Proposal: 4-week relinking period and initial server placement after relink

[Dawdler.8521]

8 minutes ago, TheGrimm.5624 said:

Getting initial placement correct is a bit of a deal in a 4 or 8 week mix. Its hard to tell if some of the betas that people see as worked or failed were due to bad initial placements and might have been different if the initial placement worked. This is a similar issue to accounting for not only players, play hours but time of play is important and to some sense tags. Else you might end up with equal numbers on opposite sides of the clocks and the matches are way imbalanced in a given timezone. 

The initial placement can only be approximated, doesnt matter how many weeks it is. It have to be random by design if we assume a more random population distribution. I would even say that judging the initial placement based on any previous performance is anti-competetive. The entire point of a true reset is to start everyone from the same position. The tier system would initially be completely flat.

The problem is how you go beyond the first week when those tiers start taking shape.

[TheGrimm.5624]

12 hours ago, Riba.3271 said:

My point isn't to get rid of transfers, it is to allow people to transfer to a server that fits them and make it so that server doesn't get ruined by periodical changes in playerbase. Of course servers will still lose or grain players gradually, but that will be much slower than relinkings and reflect how entertaining they are. The best server for a player or a guild would stay best server for years, rather than 1 or 2 months.

Without weighting these, what creates imbalances:

• People leaving the game - can't be helped
• New players - can't be helped
• Transfers - is limited, could be helped by removing in the WR
• Dual Accounts - Can't be helped - I know this one may be questionable but if a server has a group of 40 that were there all last week and then gone the next week if can be impactful to a given week
• Seasonal players - can't be helped, similar in aspects of dual accounts

So right now we have one factor that can be helped. On paper the intent of the WR is to address all of them. 

12 hours ago, Riba.3271 said:

This is the power of solo servers. System we already had. Asking for it, but with approriate amount of servers for the playerbase, isn't "wishful" but completely reasonable. They just need to copy paste and adjust some code from the past.

Solo servers allows players to once again get used to who what and where and have a potential to coordinate. Doesn't mean it would happen. I don't think there is a way to measure good transfers from bad. We know some people will stack, and some people will appear to be helping by moving their group down to a lower tier where as they might also just be looking to farm players and bags. So not certain if transfers haven't been one of the biggest imbalancing factors we have had overtime personally.

[TheGrimm.5624]

6 hours ago, Gop.8713 said:

I like that the shorter relink period gives tanking servers less time to manipulate their way down into weaker tiers, but perhaps that's not as much of a problem on EU idk. Tbh it's been a while since I've been a part of a really unbalanced matchup in NA. I don't pay attn to matchups I'm not involved in, so perhaps I've just been lucky . . .

Based on a quick snapshot from GW2matchup NA I would say T1 & T4 are still open, T2 & T3 are already over for the week.

[TheGrimm.5624]

1 hour ago, gloflop.3510 said:

I would instead voice for a quick solution of the other problems by anet and the introduction of alliances as intended with world swapping to a limited extent and with longer link periods.

Not sure how they would be able to limit transfers and not certain without shorter links how the WR would be able to rebalance something that players may throw off after a sort is done. That's assuming people like the new matches & sorts.

[TheGrimm.5624]

14 minutes ago, XenesisII.1540 said:

Placement in the betas shouldn't matter, cause like Dawdler mentioned, ideally with the WR system worlds should be closed to "even", more so when they actually start using time zone in the sorting, so what should make the difference is the ppt work put in during the match. But the other problem is because the betas are temporary one week things, people are not going to care to "win" a match, much like the last week before a relink. And then there's the lopsided random ratio of ppt and ppk players in every world. So the betas will feel uneven, we'll only see the full effects of people wanting to win once WR is permanent.

In the current system, placement obviously matters because there's a huge range between populations and activity from T1-T5. But we're moving to WR within maybe the next 6 months. So again really, not much point to a fixed placement after a relink/WR, unless they move to a fix matchup system like a tournament, which they should and get rid of the free styling tier system. KD or KDR is still a bad statistic to base placements on as you'll just have people gaming the system as usual (guess what could happen on the last week of the season, servers already do it for better links), so placements should remain random, and it should potentially be better when WR is fully on. 🤷‍♂️

To Dayra's original point about what could be used for placement. I hope they are also weighing groups for deeper attributes. Groups may look the same on paper based on numbers but you can't measure some aspects of a group. Take as an example just using KD shows one thing for PPK but may not reflect the same picture for considering PPT. But to get a good sort you may want to weigh the group attributes first and then also weigh their effectiveness so that you sort the PPK and PPT groups evenly out versus end up with a server that has plenty of fighters that hold nothing and therefore drop to the bottom tiers with nobody to fight. I enjoy database design so been starting to run thru this is my head since the thread popped up since I was already thinking along those lines and what might or not be measurable factors they might be using. Course I am out of coffee so that means its time to leave Forum Wars 2 till later. See you in the mists!

