Showing posts with label server population. Show all posts
Showing posts with label server population. Show all posts

Thursday, February 27, 2014

[GW2] Server Populations and Achievement Scores

After observing large differences in Marionette success rates across servers, I've become interested in understanding a bit more about server populations.

So the achievement point leaderboards were of particular interest to me.

Players earn achievements by completing milestones, participating in living story events, and completing dailies. They're a great measure of player activity, since you get achievement points for just about everything. Even better, ArenaNet publicly tracks achievement points for the top 1000 players on each server.
Here, I've plotted the achievement points of the top 1000 players on each server.  The servers are organized by the average achievement point score of these players (highest top left to lowest bottom right), and colored by their WvW season 1 tier (gold / silver / bronze).

There are vast differences between servers.

As a snapshot, the leaderboards are interesting.  But tracked over time, they contain much more information. For this, I collected data from the leaderboards during the Edge of the Mists update. This gives three measures of server activity:  (1) transfers, (2) inactive players, and (3) points gained among the active players.

Server Transfers: Jumping Ship
Server transfers show up on the leaderboards in two ways. A player could appear on one server, and later on another. So I can track accounts that move between servers.

Other players disappeared from the leaderboards. So, for example, if a player on Eredon Terrace had over 4217 points at the start of the Edge of the Mists update, they should appear on that server's top 1000 list unless they transferred off the server.
Overall, 1.6% of players listed on the leaderboards transferred servers. Given the small number of transfers from each server (and the reasonable expectation that some of the 'large' moves are due to guilds) it is hard to make a general pattern from the server transfer data, but I'll look at it again in future updates.

However, of the transfers I observed, 84% are to servers with higher average achievement point scores than the starting server.

Inactive Players
Overall, 17% of players on the server top 1000 didn't earn any achievement points during the Edge of the Mists update, and therefore likely didn't play GW2 during the update. Here are the percent of inactive players by server:
Inactive players are more common in lower-tiered servers, with the Bronze tier averaging over 23% of players inactive, compared to an average 12% for the gold servers.

Points Gained by Server
During the two-week update, active players on the leaderboards gained an average of 170 points, around 12 points per day. Pretty impressive, considering that the daily requires only 5 points. (The average top 1000 player earned about one-third of the total possible points to be earned from living story / daily content.)
Across all servers, players with more achievement points earned more points than those with fewer points (rich get richer faster).

Several Thoughts
By now, it is clear that there are large differences in server populations and activity. It is simply a myth that there are 'pve servers' that just happen to do poorly in WvW. Players tend to transfer to higher-tiered servers. Those servers have fewer inactive players. And the players on higher tiered servers tend to be more active.

Time to think about a server transfer myself...

Sidenote: the results presented here don't change under more complicated statistical models (which I ran, but excluded from the presentation because this is a long post already). Though, I did have to run a nonparametric model due to the interesting structure underlying the data.

Monday, May 7, 2012

The population numbers behind server status

All too often, I find myself pointing players to TorStatus to get a picture of what SWTOR's server population looks like.

Trouble is, knowing that a server is 'standard' (code 2) doesn't tell you much about the game experience. If a server can be 'standard' with 100-500 players, that is a different world than if the same code means 1000-3000 players online.

To settle this a little, I set up a monitoring program to alert me when a server changes its status as listed on SWTOR.com. When that happened, I would log onto the server and /who everyone within level brackets until I'd collected data on both the Republic and Empire sides. I did this just over 100 times.

From this, it looks like the cut-points for each server status are:
Light >> Standard:  500
Standard >> Heavy: 1500
Heavy >> Very Heavy:  2250-2500(?)
Very Heavy >> Full:  3000

There are a couple ways to use these cut-points to give context to the TorStatus numbers. For example, in April, servers were listed as 'light' 65% of the time. But players aren't evenly spread across the servers. While the highest population servers were 'light' under 30% of the time, the bottom 46 servers are listed as 'light' 90% of the time or more.  This means the bottom 46 servers spend 90% of the time with less than 500 players on both sides (combined).

As a side note: the average server status figures for the last 3 months (and a bit of May) are:

Average Server Status by Month
February March April May
1.49 1.38 1.29 1.19

The drop from February to May (a 20% decrease) closely matches the drop of subscriber numbers over the same period (from 1.7 to 1.3 million: a 25% decrease).

