced depending on their activity, the tiers could be sorted based on the activity "while being in a squad". Teams with squads tend to be more successful, thus such teams should rather start in an upper tier according to him.

[Dawdler.8521]

1 hour ago, Bizgurk.5639 said:

There's another suggestion of one of my guildies. Since teams are balanced depending on their activity, the tiers could be sorted based on the activity "while being in a squad". Teams with squads tend to be more successful, thus such teams should rather start in an upper tier according to him.

So just buy a tag, run in your own private squad all the time to automatically get sorted to more successful T1 class teams? 🤔

 Tbh I find it it amusing that people suggest a complete segregation of tiers based on the “skill” of players (ie how many boonballs they can muster).

Starting to think nobody really wanted any balance to begin with. 

[Reztek.7805]

8 hours ago, RisingDawn.5796 said:

The system is currently too slow after new team creations, just like after new World linkings, which also had problems controlling transfers.

You have mirror of Lyssa like servers, this time bava nisos from the team creations, which takes weeks from tier 4 not tier 6, to steamroll most of the other servers up to T1.

https://ibb.co/fqnrDK7

To be fair, Lyssa also steamrolled in T1. And so will Bava Nisos.

The purpose of the system is to create teams that are equally strong, which means there isn't a big difference between tiers. Well, it's not really good at achieving this ... and that's why Lyssa, Bava and so on happen. If they started in T1, 1 team could be stuck with them for the whole time ...

 

Edited August 5 by Reztek.7805

[Mabi black.1824]

38 minutes ago, Dawdler.8521 said:

So just buy a tag, run in your own private squad all the time to automatically get sorted to more successful T1 class teams? 🤔

Depends on. if you are able to do it 1 or 2 hours a day I guess not if you are able to do it 6 or 8 hours a day I guess absolutely yes 🤔

[Troubadour.6397]

9 hours ago, RisingDawn.5796 said:

You have mirror of Lyssa like servers, this time bava nisos from the team creations, which takes weeks from tier 4 not tier 6, to steamroll most of the other servers up to T1

I heard this weekend from some friends who play on different teams.  I was told of two guilds who, over a week after team creation moved as an entire guild to another team, and in one case moved to an upper tier team already swamping their opponents. The algorithm is already kitten awful but now? Anet is providing band wagoning services even old server guilds inclined to do so couldn't manipulate.

An entire Guild moved for free to an upper Tier mid team matchup. Anet doesnt want it to work, they are clearly supporting elite guilds to do whatever they want.

Edited August 5 by Troubadour.6397

[Dawdler.8521]

1 hour ago, Mabi black.1824 said:

Depends on. if you are able to do it 1 or 2 hours a day I guess not if you are able to do it 6 or 8 hours a day I guess absolutely yes 🤔

Tag up, run into a wall on EBG for 6-8 hours a day, cheat the algorithm, score!

[MedievalThings.5417]

On 8/3/2024 at 12:13 PM, Evenge.4067 said:

Why does it seem so hard to place teams in correct tiers to start relinks? Last week, in Tier 3 NA, we were against a red team that was so overwhelmingly stacked and dominant. Why weren't they put in Tier 1 to start the relink?

The running joke in WvW is that a dartboard is used to randomly set relink matches. Seems no use of any previous month's numbers. No algorithm? It can't be that hard, yet it's so obviously broken to regular WvW players. 

Many of us love this game mode, but it is so poorly managed. Constantly frustrating. 

 

Not soo much a dartboard.  Don't ask me why, but EVERY beta, reshuffle, whatever you call it has put a superstacked server, and medium server, and a "look, we're outnumbered again" server in every single tier.  The algorithm isn't trying to match the superstacks against each other.  I'm not going to speculate why they programmed the thing to do this, I just don't care anymore.  

[Mabi black.1824]

6 hours ago, Dawdler.8521 said:

Tag up, run into a wall on EBG for 6-8 hours a day, cheat the algorithm, score!

well, I didn't mean cheating, I was talking about true and genuine content. I know that in WVW there is confusion between these things, but I was speaking in absolute good faith.

