could it be a mistake to seek balance in relation to playing time?

[Mabi black.1824]

if the algorithm builds the balance in reference to the playing time we could get a team of 2000 players vs a team of 1000 players and since one plays twice the time for the algorithm we would have a perfect balance.

at this point I can only imagine that most of us get some recreation only after work . only for hypothesis we say that 80% of the players will be active let's say from 18.00 to 22.00 continental time (80% of 2000 players vs 80% of 1000 players)

the difference in hours played could be made by the remaining 20% who can afford not to work or to have particular working hours or to stop until late in the evening etc.etc.

The result we get is that 80 percent of players will feel like they're playing outnumbered by about 50 percent.

but the algorithm did its job perfectly. that's perhaps why many of us don't find ourselves with the balances that are calculated and constructed.

also the message / advice if you are outnumbered practically is............... if you are losing do not get busy let them take all your map indeed play little just 30min and go, so in two months you could have more players.

or you will have this game mode, do not play it or you will be constantly inferior.

am I wrong? what do you think?

[kash.9213]

That would be dumb and horribly skewed if it factored in play time. Some people are All Stars during the quick hour or so after work or classes while some people are running into walls for hours on end doing nothing or just camping out in some squads fields pressing auto attack for the entire prime time. Trying to correct that any imbalance of low time played with more imbalance of high play time but low effort or activity would compound that mess. 

[Chaba.5410]

23 minutes ago, kash.9213 said:

Some people are All Stars during the quick hour or so after work or classes while some people are running into walls for hours on end doing nothing or just camping out in some squads fields pressing auto attack for the entire prime time.

And some people are All Stars for 8 hours a day while other poor bastards are suck at work or school and can only log on for an hour or so, maybe do some dailies and barely contribute.  Those players who can't keep up in play hours with the no-lifers deserve to lose.  Such competition.  Much winning.  /sarcasm

Edited February 2, 2022 by Chaba.5410

[subversiontwo.7501]

I wrote at length about this the other day. I think we can see the two-grade system we have now move towards what could perhaps be a four-grade system or such. I don't think there is much reason for more granularity than that because of how sensitive activity is and how many other factors are at play. Among them: Activity fluctuates so it is difficult to measure in detail, being overly focused on small activity differences is going to limit possible combinations or variations in the world creation (ie., if it plays too large a part it risks recreating too similar worlds repeatedly) and things like alliances make up such large building blocks anyway, so measuring activity in detail becomes rather pointless. So the system is likely to get more grades but broad strokes is still the ideal measurement.

[Chaba.5410]

I recall an Anet post once mentioning how WXP gain was tracked to sanity-check the playhours algorithm.  WXP gain tracked closely which means the vast majority of players are actually playing during the time they spend in WvW.

[Dawdler.8521]

11 hours ago, Mabi black.1824 said:

if the algorithm builds the balance in reference to the playing time we could get a team of 2000 players vs a team of 1000 players and since one plays twice the time for the algorithm we would have a perfect balance.

This assumes a straight linear curve, that you do "twice as much" if you play 2h vs another guy playing 1h for the overall team score. On a micro scale between 2 people that may be true, but when you draw that out over a week with a total of maybe ~8000 players across 3 sides, its not that basic. I highly doubt thats how Anet counts populations even now.

Why isnt it 1.1 worth? 1.2? Is there anything that back up the idea it would be balanced 2000 vs 1000 rather than say 1600 vs 1400?

Edited February 3, 2022 by Dawdler.8521

[Mabi black.1824]

10 hours ago, Chaba.5410 said:

And some people are All Stars for 8 hours a day while other poor bastards are suck at work or school and can only log on for an hour or so, maybe do some dailies and barely contribute.  Those players who can't keep up in play hours with the no-lifers deserve to lose.  Such competition.  Much winning.  /sarcasm

hi chaba,

try to look at what I wrote with an open mind. do not take into account the special exceptions that may be present on all sides of the game.

do you think my reasoning could be correct? after all, there must be a reason and therefore an error in the algorithm that builds the teams. it is not possible that for everyone they have an absurd composition, which does not correspond to the sensations you have at stake.

the algorithm could really do its job well, but the result it gets for the majority of players is absolutely out of place.

