Is it possible to get some more transparency on the team-making algorithm now?

[Visirale.6097]

This is the third set of new teams, and we can say the 2nd using data from post-world-restructuring. This time we were able to glean a little bit more insight of the teams from the API before the actual new teams started. I wanted to give some time to make sure that information ended up being correct (it did) before asking some questions about the system.

How are different alliance guilds/battle guilds, or whatever we want to call them, sorted into teams for the purposes of team-making? What metrics are being used? We got a little insight into how off-hours players affect the information given to a disproportionately NA-heavy alliance. But at this point, I'm really wanting to know more of anything going into it. From the outside looking it, its hard to think of metrics being used outside of pure numbers of members in a guild. I'm in a medium-sized alliance, but one that is full of very active players. I'm sure our average hours played per week as an alliance is more than other alliances double or more our size. But we're getting stacked with other groups that are similarly as active and getting stuck in queue hell. Because we have access to the information, we can see other teams that look empty in comparison. How are these groupings being made?

Do you have access to information like average length of playing session for players? Average number of structures flipped vs. average number of player kills in a session? If you told me it's completely random and you don't have a lot of information to base an algorithm on, I'd believe you... but I'm also assuming that surely can't be the case in 2024 with an API that gives almost all the information I've mentioned. 

Surely a once over human review of the team assignments could catch some of the outliers and work to make changes before the teams go live. I know that's a lot to ask, and probably not in anyone's job description. But I also assumed these team-matching assignments would get better as time went on in the new system, and objectively (with an admittedly small sample size) they're getting worse. 

I appreciate all the time and effort that goes into the system, I see the devs commenting on it and working with it. Just really wondering when we can expect some more transparency on the system or reflection from your side on how you've viewed the system's performance so far. 

[XenesisII.1540]

1. They are obviously not going to release information that the players could/would exploit to their benefit.

2. Anet doesn't make any effort for wvw discussions in the forums, so good luck with that.

[Evenge.4067]

Dartboard is the algorithm.  🎯

[MedievalThings.5417]

The algorithm isn't "adjusting" based on guild/server performance.  They are just matching play hours and NOTHING else.  If you have a guild of yak walkers and a supercomp of equal size and play time, the system treats them as equals.  The algorithm places a superstack, a medium server and a "stop farming me bro" server in every single tier every reshuffle.  This is what it has been programmed to do and what it has been doing since the very first beta.  Why, we have no idea.  We can only assume it's intended since that is what it has always done, EVERY single beta and reshuffle.  You have to remember this was designed by people who stood in a 40-50 man blob farming 5-10 guys over and over and when they said it was fun chat asked if the guys being farmed were having fun too...and they said yes (they also commented if they weren't they could just transfer, but the sentiment shows their thoughts).  The world restructuring was never going to make things even, it was never designed to.

[One more for the road.8950]

5 minutes ago, MedievalThings.5417 said:

The algorithm isn't "adjusting" based on guild/server performance.  They are just matching play hours and NOTHING else.  If you have a guild of yak walkers and a supercomp of equal size and play time, the system treats them as equals.  The algorithm places a superstack, a medium server and a "stop farming me bro" server in every single tier every reshuffle.  This is what it has been programmed to do and what it has been doing since the very first beta.  Why, we have no idea.  We can only assume it's intended since that is what it has always done, EVERY single beta and reshuffle.  You have to remember this was designed by people who stood in a 40-50 man blob farming 5-10 guys over and over and when they said it was fun chat asked if the guys being farmed were having fun too...and they said yes (they also commented if they weren't they could just transfer, but the sentiment shows their thoughts).  The world restructuring was never going to make things even, it was never designed to.

Pretty contradictory to claim the algorithm is set to make a super team, medium team and farm team but is also only counting player hours and nothing about what you actually do in WvW - how can they make three types of teams if they have no idea who can press buttons or walk into a tower? And what would be the benefit of making each tier like that when it would change next week anyway?

[Skub.8240]

I think it's something like this:

They go to a pet store and adopt a thousand mice. They assign each mouse a unique number based on player IDs, then put the mice in a maze where the exits are the various teams. And when a mouse makes it to an exit, that's it! You're sorted into a team!

Then they release a hungry snake into the maze.

[MedievalThings.5417]

47 minutes ago, One more for the road.8950 said:

Pretty contradictory to claim the algorithm is set to make a super team, medium team and farm team but is also only counting player hours and nothing about what you actually do in WvW - how can they make three types of teams if they have no idea who can press buttons or walk into a tower? And what would be the benefit of making each tier like that when it would change next week anyway?

Not really.  Anet is the one that said they were only matching play time.  I am pointing out that for 3 years now, including the current 3 relinks, they have always placed a superstack, medium and underdog in every tier.  Some people had this idea that restructuring was going to fix imbalance, but it doesn't, and 3 years' worth of betas has proven they aren't trying to.  I don't know why they wanted to get rid of the old servers/communities, but they really wanted to.  It's like trying to explain why Anet really doesn't like people defending.../shrug.

[cannibalkitteh.8376]

I sure would love some sort of high level discussion too.

I think the algorithm lacks context for a lot of things. Papering an empty BL and fighting a mapqueue while papering a BL could result in similar points with reference to how active you are as a group. Similarly, I'm not sure if it can draw a distinction between kills in a meaningful way. If you often played with another group's comped zerg in your last match, you might appear to be performing well as far as PPK, but will it continue next shuffle?