[Gop.8713]

Well the new system is supposed to be balanced, so there shouldn't be teams that 'belong' in t1 or t4, there should be a lot of close matches and movement up and down week to week. That's not how it's working out so far, but the only solution I can see is stretching the matches out for more than a month before reshuffling. I think two months would give teams time to find their place and some weeks to actually compete there before getting rebuilt . . .

[Evenge.4067]

17 minutes ago, Gop.8713 said:

Well the new system is supposed to be balanced, so there shouldn't be teams that 'belong' in t1 or t4, there should be a lot of close matches and movement up and down week to week. That's not how it's working out so far, but the only solution I can see is stretching the matches out for more than a month before reshuffling. I think two months would give teams time to find their place and some weeks to actually compete there before getting rebuilt . . .

Nah, I like the 1 month, but Anet should do the initial placement of alliances better to start the relinks. No low population (or skill) alliance should be fed to stacked alliances for the miserable few weeks until they sift down to similar play levels. 

There must be enough data from alliances playtime, K/D, sheer numbers,  to arrange matchups with comparable competition to start each reshuffle.

[Gop.8713]

I just don't see how that would be possible with the shards being reconstructed for every match. The 'dominant' shard from last month no longer exists . . .

[joneirikb.7506]

(Just picking a random quotes to work from)

Quote

place teams in correct tiers

Quote

The purpose of the system is to create teams that are equally strong,

The purpose of the WR system is to make servers of close to similar size/activity (hours/players). And then lets the tiers sort out how well each Team works and move up and down until they hopefully find a tier they fit in. That is part of how WvW is designed/structured.

It is not meant to sort players by skill, organisation, PPT or PPK etc. It's meant to build up a team of equal "opportunity", and then leave it up to players skill and organisation to decide how well it goes.

This would work better with a 2 month Team duration though, to let teams move to the tier they fit into in first month, and then get more even matches in the second month. I assume ANet is planning to move back to 2 months after the initial period of testing/stats and when most players gets put on the right teams, and learn how to use the guild system.

----

What a lot of people here are talking about is essentially to change how WvW works on a very core level, in order to change it into something else. Some examples:

• Handicap system

Any system that tries to adjust score or combat in any way is called a handicap system. These systems are made to help whoever is in last place feel like they have a chance (like in Mario Kart when you have a higher chance to get better items like the shell that can shoot down the leading player).

The reason that WvW doesn't use any kind of handicap system (nope, Outnumbered EFFECT is not a handicap, it's a warning system), is that it tries to uphold some respect toward "fair competition". It might be misplaced, considering how many other things in the game mode works against "fair competition". But if you start adding handicap systems to the game, then whatever little respect people still have for "fair competition" goes out of the window.

Not saying it wouldn't be possible or even a good thing. Just that if ANet adds a handicap system, they might as well throw the WvW mode as it is under the bus and remake it as something else entirely, without any competition.

----

• PVP Matchmaking

A lot of the talk in this topic is about matching Teams up against Teams. This is inherently not how WvW works, nor how Team-shuffling works (not even back in relinks, it randomized based on glicko scores, and then returned to 1up1down). What you're suggesting is basically PvP MatchMaking. That it sorts players (teams) after their PVP ranks, and set them up against similar ranked players.

In WvW terms this is what Tiers are. And moving up and down in Tiers is the Team adjusting its rank according to wins and losses. But this will reset each time we have a Team-Shuffle, obviously. And considering how the whole Team system works it has to be that way, ANet can't really predict how strong a new Team will be or not, as they can't predict how the combination of guilds and players and how well they'll organise or their separate skill levels.

Now again there could be ways to work around this, but the short version is again that it would have to throw the WvW design under the bus, and create something else, especially without the 24/7 format.

----

 

[Mabi black.1824]

3 hours ago, joneirikb.7506 said:

they might as well throw the WvW mode as it is under the bus and remake it as something else entirely, without any competition.

For those who have participated in this '' teams / server '' game mode putting their hinge, as you should expect right in the '' team / server '' we are already well beyond this point. Comparison and competition have not only passed under the bus, but you have taken the bus and thrown it off the bridge, so as not to leave any trace.