[Mabi black.1824]

2 hours ago, Dawdler.8521 said:

This assumes a straight linear curve, that you do "twice as much" if you play 2h vs another guy playing 1h for the overall team score. On a micro scale between 2 people that may be true, but when you draw that out over a week with a total of maybe ~8000 players across 3 sides, its not that basic. I highly doubt thats how Anet counts populations even now.

Why isnt it 1.1 worth? 1.2? Is there anything that back up the idea it would be balanced 2000 vs 1000 rather than say 1600 vs 1400?

hi dawdler,

since I was told that teams are built based on the number of players and the time they play, I just gave a mathematically practical example.

there is nothing to support I hypothesized a team that plays exactly twice the time of another.

and my reasoning mathematically works. Of course, in real matches we will have different proportions that are not exactly twice as good as the other.

but my goal here is to reason that 20% of a team can really influence the playing experience of the remaining 80% as I indicated above.

[Mabi black.1824]

10 hours ago, subversiontwo.7501 said:

I wrote at length about this the other day. I think we can see the two-grade system we have now move towards what could perhaps be a four-grade system or such. I don't think there is much reason for more granularity than that because of how sensitive activity is and how many other factors are at play. Among them: Activity fluctuates so it is difficult to measure in detail, being overly focused on small activity differences is going to limit possible combinations or variations in the world creation (ie., if it plays too large a part it risks recreating too similar worlds repeatedly) and things like alliances make up such large building blocks anyway, so measuring activity in detail becomes rather pointless. So the system is likely to get more grades but broad strokes is still the ideal measurement.

hi subversiontwo,

I am not talking here about ''small differences''.

if you enter the time parameter in the algorithm since the teams have a limit, it is absolutely possible to find a team of 2000 vs a team of 1000 (assuming the maximum limit of 2000)

in fact the algorithm has done its job, but the experience you will have in the game is what I have indicated above.

also give a message to all the wrong players in my personal opinion, that is, if I play twice the time of my opponent I could finish the weekly game to the parts of my opponent who has done half of what I did, to win with a little margin practically I have to commit 3 times as much.

passion and consequently time must be a quality not a defect.

and if we do not reason about this in the current system we also drag it into alliances. the algorithm could still build teams one twice as much as the other by doing its calculation correctly.

[gloflop.3510]

It will be a challenge for anet to implement further measures than only "time online" to balance the teams. Anet can e.g. also consider coverage, hours commanding a squad, structures flipped while online,...

 

Another problem is that all of anet's data relies on the past. To create the worlds, they assume that player will play in the future like they did in the past (playing hours, etc.). This assumption is unproblematic as long as a drop in their measures is non-systematic (meaning: if the playing time in WvW for all players goes down by 10%, there is not one worlds more affected than another). The hurdle is that entire guilds can leave the game mode or players can go to their secondary accounts. It both skews the balance.

[subversiontwo.7501]

3 hours ago, Mabi black.1824 said:

if you enter the time parameter in the algorithm since the teams have a limit, it is absolutely possible to find a team of 2000 vs a team of 1000 (assuming the maximum limit of 2000)

Quote

passion and consequently time must be a quality not a defect.

It depends on what you mean by this stuff. The system does not work in that way, at least not directly. It could possibly create a situation where you have 1000 players highly active players and another 1000 players who play so little that they have trouble making any impact (but still reach the minimum). Then the first 1000 players may feel like they play 1k vs. 2k. That would be a problem. But that is not a likely situation. Directly making some worlds smaller (ie., actually create worlds of 1k and 2k players) would be even worse. Neither of those extreme situations are healthy considering what I pointed to above about activity and players feeling forced to play.