[Evenge.4067]

To me, they seem to be putting one very strong alliance,  one medium strength, and one weak alliance in each tier. That is how the scores and KDR look. Why they would do this makes no sense.  This only allows the strong alliances in the low tiers to stomp their way up for weeks to Tier one where they should have been placed to start the relinks. It's not competitive. It only tends to discourage weak alliances. 

Place strong against strong in Tier 1 to START the relink month! 

[RisingDawn.5796]

At least on the EU side, there are new Alliances which formed, during the past few weeks, so the algorithm  needs to adjust to that shift in population after the new reshuffle, just like how guilds used to transfer so often, thanks to the world linking system.

Then again in T5 you clearly have one large alliance, medium and a lot of pugs, so more activity than others in this tier.

With one up/down, at least it will take them weeks to get up T1, if they do have one of the most active populations though.

Edited September 1 by RisingDawn.5796

[Palikka.8249]

I tought new system was supposed to cover content for different times. Currently blues on T3 has only activity at morning, other time we are fodder and so few playing. 3rd week and this system is as s**t as when this clownfiesta started.

[Sonderm.4639]

It is probably just a fortunate accident that the "algorithm" put a big streamer's guild, which had a recruitment event for wvw attended by anet devs, on a team with the currently highest k/d in EU instead of on some farmville team - or said guild is pumping hard and carrying, in which case I humbly apologize and eat my tin-foil hat as penance.

Edited September 1 by Sonderm.4639

[Chaba.5410]

12 hours ago, MedievalThings.5417 said:

Anet is the one that said they were only matching play time.

You're misrepresenting what has been published.

[Dinas Dragonbane.2978]

13 hours ago, MedievalThings.5417 said:

  The algorithm places a superstack, a medium server and a "stop farming me bro" 

Thumbs up for this, it's funny and we have all seen it multiple times.

[Palikka.8249]

6 minutes ago, Dinas Dragonbane.2978 said:

Thumbs up for this, it's funny and we have all seen it multiple times.

EU T3:

Red = stacked, Green = medium, Blue = why are we even here. 

And still both reds and green zerg our borderland and not each other. We are constantly outnumbered even if we are against one team. Great system, how many more weeks we need before this WR does not suck?

Edited September 1 by Palikka.8249
typo

[bigmouse.2163]

I do wish they would give us some kind of indication of how matchups are made. So far they seem almost random. I have my doubts that there even is an "algorithm," and if so, I doubt it's based on much more than play time.

[Leaa.2943]

We are still running on One up/One down matchups. This means that the only thing Anet decide about is how the first match will  be played out. After that it is One up/One down that settle who you fight and at that point what ever they paired us with have no longer meaning until next re-shuffle and that one first match.

[Zephyrus.9680]

I'm generally skipping the first week or two because of how bad the balance is. Mid-day NA on ebg had outnumbered buff and basically no one playing. On Sunday. First step would be removing the extra tier but that might require investing more money in servers to handle more simultaneous fights and they recently downgraded servers. Servers cost money. 

[One more for the road.8950]

12 minutes ago, Zephyrus.9680 said:

I'm generally skipping the first week or two because of how bad the balance is. Mid-day NA on ebg had outnumbered buff and basically no one playing. On Sunday. First step would be removing the extra tier but that might require investing more money in servers to handle more simultaneous fights and they recently downgraded servers. Servers cost money. 

What extra tier?

It's back to what it was.

[Zephyrus.9680]

1 hour ago, One more for the road.8950 said:

What extra tier?

It's back to what it was.

Could have fooled me. Then I wonder how they are making such dead teams? Every time I looked there was nothing on any map, except once. Roamed a couple hours but it was a snooze with the other two teams just fighting each other. 

Edit: Just checked mists site and ours (T3 NA) is the most blowout tier by far from both regions. Bad dice roll I guess. EU is actually pretty balanced by the numbers, though 3/4s of NA are blowouts. 

Edited September 2 by Zephyrus.9680

[Dawdler.8521]

1 hour ago, Zephyrus.9680 said:

Could have fooled me. Then I wonder how they are making such dead teams? Every time I looked there was nothing on any map, except once. Roamed a couple hours but it was a snooze with the other two teams just fighting each other. 

Edit: Just checked mists site and ours (T3 NA) is the most blowout tier by far from both regions. Bad dice roll I guess. EU is actually pretty balanced by the numbers, though 3/4s of NA are blowouts. 

Been checking the kills+deaths (ie overall activity) since the start of WR and mondays is one of the usual datapoints. Today it's just broken 1 million KD, havent seen that since the first weeks of WR (but granted I have been missing the last couple of weeks due to the API and me starting to forget doing it). It's averaged ~800k on mondays over the summer.

Edited September 2 by Dawdler.8521

[Dinas Dragonbane.2978]

Don't forget this weekend was a holiday weekend for the US at least!

[Mimizu.7340]

I'm in Skrittsburgh EU Tier 1 and it's DEAD after midnight until dawn.

AlGoRiThM has decided everyone has to go to bed early on our side whilst the other two sides ppt all night long.

They are still losing in points and k/d but I can't play in my usual time slot anymore.

When our side is awake, it's Qs on all 4 maps...

Pathetic doesn't even begin to describe it.

Funky Little Shack.

 

[ChrisWhitey.9076]

We have this and the last update was it was not working due to off hour stacking

 

[Dawdler.8521]

14 hours ago, Mimizu.7340 said:

AlGoRiThM has decided everyone has to go to bed early

I believe you've mistaken the algorithm for parents.