Note:  I'm doing some checks to examine whether those cut-points are stable over time, and do more to pin down the limits on the 'Very Heavy' status. That, combined with a few crazier statistical models, should help to put actual numbers to the server status figures.

Wednesday, May 2, 2012

LFG: How Many Make a Party?

In what I can only describe as fantastic news, it looks like a group finder is coming with patch 1.3 (via both the April 27th Q&A and earlier requests for feedback). Unfortunately, the initial release of the system seems limited to matchmaking within a single server (source).

But what about players on low-population servers?  Do they have the population to support a reasonable LFG system?

Ideally, Trion or Blizzard would hand me a large data set on player behavior, and I'd get to go to town analyzing segmentation in the MMO playerbase and its influence on the availability and fun of small-group content. Mmmm... happy data dreams...

Absent that, there is another way to get some insight on whether low-population servers in SWTOR can support a LFG system: SIMULATION!

The basic idea is simple: I wrote a program to generate test data given some assumptions about how players behave. I then ran that program thousands of times, with different assumptions about player behavior, to see how the results changed.

The 'Perfect' Balance
For example, suppose that SWTOR was a utopia - the holy trinity of roles was distributed perfectly across the lvl 50 playerbase: Tanks were 1/4 of the population, Healers were 1/4 of the population, and DPS were 1/2 the population. At regular intervals, one of these players (randomly chosen across the three roles) enters the que. Imagine further that players were perfectly patient, never ditching the que. Using this, I can set up 100 sample ques of 1000 players, and take a look at the average wait of each.
Egads! Wait times spiral off into infinity - and beyond!* As players enter the que at random, a surplus of players for a role develops, causing the wait times for players of that role to increase.

So impatience (or limited playtime) is a great thing for a LFG system.

Impatient Players
Now suppose that players are willing to wait patiently for some amount of time, and after their patience wears thin, they have a 20% chance of dropping the que for each extra minute they wait. LFG que waits dramatically decrease:
The surplus of players for a role gets evaporated by the que droppers.

But just how much (im)patience is needed for the shorter, more stable ques?
A little impatience goes a long way.  Even if players are willing to wait half an hour before they slowly start dropping, just that little bit of impatience lowers que times by about 10 minutes... and it does this without throwing too many players out of the que - 92% of players still get placed.

Distributing Time
So far I've assumed that players stroll in at even intervals. The population of a server doesn't look like that, though.
Server population peaks late at night, and disappears during the day (approximated by the plot on the left). The effect on the que is clear: que times shorten during peak hours - dropping to less than 5 minutes - shown by the smoothed averages in the top right plot.  Unfortunately, this comes at a cost: ques all but disappear during the day - only a handful of players get groups from 4am to noon (go to sleep!).

Tank Shortage
All of this leaves aside at least one crucial component of ques: tanks and healers tend to be rare.

To take a look at this, I'll run the same simulator, only I'll vary the percentage of tanks and healers from 1-25% (and all combinations).  For each of those 625 models, I'll look at the percentage of players queing that obtain groups, and look at the average wait time of DPS. I'll do this for a server of 500 players queing, 1000 players queing, and 10,000 players queing.  For ease of presentation, I'll assume that players are moderately impatient - waiting 15 minutes before they start dropping the que.
Moving from left to right across the facets, as the percentage of healers and tanks increases, wait times generally decrease, and the percentage of players getting groups increases (gasp!).

The most remarkable effect, though, is that just having a large pool of players to pull from keeps wait times down.  Regardless of the percentage of healers or tanks, wait times for a pool of 10k players stay around 10 minutes - about half that of the pool of 500 players.  Balancing the percentage of healers/tanks increases the percentage of groups that form.

Amazingly... if tanks are truly scarce, only 1 in 5 dps that que will get a group, which is absurdly low, so low that I wonder whether the tank shortage is smaller than most people suppose.

So can low-population servers support LFG?
Even on the lowest-population servers, LFG will help match players during peak play times

But on those servers, unless a large percentage of players are willing to play tanks/healers, wait times are likely to be lengthy, and a high percentage of players are likely to drop group in frustration.  (Expect some heavy forum QQ from those that do... but just remember that they're helping to keep wait times down for everyone else!)

As always, serve with a large heap o salt: there are a ton of variables I didn't model, or modeled pretty simplistically. If anyone knows a good source of data on players to help reduce some of that uncertainty... I'd love it.

Sunday, April 22, 2012

Slicing for Augments in 1.2


I just want to say one word to you.  Just one word.  Augments.