Even if someone builds a hyper active alliance of 500 players, it is not a good thing to repeatedly put them into a situation of constantly having to carry a workload alone, especially not if they will ping between being hyper active and just highly active. That will burn players out, make them quit or have them drop their ambition or manipulate the system. It represents the same flaws we see now with the linking system where the unlinked servers may be the largest but they still suffer heavily from having no link. That is a problem now and any new system is not going to try to reproduce it.

The same goes for words like "passion" or "consequent time", it depends on what you mean by them because in the context of time and activity balance they could be interpreted as being expected uphold coverage. Those are some of the issues in the system that we look to get away from if we are to have healthy matchups. Even if some groups form that are very active in their spare time, they still tend to have school or work to occupy them and need as much sleep as everyone else. Activity does not, generally, operate outside of that framework and it is important to have a grounded idea of what activity actually is.

While the system may look to balance things out and make them more even, it should be understood as primarily looking to create good matchups where players can be matched up against each other at the same time. It is not there to mismatch them to take out each other for balanced scoring results (ie., matching a night-world with a day-world, or give another world a night-coverage to compensate for getting stomped in prime). Even in prime alone, you can't have a very active group of players be expected to hop around maps to cover opposing groups with numbers to hit all maps during half of prime. That is not desirable so you don't build a system for that.

It is more about content than it is about results. That way player numbers is always more important than activity in the system and activity is more important than- or should not be conflated with other factors. Correct me if I am wrong here but I get the impression that you may be a bit more concerned with the latter: of balancing results. If they want to address and shape results that is best done through scoring algorithms, not world-creation algorithms.

Think of the world-creation algorithm more like this: It will categorise players based on whether they are group or solo, based on a broad category of activity and play hours so you will get a set of categories, like this example:

1. Group, activity A (makes diamond) ---> worlds A-L (12 worlds, 4 matchups) or whatever
2. Group, activity B (makes gold)
3. Group, activity C (makes wood)
4. Solo, activity A
5. Solo, activity B
6. Solo, activity C

If the system adds play-hour (timezone) tracking it may look like this:

1. Group, prime, activity A
2. Group, offhours, activity A
3. Group, prime, activity B etc.

They will then be spread out as far is possible within each category and be filled up with the next order of grouping. In a first draft of the algorithm they may not even track what hours are being played and just categorise the system over group-solo and however many grades of activity they aim for (3, 4, 7, whatever). So the system will take "Category 1" and then divide them up over as many worlds as it can cover. Then it takes "Category 2". The size of groups sits behind this, right, because they can just even out totals by placing a 500-player category one block on one world and two 250-player category one blocks on the next world or five 100-player blocks on the third world.

So the algorithm itself does not differentiate between the sizes of groups, they just fill the numbers up as they go, making sure totals are even for each category. There is no practical difference between world A getting a 500-player alliance, world B getting a 500-player guild, world C getting two 250-player guilds and world D getting five 100-player alliances as long as they are all category one and total 500. Then the system isn't even because when we hit world L or whatever, we may be out of category one groups to seed but the difference never stacks beyond that.

If you understand the concept of the system you'll also understand that factors like activity or play hours could have as many grades or the system could have as many categories as they would like and it would matter little for the seeding and just be a question of how many categories they want their sample in and how complex or detailed they want to make the system (or precise but sensitive, as a system). They could have no play-hour tracking, they could have a system of two (prime/off) or they could divide a 24-hour day into four 6-hour slices and use that to identify a group's or player's general activity, essentially labling players morning, day, evening and night and then use that to seed them on coverage spread. Putting the coverage factor in a hierarchy of factors superceeding each other.

As always, I think the biggest misconception among players of, let's say, category six up there in the list, is that they are afraid that category one will be one world and category six another world. That's just not how the system works even if the betas, without the algorithm, may in some specific case give that impression. Players like that tend to conflate alliances with worlds, simply not understand the system and be worried about their own misconception.