[TheGrimm.5624]

10 hours ago, Dawdler.8521 said:

The initial placement can only be approximated, doesnt matter how many weeks it is. It have to be random by design if we assume a more random population distribution.

I am not sure I agree here, I think they are working on less random. I think they are looking at weighting on top of sorting. Which is a reason to apply the 4 weeks while still in server linking to test the math.

10 hours ago, Dawdler.8521 said:

I would even say that judging the initial placement based on any previous performance is anti-competetive. The entire point of a true reset is to start everyone from the same position. The tier system would initially be completely flat.

That's the issue that Dayra raised, is that there is no prior performance impactws after a resort which is the point of this conversation. And as many noted the first placements have been questionable and potentially impact the first 4 weeks of matches.

10 hours ago, Dawdler.8521 said:

The problem is how you go beyond the first week when those tiers start taking shape.

The answer here is fix the initial placement so we don't start a new sort, or currently a new relink, in so much of an out of place then they should be so we get better matches.

[Ironbars.2857]

A simple bandaid would be something like

1. Week 1 - Random placement

2. Week 2 - Every Tier winner from Week 1 placed in top 5 spots, 2nd place 6-10 spots, 3rd place 11-15 spots

3. Week 3 onwards, one up one down as usual

This does a better job of getting the strongest servers quickly to at least the right third of the Tier list after Week 1.

Edited December 8, 2023 by Ironbars.2857

[Dayra.7405]

Yeah, in lack of a good way to do an initial placement, faster sorting seems to be the solution. In the end

5 hours ago, Ironbars.2857 said:

1. Week 1 - Random placement

2. Week 2 - Every Tier winner from Week 1 placed in top 5 spots, 2nd place 6-10 spots, 3rd place 11-15 spots

3. Week 3 onwards, one up one down as usual

@Dawdler.8521's proposal of return (that's how WvW-matchmaking was started) to Glicko with glicko-score-reset on every relink (that was missing from the initial system, such that it created barriers between tiers after 9 month running) and my Swiss-Tournament (as we had it in WvW-Tournaments) proposal, will do more or less the same over the 4-5 weeks of a linking Period.

I think @Dawdler.8521's is the most elegant 🙂

 

Edited December 8, 2023 by Dayra.7405

[Dawdler.8521]

3 hours ago, TheGrimm.5624 said:

I am not sure I agree here, I think they are working on less random. I think they are looking at weighting on top of sorting. Which is a reason to apply the 4 weeks while still in server linking to test the math.

That's the issue that Dayra raised, is that there is no prior performance impactws after a resort which is the point of this conversation. And as many noted the first placements have been questionable and potentially impact the first 4 weeks of matches.

The answer here is fix the initial placement so we don't start a new sort, or currently a new relink, in so much of an out of place then they should be so we get better matches.

Not how that works. WR (reset, links, whatever) is overall balance of worlds against all other worlds. You cannot sort them by their “proper” tiers at that point because that would be stupid and go against the point of a reset. Hence it would be random. You can’t fix that unless it’s not a reset.

It’s the same as in any competition. You start everyone at the same point, you don’t put Usain Bolt 10m from the finishing line “because he’s so good he’ll win anyway”.

[joneirikb.7506]

Overall agreed with Xen, this is a temp/transition change. And I don't see them making any further changes on it until they release WR in whatever state, as WR should in theory fix these issues.

ANet starting to make fixes to things inbetween other things won't really benefit us, all it will do is give us more half-baked systems, that's being replaced by WR, which gets further delayed by the time needed to make half-baked fixes. At that point, it would be much easier/better/cost and time efficient, to just roll back to 8 weeks when they're done testing what they want for now.

Edited December 8, 2023 by joneirikb.7506

[gloflop.3510]

14 hours ago, TheGrimm.5624 said:

Not sure how they would be able to limit transfers and not certain without shorter links how the WR would be able to rebalance something that players may throw off after a sort is done. That's assuming people like the new matches & sorts.