Despite data-nerding-out about lockboxes, they aren't my main source of profits with 1.2.

I've been selling augments on the GTN as fast as I can make them, and the profits are absurd. An augment that costs less than 20k credits to craft will easily sell for 60-100k on my server. I'm considering buying the legacy-unlocked neutral GTN...

With 1.2 augments are no longer obtained via slicing missions.  Instead, they are crafted by the following professions:
  • Armormechs: Aim, Cunning, Shield, and Absorb Augments.
  • Armstechs: Endurance, Surge, Critical, Accuracy, and Power Augments.
  • Synthweavers: Strength, Willpower, Defense, Alacrity, and Presence Augments.

Getting the Patterns
These patterns are only found with crits of 'Sliced Tech Parts' missions. 95% of the time, these will give to 1-2 schematics, though there is a 5% chance of no pattern on a crit (ht to GnatB on the forums for observing this first).

The patterns teach a premium (green) quality schematic.  RE'ing the premium items gives a 20% chance to research a prototype (blue) schematic; the blues have a 20% chance on RE to research a purple (artifact) schematic.

Gathering the Materials
In what follows, I'm going to focus on the top tier of augments (the 22s). While 25s are listed on Torhead, these do not appear to be in game.

In addition to some base components from the appropriate gathering profession (arch for synthweavers or scavenging for armor/armstechs), the blue augments require 4 Subelectronic Data Module and the purple augments require 4 Advanced Neural Augmentors.

The bottleneck for crafting the top-tier augments is the Augmentors.  At the moment, these can only be obtained via the 340 mission discovery 'A Test of Intelligence' or crits on the Grade 6 abundant missions Mysterious Funds / Devoted to Duty. I've listed the returns from each mission in the table below:

Blue (reg) Epic (reg) Blue (crit) Epic (crit)
A Test of Intelligence 8 4 10 5
Devoted to Duty 4 0 6 3
Mysterious Funds 4 0 6 3

After running just over 200 of these missions on companions with 10k affection, I'm getting a 20% crit rate.

Since the missions cost 1,780 credits to run, at a 20% crit rate and 3 purples per crit, that is about 1068 credits per purple material, or 4272 credits in slicing materials per purple augment made. Adding the cost of the vendor mats (1600 credits from vendor, 600 credits if gathered via a crewskill) and the AH price of the gathered components (around 500 -1500 per component), that puts each augment at around 9-17k credits to craft.

Given the prices on my server right now (60-100k), that is a phenomenal profit.

What is the catch?
There are two downsides to crafting augments.

The first is time. Each slicing mission takes just over an hour to run. That makes for 7 hours of slicing missions per augment. With only 2 tech parts missions available at any given time (3 with the discovery), this works best with several characters with slicing (or, selling in low volume).

The second downside is that players can recycle their augments, removing them from their gear for 22k credits. Eventually, this may serve to limit the demand for augments.

One more upside
In the near future, the market for augments is likely to grow. Right now, using custom gear with augment slots means losing set bonuses on Columi and Rakatta gear. As more characters acquire Campaign and Black Hole gear, there is no downside to picking up an augment slot for better stats, widening the market.

Happy credit hunting.

NOTE:  Take a look at the cost of the [Augmented] custom gear on the GTN - made via crit-crafting an orange item. At least on my server, even the lvl 11 oranges are going for 50-100k if they have an augment slot. (Another remarkable profit - since the items cost ~3-5k to craft each).

Friday, April 13, 2012

Server Populations: What Difference a Day Makes?

In my last post on server populations, I noted that there was some support for the claims made by players on the forums pushing for server transfers. Low-population servers had very few players online, at levels that would make group-based activities exceptionally difficult. Recognizing this, the disproportionately large numbers of low-level characters on high-population servers hinted at players migrating to those servers.

But my data was fairly limited: just a slice of a handful of servers on one Friday night by /who-ing level ranges. As Shintar pointed out, there are reasons to wonder about whether Friday fits other nights.

So do the numbers hold up after looking at other nights?
Much like TorStatus, I've been collecting the server status of the SWTOR servers for a few months now. The above plot tracks the average server status for East- and West-coast servers for the past month. The shaded areas cover 'primetime' - 6:30 to 10pm within each timezone - and the black bars are the average server status across primetime.

Friday looks a bit like a weeknight, with similar numbers to Mon-Thurs. There are definitely fewer players than on Saturday / Sunday, where the evening playtime window is much wider. But from server status alone, it looks like a representative night.