Like I wrote the other day, if they track results (and go in one specific direction with regard to rewards and matchups) they could add results as a factor into the system, but it would still be superceeded by essentially everything else: Mainly be a factor that would separate successful groups within the same categories from ending up on the same worlds so they got to match up more. There is no healthy scenario where the system is going to punish a successful group by making them play on a smaller world or a dead world. That is not the balance the system seeks.
 

Edited February 3, 2022 by subversiontwo.7501

[GoldenPants.1870]

All I want is to not constantly be outnumbered even in prime time.

[Dawdler.8521]

3 hours ago, Mabi black.1824 said:

and my reasoning mathematically works.

No it doesnt because you dont know how Anet weighs any of this in the algorithm.

This is the same as saying 1 power adds 1 damage and calling it mathematically correct.

[Mabi black.1824]

2 hours ago, subversiontwo.7501 said:

It depends on what you mean by this stuff. The system does not work in that way, at least not directly. It could possibly create a system where you have 1000 players highly active players and another 1000 players who play so little that they have trouble making any impact
 

at this point I'm a bit confused, on the forum someone explained to me that today teams are created in reference to the number of players and the playing time of these players.

since the teams have a cap (I assume 2000) it is impossible as you say that we will find 1000 active players vs 1000 less active players.

instead you can get a team of 2000 players vs another team of 1000 players since the latter play twice as much time as the former. ( and I would ask the administrator for confirmation if the algorithm reasons in this way ).

there would be nothing wrong with this logic probably the intention is improving in the count of players.

the problem is that the result it gets is not what you wanted. Why? now reason are approximately / statistically with the following questions.

Is it true or not true that most of us play in our spare time?

is it true or not true that most of us have to work/study to live?

so exceptions aside, as I said in the first port, let's say between 18.00 and 22.00 most of us could be online to have some fun.

only by hypothesis 80% would get the experience of 80% of 2000 vs 80% of 1000 for all the evenings of the week.

is it perhaps for this reason that when they build the combinations we are so astonished?

Is this why you see your team full for months and years despite entire groups of guilds leaving? or do you never find things on any map?

I still don't want to talk about coaches or all the parameters you can add to build and associate teams. I would like to remain at the root of how the algorithm thinks.

[The Boz.2038]

Playtime can mean "how many hours per month do they put into WvW" and "at what times does this player normally log on". I think your misunderstanding is between the two.

The first is used by the algo, AFAIK. The second would be... confounding to calculate with.

[subversiontwo.7501]

30 minutes ago, Mabi black.1824 said:

at this point I'm a bit confused, on the forum someone explained to me that today teams are created in reference to the number of players and the playing time of these players.

since the teams have a cap (I assume 2000) it is impossible as you say that we will find 1000 active players vs 1000 less active players.

instead you can get a team of 2000 players vs another team of 1000 players since the latter play twice as much time as the former. ( and I would ask the administrator for confirmation if the algorithm reasons in this way ).

there would be nothing wrong with this logic probably the intention is improving in the count of players.

the problem is that the result it gets is not what you wanted. Why? now reason are approximately / statistically with the following questions.

I still don't want to talk about coaches or all the parameters you can add to build and associate teams. I would like to remain at the root of how the algorithm thinks.

I can only refer you back to the post you are qouting. I feel as if it already explains both how the system is likely to work and why it would be a bad thing if it worked as someone else gave you the impression it did.

The easiest way to understand it is to understand it in terms of factors and hierarchies. The system first looks to balance populations. After that it looks to balance other factors. It does not balance one part (population totals) with another part (activity). Instead it takes activity as a factor under or inside how it balances totals. See the list.

It is bad for a system to punish players for being active or successful, that's why no system is going to purposely try to do that. That's the discussion part before the list. Furthermore it can't really punish players for being active or successful by just stacking more players against them when they don't play. That would just recreate the number one problem they are trying to solve: population imbalance and matchups won by players who play without opponents.