It would already be interesting to limit the number of transfers per world to maybe 25 players. Then full guilds cannot transfer anymore but individual players.

[SweetPotato.7456]

One up one down scoring system is the problem. For initial placement at least that system needs to be replace.

[TheGrimm.5624]

7 hours ago, Dawdler.8521 said:

Not how that works. WR (reset, links, whatever) is overall balance of worlds against all other worlds. You cannot sort them by their “proper” tiers at that point because that would be stupid and go against the point of a reset. Hence it would be random. You can’t fix that unless it’s not a reset.

It’s the same as in any competition. You start everyone at the same point, you don’t put Usain Bolt 10m from the finishing line “because he’s so good he’ll win anyway”.

Random is extremely hard to code as a side not. Early on they had already mentioned they were planning on setting up metrics so they could better sort. The intent was to build a glicko style history for groups so that thy could refine how the sorting works after they had player history. Its probably one of the links under the wiki entry. They don't list things that get shelved so who knows. In either case it won't be random even if we don't know what logic is in place. Unless you are imagining they have a ping pong filled air tank with server names on it that will list 1-30 for the brackets. I mean could be, but I think it would be easier to have a break by statement in the code.

[Dawdler.8521]

3 hours ago, TheGrimm.5624 said:

The intent was to build a glicko style history for groups so that thy could refine how the sorting works after they had player history.

No it isnt, you keep mixing up tier placement with population balancing. The point of the "historical performance" (ie playing in certain timezones, playtimes, guild strength, commanders, whatever) is to balance the team against all other teams to achieves as flat a starting field as possible. It has nothing to do reset tier placements, which by WR design would be random. 

Otherwise what's even the point of having a WR population reset if you always go to the tier you are "supposed" to be? In fact why even use tiers, just match teams against each other for 4 weeks straight fighting only the 2 enemy teams they are "supposed" to fight by your historical group metric.

[TheGrimm.5624]

9 hours ago, Dawdler.8521 said:

No it isnt, you keep mixing up tier placement with population balancing. 

Actually no. Re-read the OP, they were asking how to fix the initial placement. Do you agree or disagree after you reviewing?

"While I am a fan of the 4-weeks relinking (and voted for it years ago) I see a bunch of problems in combination with the pseudo-random (I know Glicko is involved in it, but the Glicko of server is completely unreleated to its strength since linking so it's quite random as well) initial server placement."

Where I cross into WR is that we already have issues in tier placements while we have numbers, let alone in WR where we don't have numbers. Does that make more sense?

[TheGrimm.5624]

So question to others, are you noticing differences in attitudes in the 4 week links? Same or different vs 8 week links?

[Dawdler.8521]

1 hour ago, TheGrimm.5624 said:

Actually no. Re-read the OP, they were asking how to fix the initial placement. Do you agree or disagree after you reviewing?

I am talking about the initial tier placement. You keep talking about how Anet has said they are going to balance team populations.

But I digress, lets just agree to disagree on everything. It's easier that way.

Edited December 9, 2023 by Dawdler.8521

[TheIceman.1039]

On 12/6/2023 at 8:15 PM, Dayra.7405 said:

While I am a fan of the 4-weeks relinking (and voted for it years ago) I see a bunch of problems in combination with the pseudo-random (I know Glicko is involved in it, but the Glicko of a server is completely unreleated to its strength since linking so it's quite random as well) initial server placement.

If the initial placement put a server into (EU) T5 it cannot even reach T1 in 4 weeks, similar a server placed in T1 cannot even rech T5. That means the 4 weeks are barely sufficient to sort the links by strength, and we don't have any saturation phase anymore where server are placed where they belong and can enjoy play.

How could that be improved?

I think a possibility is to measure the K:D per server (or guild/alliance later on) over some period and (weighted by play hours) average it over all server (guilds).

Put the high KD into the top tier and the low KD into the bottom tiers. This is unlikely be worser than the random placement of WSR into T5 this relink period.

If anyone has a better proposal for the initial placement: your welcome 🙂

Another possibility could be the swiss tournament we had once in wvw-tournaments, but the winner-up/looser-down is to slow for 5 tiers and 4 weeks.

