Guild Wars 2

Moving Averages are always behind trend changes

A moving average is very nice to smooth variation
and the longer the period that goes into the average the smother is the result.
From the Devs comments on the intro of the new population it is clear that it
averages at least one week (such that it can smooth hourly and daily variations),
most likely more (to smooth fair-weather commitment it must be severaly weeks).

From the behaviour of it: I guess it averages one month

Let’s look what this means to the server population status.
Assume a server has 150% population of what is considered FULL
Let’s asume this is the case since a while, such that the moving average is 150 as well.
Assume further that on day 15 several guilds transfer away such that on day 15 the
population drops down to 90% of what is considered FULL.

When will the 30-days moving average drop below 100% and flag the server as very high?

25 days after the transfer happend the 30-days moving average will cross 100%.
So don’t wonder if it takes a while you are flagged very high.

This feels bad, but this is not the problem. The problem is: Any (not-full)
server has several days of overstacking till the moving average flags it as full.
If a lot of people decide to go to not-full server XYZ on a friday before reset,
they could easily stack several times as much as allowed before the server becomes
flagged as full.

I strongly recommend that transfers influence population immidiately
So immiatedly when a transfer is executed remove 1 from the moving average of the source and add 1 to the moving average of the target. Transfers are only done for WvW, an wvw-inactive unlikely spends gems on ransfers and they aren’t noise that need to be smoothed. They affect both servers immediately.

What is counted? People or Personhours

At first sight it is naturl to count people. But is this really adequate?
A 24/7 match with a capacity of 300 people at the same time has a capacity
of around 50’000 personhours not persons. How full a match is depends
only on the total personhours a server commits to a match and not at all
on how many person are involved in this commitment. In fact the more people
are involved in commiting a specific amount of personhours the worser the
server plays (400 people spending 15min per day to do dailies are the same
personhours than 100 people fighting 2h a day).

Even if it appears natural to count persons on first sight, personhours are
a much better reprensation of the fullness of a server

To combine it with the immediate transfers above it is probably sufficient to use the mean personhours of the source server to caculate the the transfer effects for source and target.

The only thing that’s known for sure about WvW populations is that any thread talking about that subject will be deleted.

if you dont know where the bar for full is, how will you know when its crossed?

Swamurabi.7890:

The only thing that’s known for sure about WvW populations is that any thread talking about that subject will be deleted.

If they send it to the developer to read it when they delete it it is fine for me

insanemaniac.2456:

if you dont know where the bar for full is, how will you know when its crossed?

I made my computations in percent not in actual numbers, so 100% is full by definitions, it doesn’t really matter how much that is.

Dayra.7405:

Moving Averages are always behind trend changes

A moving average is very nice to smooth variation
and the longer the period that goes into the average the smother is the result.
From the Devs comments on the intro of the new population it is clear that it
averages at least one week (such that it can smooth hourly and daily variations),
most likely more (to smooth fair-weather commitment it must be severaly weeks).

From the behaviour of it: I guess it averages one month

Let’s look what this means to the server population status.
Assume a server like SFR (or NA-T1) has 150% population of what is considered FULL
Let’s asume this is the case since a while, such that the moving average is 150 as well.
Assume further that on day 15 several guilds transfer away such that on day 15 the
population drops down to 90% of what is considered FULL.

When will the 30-days moving average drop below 100% and flag the server as very high?

25 days after the transfer happend the 30-days moving average will cross 100%.
So don’t wonder if it takes a while you are flagged very high.

This feels bad, but this is not the problem. The problem is: Any (not-full)
server has several days of overstacking till the moving average flags it as full.
If a lot of people decide to go to not-full server XYZ on a friday before reset,
they could easily stack several times as much as allowed before the server becomes
flagged as full.

I strongly recommend that transfers influence population immidiately
So immiatedly when a transfer is executed remove 1 from the moving average of the source and add 1 to the moving average of the target. Transfers are only done for WvW, an wvw-inactive unlikely spends gems on ransfers and they aren’t noise that need to be smoothed. They affect both servers immediately.

There is a flaw with this. Being on SOS I and receiving a lot of transfers I can tell you that most guilds who transfer do not run WVW 7 days a week. Most seem to run about 2-3 times a week, sometimes even less, sometimes more with as little as 1 day a week or as high kitten -6 days a week.

The remaining time their people who play in WVW and run on other guilds / public raids, are about 1/2 or often even less then when they do guild raids. This number also varies up and down.