Ps. It is like God is saying below. I think you are confusing yourself a bit with that example of 1000 players as that has no reference in the system. Think about it in its actual chunks of 500 players instead and you can see that the ideal of the system is rather that most servers will have roughly 500 more active group players, 500 more active solo players, 500 less active group players and 500 less active solo players. Once you burn that image into your mind you can begin to look at how the system is likely to count and see those 4 pieces break into more pieces.

27 minutes ago, The Boz.2038 said:

Playtime can mean "how many hours per month do they put into WvW" and "at what times does this player normally log on". I think your misunderstanding is between the two.

The first is used by the algo, AFAIK. The second would be... confounding to calculate with.

In all fairness, Anet have said that they are considering using both factors. Though it is likely that "how much" will be tested first and "when" may be added later.

Edited February 3, 2022 by subversiontwo.7501

[God.2708]

12 minutes ago, Mabi black.1824 said:

at this point I'm a bit confused, on the forum someone explained to me that today teams are created in reference to the number of players and the playing time of these players.

since the teams have a cap (I assume 2000) it is impossible as you say that we will find 1000 active players vs 1000 less active players.

instead you can get a team of 2000 players vs another team of 1000 players since the latter play twice as much time as the former. ( and I would ask the administrator for confirmation if the algorithm reasons in this way ).

there would be nothing wrong with this logic probably the intention is improving in the count of players.

the problem is that the result it gets is not what you wanted. Why? now reason are approximately / statistically with the following questions.

Is it true or not true that most of us play in our spare time?

is it true or not true that most of us have to work/study to live?

so exceptions aside, as I said in the first port, let's say between 18.00 and 22.00 most of us could be online to have some fun.

only by hypothesis 80% would get the experience of 80% of 2000 vs 80% of 1000 for all the evenings of the week.

is it perhaps for this reason that when they build the combinations we are so astonished?

Is this why you see your team full for months and years despite entire groups of guilds leaving? or do you never find things on any map?

I still don't want to talk about coaches or all the parameters you can add to build and associate teams. I would like to remain at the root of how the algorithm thinks.

Why do you think creating teams based on playing time means a directly correlation to playing hours?

They could be using any number of metrics to find that out @subversiontwo.7501listed one where they use how far you make it in pips each week. But in that case a new player has to 'play' almost twice as long as a person ranked diamond.

Your attempt at being logical is failing because you aren't considering the logic of your premise.

[subversiontwo.7501]

5 minutes ago, God.2708 said:

They could be using any number of metrics to find that out @subversiontwo.7501listed one where they use how far you make it in pips each week. But in that case a new player has to 'play' almost twice as long as a person ranked diamond.

Indeed, it is a simplification mainly used to give people some sort of visual reference that they can understand. While drifting a bit off topic I also think that layering different systems on top of each other is not a good thing, so if I was ArenaNet I would probably do away with some of the more problematic factors of that separate system so they could be used in the same way like that for sake of simplicity and visualisation. It isn't a good thing that large-scale gains higher pip levels so much quicker than small scale and it isn't a good reward to give experienced players bonus pips so they have to play less to achieve the same. They could easily throw that bad design out and then they could use the chests as the actual measurement of time and not just as a simplified example. That is another discussion though that I think may not help people with the topic at hand 🙂 .

[Mabi black.1824]

9 minutes ago, God.2708 said:

Why do you think creating teams based on playing time means a directly correlation to playing hours?

They could be using any number of metrics to find that out @subversiontwo.7501listed one where they use how far you make it in pips each week. But in that case a new player has to 'play' almost twice as long as a person ranked diamond.