Generally I see 2 possible directions:

• make the initial placement better (that's what the KD is targeting)
• make the sorting faster (that's what the Swiss tournament is targeting)

Anet has done a good job of getting more players into WvW.
But recently, most servers have a constant queue for EBG, which has created a new problem - no one dares to go to the borderlands for help. Because they are afraid that they won't be able to get EBG back afterwards. I hope that the new WR system is not based on the old tier 1-4(5) principle, otherwise this problem will continue. Or if WR comes like this in 5 years, then it might be time to eliminate relinking altogether + merge some smaller servers.

Placement by kdr rating does not help at the moment because it does not matter if you are in T1 or T5. The Tiers system is outdated, they need to be modernized. Significantly better rewards, etc. so that players have the motivation to push up to T1. Also, the importance of the ppt in the placement calculation must be significantly reduced.

[TheGrimm.5624]

21 hours ago, Dawdler.8521 said:

I am talking about the initial tier placement. You keep talking about how Anet has said they are going to balance team populations.

But I digress, lets just agree to disagree on everything. It's easier that way.

Wot!? I thought we were doing this as normal:

But I agree we can disagree on what we were or were not talking about and what those topics and subjects were or were not about and which points were or were not about.

Normal Dawdler and Grimm conversation, move along, move along. Nothing to see here. And this is why I see you as a friend lol.

Tea or coffee? Pie or cake?

[Deniara Devious.3948]

Measuring K/D ratio is a not a good way to determine initial tiers as this might place solo servers in tier1. A better would be to tally kills + deaths (KD total count). But the best would be to measure total active ticks per players and this should be also the only way to determine server population status. Let me explain:

After each 5 minute tick ends, the game records how many players on each side (also those AFK in spawn would be counted). Then tally all those ticks together at the end of the week. If one server has a super active player who plays 200 ticks (1000 minutes) in a week, he will count as much as 10 players who play just 20 ticks each. Any other way and we will actual player presence imbalance.

My home server Desolation is once again without a link. And to make matters worse we were placed once again to tier1 after the relinks. First two weeks were of course losses and in fact all the 3 servers without a link lost their match ups and then again. We are 3rd on our match up and maybe dropping again. I hope once again relinks will not lift Desolation up to tier 1 and give us no link. This has happened way too many times. This is basically a recipe for 4 weeks of losses.

At least those servers who don't have a link, should be placed in tier 5. Why there needs to be 5 tiers in EU anyways?

Ayna

[RisingDawn.5796]

@Deniara Devious.3948 at least deso, hasn't been seen a bandwagon to their link in a few years.

No link sucks, but so does that too over the years, which destroyed many communities previously, that I enjoyed.

I'd liked to see transfers closed after the first week too tbh, but if WR is soon after the next beta like 3-4 months, I'll just take the current system and tweak the matchmaking after the relinks a bit, especially if we see so many transfers like this time around.

Edited December 10, 2023 by CrimsonOneThree.5682

[Ubik.8315]

7 hours ago, CrimsonOneThree.5682 said:

I'd liked to see transfers closed after the first week too tbh, but if WR is soon after the next beta like 3-4 months, I'll just take the current system and tweak the matchmaking after the relinks a bit, especially if we see so many transfers like this time around.

It should be the other way around, transfers open at the end of a "season" - last week of relink, which counts towards server pop in relink, then are closed for the following 3 weeks.

Edited December 11, 2023 by Ubik.8315

[Mabi black.1824]

8 hours ago, Deniara Devious.3948 said:

A better would be to tally kills + deaths (KD total count)

the sum of these two numbers should define both the levels and the status of a server. it is an excellent parameter for measuring the flow (players playing) and it is already there, available. We really have to ask ourselves why we are not using it.✌️

Edited December 11, 2023 by Mabi black.1824

[Mabi black.1824]

8 hours ago, Deniara Devious.3948 said:

My home server Desolation

good since you mentioned the mother deso, eu t3 is clearly broken. the teams are not even remotely similar for a credible competition. desolation has 20,000k+d wsr has 42,000k+d and it's only sunday night. one team has more than double the activity (players) than the other.

[Bigpapasmurf.5623]

On 12/7/2023 at 4:21 AM, TheGrimm.5624 said:

Have lived that a number of times now in my first server, from both sides, the ktrain side and the fight side. Both are limiting their options and impacting others that aren't playing all aspects. Wasn't a fan of being asked off map to allow either more fighters or more people to take empty objectives. It lead to losing the week in different ways.

If you ran into that with SBI and EBAY, I admit, makes me kind of grumpy. 

I know what you are talking about (Im on EBAY) and that very rarely happens, thinking it was SBI asking.