If you have a system based on the number of received transfers vs the actual WVW activity, then the system will grossly over represent the actual population in WVW as it will falsely assume everyone that transferred in will be playing 7 days a week, entire length of their respective prime time, which is extremely not the case.

The second issue would be with failure to account for timezone / weeklong balance. As you know most servers populations are high on resets and weekends and scale down drastically during the week, some more then others to various degrees. If you ignore the activity moving average in favor of the false assumption outlined above you will get very skewed results.

I think I assume only:

A person not playing WvW will not spend gems to transfer and a person will roughly play the same before and after the transfer, whatever this is. Timezone and day are still completely smoothed away by averaging.

And of course, if the population count is a person count it assumes implicitly that every person counts the same, which is wrong, why I proposed in my 2nd post to switch to personhours instead of persons, where the “mean personhours” is only an approximation to be used till the moving averages adapt to the new situation.

It doesn’t mean it’s fair, when populations are imbalanced; there is no competition to take seriously (like many on Deso believe now), you are suggesting because it suits one party it’s fine, that is anti-competition. I generally don’t care anymore, as suggested you shouldn’t too, they will qq @ people who are fights and complain about us GvGing rather than PPTing, even though large scale fights are a feature in WvW.

It’s as though people believe PPT is any fun or has any purpose anymore, but they hold on to Server Pride, which I guess is true for Servers who don’t have continous streams of bandwagoners and fairweathers.

T1/T2 NA seems more balanced from what I’ve read, the EU side is a mess for whatever reason, maybe NA takes the PPT Competition more seriously and has more balanced populations in some cases, but things like Kodash being Full is a joke.

Oh and I agree, transfers need more restrictions, less freedom to simply move to the top stacked servers; if they’re open and we need more dynamic (and accurate) population calculations.

Tongku.5326:

If you have a system based on the number of received transfers vs the actual WVW activity, then the system will grossly over represent the actual population in WVW as it will falsely assume everyone that transferred in will be playing 7 days a week, entire length of their respective prime time, which is extremely not the case.

We don’t know enough about the current system to say how “WvW activity” is being measured to form the rolling average. We don’t know if the guy who logs into WvW to play for 2-3 hours once a week is counted.

I agree with you, I personally think that someone should be considered an average WvW player (in person-hours or standard population) when they transfer until their game data proves otherwise. When it is determined that they are not, the person is unflagged and the population reading on the server-select screen adjusts immediately (they could be flagged for 1 week or 2 depending). This is to avoid getting players “laying low” so they can cram as many people as possible into the server.

Adjustments in increasing population should be adjusted immediately. This will prevent over-stacking.

Adjustments in decreasing population should be every 3 to 4 weeks. This should prevent blackouts from opening up a full server.

I think the current way it is implemented is only good for decreasing populations…

Decreasing population due to transfer off should be immediate – these people have 0 chance of entering wvw on a server they are no longer part of….. Believe this was Chaba’s idea, but makes perfect sense.

It was silly seeing t2 pops remain the same for several weeks after the gvg guilds left.

That said overstacking before hitting full (to make a server extra over full) could be the bigger problem with hot coming soon.

Why is it always SFR people starting these threads?

BrazenNL.9857:

Why is it always SFR people starting these threads?

I am not on SFR nor a friend of SFR (try to grasp my signature).

Anyway a solution on this problem will reach us (if ever) long after SFR dropped to very high.

But as massive overstacking is still possible (the reason why I wrote this thread), it will hit other servers as well.

This guy is not from SFR, learn to read signatures or at least google if you have no clue……

Greets Myrmi

I guess the only solution is to move away from the server idea…ESO made few things right, WvW campaigns were one of them.

Basically there are active like 10 campaign and player would join what they prefered. This would need some adjustment/balance, might even be tied to a guild level decision, dunno.
But that would resolve population issue greatly.

Gaaroth.2567:

I guess the only solution is to move away from the server idea…ESO made few things right, WvW campaigns were one of them.

Basically there are active like 10 campaign and player would join what they prefered. This would need some adjustment/balance, might even be tied to a guild level decision, dunno.
But that would resolve population issue greatly.

Every campaign in eso has population issues lol

BrazenNL.9857:

Why is it always SFR people starting these threads?

selfowned

If i must guess you are a player from FSP or Deso

@Topic

Yes it is a Problem and some Server already doing this, on our server there are already Map messages that if you transfer you will get gold from them.

And again people overstack a sinlge server and it’s only high at moment, lets see how long the population function need to classify it as very high and then full

And it doesn’t matter which server it is today, or which it will be tomorrow, what worries me is that the population function doesn’t even recognize it not talking about hindering it.