Your attempt at being logical is failing because you aren't considering the logic of your premise.

hi god,

here I'm just trying to figure out/share where the error might be in the current algorithm.

and I would also ask the administrator for help.

no one told me about other metrics, the calculation I suppose is done based on the number of players and the time they play, at least so I was told.

consequently if the teams have a cap and I play twice as much as you, we could be in 1000 players like me vs 2000 players like you.

and for the algorithm we are two perfectly balanced servers.

[Mabi black.1824]

an explanation there must be we see it again this week in the t5 , I do not find there but I can only imagine 2000 vs 1000 vs 1000.

how can a mathematical calculation get that result I can't really figure it out.

[subversiontwo.7501]

13 minutes ago, Mabi black.1824 said:

here I'm just trying to figure out/share where the error might be in the current algorithm.

What is so hard for you to understand about this? I'm rarely lost for words, but I am getting there.

1. The current system: Servers, has no algorithm like that.

2. The beta system: Does not have that algorithm at its current stage (beta 3).

3. The beta system: Will later have some sort of algorithm like that based on logic and developer comments.

The current system (Servers) have different sizes because some servers has never had larger populations or because players transfer around. Server sizes are not reset, at all. If your server is smaller or larger than another server that only has to do with how many players choose that world when they started the game, how many players have paid to transfer there or how many have quit since. That is what the devs want to change and what we have betas to test.

When people talk about an algorithm they talk about the beta 🙂 .

Edited February 3, 2022 by subversiontwo.7501

[Mabi black.1824]

10 minutes ago, subversiontwo.7501 said:

 

1. The current system: Servers, has no algorithm like that.

 

 

[Mabi black.1824]

1. The current system: Servers, has no algorithm like that.

 

in fact they often tell me that it takes me a while to understand ☺️

so we can say that the current system calculates only in number of active players without considering their playing time.

a little more worried than before at this point, I was hoping to have found a possible error in the logic of the algorithm.

[subversiontwo.7501]

Haha, well yes, I honestly did not even consider that the existing system confused you because it doesn't really calculate anything 😅.

Assuming that you play on EU given the time of the day:

• There are 27 servers.
• They make up 15 worlds (5 tiers of matchups; 3x5)
• So 15 servers are hosts for those 15 worlds and 12 servers are links to those 15 worlds.
• This means 12 servers are linked and 3 servers are unlinked for the 15 worlds.
• That is the system now.

Then you have the players. The smallest servers are the cheapest to transfer to (200 gold or 500 gems). So many players just transfer every relink now. So what is a big server + a small server on relink day is often a big server + a big server days later.

So if you look at the current T5 in EU:

  Baruch Bay

  Abaddon's Mouth +Whiteside Ridge

  Seafarer's Rest

You see two unlinked and one linked server. The unlinked servers may feel small alone but are actually the largest servers. Then you have the special case of Baruch Bay. Baruch is treated different from every other server in the game. In order to have 1 Spanish server Anet made it so Baruch can never be full but also never gets a link as a result. So they are an exception to the system.

If you look at T1 in EU:

  Fissure of Woe +Arborstone

  Elona Reach +Kodash

  Aurora Glade +Jade Sea

 

You can see that they all have links. Fissure of Woe was also a server that a couple of months ago was one of the smallest servers, so alot of guilds and alot of players transfer there. They now have a link in Arborstone so even more players can transfer there cheaply. As a result that one world has a big+big combination of players and on top of that they have alot of active players and veteran players. As the weeks pass now, more and more players will transfer there to "win". FoW will become full. It will lose its link. The guilds can then not recruit and they can usually not get good prime time content in tier 1. They can't tank the server down because the people wanting to win keeps their PPT up. So sometime soon those guilds will begin to transfer to the smallest server again. So the cycle continues 🙂 . Guilds move away from players who want to "win" and those players follow the guilds because the guilds being there makes it easier for them to "win".

 

Edited February 3, 2022 by subversiontwo.7501

[Mabi black.1824]

2 hours ago, GoldenPants.1870 said:

All I want is to not constantly be outnumbered even in prime time.

we all agree with you goldenpants 😅