But server status isn't population.

I re-collected the original data for the same servers, this time during primetime Tuesday. I'm primarily interested in three potential differences: (1) whether overall populations differ, (2) whether the population gap between high- and low-pop server differs, (3) whether the numbers of 50s differ.
Consistent with the server status data, the average population of a server varies remarkably little between the two nights (the colored vertical bars in the top left).

But the average masks striking differences between the two nights. On Tuesday, the higher population servers have relatively fewer players, while the medium population servers have relatively more players, slightly decreasing the gap between the servers. On Friday, the typical high-pop server had 4 players for every 1 on the average medium-pop server.  On Tuesday, this gap shrinks to 3:1.

Unfortunately, this trend doesn't extend to the lowest-population servers, whose small populations remain virtually unchanged - with on average 3:1 players between medium and low servers, and a 10:1 high to low player ratio.
Perhaps unsurprisingly, the reset of the operations lockout on Tuesday sees relatively more players on their 50s for medium and low pop servers. High population servers retain an abnormally high percentage of low-level characters, most likely due to rerolling players.

As someone on a medium-pop server, I find these numbers comforting. The increase of 50s and overall population between Tuesday and Friday closes the gap between medium and high population servers, something I view as essential for server health.

But there is small comfort here for players on low-pop servers. Hopefully the uptick in the numbers of 50s on Tuesday is enough to coordinate group activities at endgame. But the bottom 5 servers don't see the same increase in overall players as the medium population servers, and the number of 50s online is still dwarfed by every other server.

Tuesday, April 10, 2012

Galactic Census: SWTOR's population

The vastness of space, without some to experience it with, is transformed from something magical and awe-inspiring... into something empty and lonely.

The SWTOR forums are filled with complaint threads about dead servers and players rerolling to higher-pop servers in order to continue playing the game. (And begging for transfers / mergers... or even just LFG.) Because the forums can be a negative place, I'd assumed that most of these posts were written by players who just weren't having much fun in the game.

But I'm a data nerd, so I did some digging.

Unfortunately, there isn't great data on SWTOR server populations. TorStatus reports the average status of a server listed on the SWTOR site, but it isn't quite clear how that relates to the actual population, and it can't break apart the balance of players inside a server. (Plus, what counts as a 'Heavy' pop server could change over time, masking a loss of players.)

I created characters on 22 different servers during Friday primetime. I selected 5 servers that were high pop, 5 low-pop servers, and 12 servers in the middle. On each server, I recorded the number of characters within 5 level brackets (1-10, 11-20, 21-30, 31-40, 41-49) and the number of level 50s.

Before I start, I want to caveat EVERYTHING in this post with the observation that this is just an initial sample of data. I've got much more planned as I explore this topic.
There are vast differences in the populations of servers. Just as an example, there were more Empire characters between level 11 and 20 on The Fatman server (420) than there were characters of both factions and all level ranges on Rwookrrorro, Thendys Noori, and Cho Mai combined (417). The smallest servers are empty places - averaging about 160 players total - compared to an average of well over 2000 players on the highest pop servers.

Even more startling is that the data suggests players are migrating to the high-pop servers.
The high population servers not only have more players, they also have more players rolling new characters. On most servers, approximately 1 in 3 of players online are level 50. On a top 10 server, that number shrinks to 1 in 5. High numbers of low level characters suggests a large number of either rerolling or new players on these severs.

To give a sense of what being on a low pop server is like, here are the chat logs from 2 minutes on Ord Mantel on the highest population server - Fatman - compared to 15 minutes on a low pop server.

This wouldn't be so alarming... except that participating in most endgame activities requires a large number of like-minded players. Even a simple Huttball match takes at least 16 players in the que. That is half to one-quarter of all the 50s online for the smallest server in my sample. A ranked WZ system on that server - or a 16-player world boss - is absurd.

Small numbers become even harder when there aren't socializing tools in-game. While cross-server LFG is typically blamed for undermining server communities, when the community on a server gets small enough, the absence of a LFG tool isolates those players even further, ironically turning an MMO into an isolating experience.

Worse still, imagine guild recruitment to replace someone in an operations group.

Again, a word of caution: this is just a snapshot of a few (mostly pvp) servers. Until I have more data (working on it) the findings here are more intended to be thought-provoking and give a little more context to ongoing discussions about the state of SWTOR.

That said, I really feel fortunate to be on a medium population server with a stable set of guildmates who remain interested in operations and endgame shenanigans. But for those who aren't, I hope Bioware resolves these issues... quickly.

Tuesday, February 28, 2012

Hitting Moving Targets

One thing I'm interested in is understanding how accurate any population rankings I observe can be (from the previous post).

Just taking a quick glance at TorStatus, its pretty clear that the population of any given server can be... well... lumpy.  Case in point:  using the same basic methods of TorStatus, but averaging the server status of all East Coast and West Coast servers (US) and across a month, the overall server population looks like this:
Not only would that be an awesome roller coaster ride, it suggests a few things about the populations of servers.  (Numbers in military EST.)
  • Differences between nights aren't as large as you might expect.  I'm pretty sure that if I ditched the weekday labels, it'd be hard to match the charts to days of the week.  (Other than the Tuesday downtime.)  Saturday looks an awful lot like Monday.  And even though Sunday is the heaviest play time... it isn't by much.
  • Time of night matters much, much more.  Just for example, the difference between 7 and 9 pm PST is huge on a West Coast server (22 vs 24 on the charts).
What does that mean? I sampled the data for the last post over three hours on Saturday night, so I'm understandably nervous that the time it took to observe the server populations could have an effect.

Too check this, I took a quick glance at how closely the data I observed matches to the listed server status, to get an idea of how accurate my earlier observations were.  First, looking at the status at the time of sampling (triangles), 15 minutes before (circles), and 15 minutes after (squares):
Thankfully, none of the servers I examined changed their status around sampling time.  Even better - only Infinite Empire (top left, green) shows a mismatch between the server status and the observed population, with more people than would be expected.

This also gives some idea of the cut-points for each server status - or at least a nice guess for where they might be, sitting at around 500 and 1500ish.  (Hopefully I can get a bit more data to pull that together.)

While the observed data broadly match the server status - the same can't be said of the rankings between servers over time. To show this, as I collected the data, I sampled two servers (Prophecy of the Five and Veela) at multiple times as I collected the data.
Because I'm drawing the data from different times, the observed total population of each server moves pretty dramatically across the scale. Unfortunately, this makes answering whether Po5 has a higher/lower pop than Mandalore something I can't tell with any confidence.

But instead of being concerned about server size... what about the balance between factions?  How much does that move over the night?
It shouldn't be any surprise that the server populations are smaller pre-primetime. Curiously, that doesn't seem to have any effect on the balance between factions, which doesn't show much movement between times.

TLDR:  Server populations are consistent across days, but change pretty dramatically over the course of the night. These swings make estimating the relative size of servers pretty difficult. Fortunately, these changes don't influence the estimated balance between factions.

-myth


*note:  as with the previous post, The Harbinger isn't a PvP server, its just there for a point of reference.

Monday, February 27, 2012

PvP server population

Hazards of reading cool stuff on the internet.

Sometimes, I get inspired to do things.  Absurd, fun, crazy things.

Case in point: I read a post by Dulfy on PvE server population and balance.  She created two level 1 characters on each server and used the /who list to record the population numbers on six planets plus the fleet during prime time on each server (6:30-10:30 EST for East Coast servers and 6:30-10:30 PST for West Coast servers).

I just had to know what the data looked like for PvP servers.

Four hours later...

These charts replicate the server pop / balance data presented by Dulfy.
- PvP/RP-PvP server pop ranking by
- PvP/RP-PvP server Imperial::Republic Ratio
- PvP/RP-PvP server Ilum Ratio
- PvP/RP-PvP server pop ranking by planet

NOTE:  to aid comparisons, I've added The Harbinger, a pve server.

In a follow-up post, I'll extend some of this analysis and comment on how reliable I think these numbers are.

PvP/RP-PvP server pop ranking
 PvP/RP-PvP server pop ranking (West Coast)

 PvP/RP-PvP server pop ranking (East Coast)
PvP/RP-PvP server Imperial::Republic Ratio
PvP/RP-PvP server Imperial::Republic Ratio on Ilum
PvP/RP-PvP server pop ranking
Note:  this chart includes information on additional planets I didn't use above (to match Dulfy's data)
For the extra geeked:  data here
(For the data - I ran page 'SWTOR pop' on Friday and 'sample 2' on Saturday.  Several servers were re-sampled, and the numbers on the side are me getting the fleet sums correctly.  I know documentation makes the data, so definitely ask if its confusing.  Just ask.)


As always, let me know if I missed anything.
-myth