Monday, January 13, 2014

Putting The Graph Into Context: 2006 - 2012

I've now read all of the entries for Blog Banter 52 through Khanid Kataphract's entry and I have to say I'm impressed by the quality of writing so far.  I'm also glad to see that some bloggers have used the data set to help explain their thoughts.  Following in Mabrick's footsteps, TurAmarth and Splatus dug deep and extracted some good information.  Not that they were the only ones using the data, but they used some different data that compared EVE and EVE's potential audience to those of other games.

While this post is not a part of BB52, I thought I'd do something I haven't seen yet.  That's right, combine the data from Chribba's chart with the data set.  So instead of playing EVE, I geeked out and played with spreadsheets for the past few days.  What follows is the result.

First, how did I get Chribba's data?  I didn't need to ask him, as I wanted to create charts that others can duplicate independently.  So instead I went to the source page for the chart and extracted the information.  The information in the chart is the average concurrent users1 for a week.  But in order to match up the data with the subscription data2, I needed to convert the weekly data into monthly data.  I did that by pro-rating a partial week (like the first or last three days of a month) with the average for the entire week.  If Chribba were to run the numbers he'd come up with something different, but for my purposes I think the method worked out well.

We have 6 years of data

Did I forget to mention that we have 6 years in which we have both concurrent users and subscription numbers?  I think that's unheard of in the MMORPG industry.  I'm sure the industry veterans that CCP hired following the Summer of Rage were flabbergasted to find out that CCP was so confident that they allowed the public access to that type of information.

So what does putting both the subscription numbers and the average concurrent user numbers on the same graph tell us?  For one thing, the numbers were going up steadily from May 2006 to January 2011.  That bump in the concurrency numbers between November 2008 and May 2009?  I attribute that to Quantum Rise and Apocrypha launching 4 months apart.  If not for that I think the blog banter question would ask what has happened over the past three years, not four-and-a-half.

The available data for the blog banter period

The second graph shows just the period Kirith asked about in his Blog Banter question.  For this graph I included the data for both the number of subscriptions and my calculated average concurrent users for the month.  The striking point in the graph occurs after January 2011.  At that point the average concurrency and subscription numbers start to diverge.

So what happened?  I am not going to credit the War on Bots for the decline as the current forever war on the mindless beings did not begin until the end of March 2011 and the decline in concurrency was evident in February.  I can only think of two events that could explain the decline.  The first was the defeat and breakup of the Northern Coalition.  I had heard that NC participation declined greatly and perhaps the concurrency numbers took a hit.  The second is the launch of the new Incarna character creator on 18 January.  I remember a lot of players posting on the forums that they were going to leave the game because of the upcoming expansion.  Apparently that happened immediately and the subscription numbers took some time to catch up.

But that was the story the graphs told me.  I wondered if the difference in scales were misleading so I made a third graph.

Figure for all accounts, not just those that logged in

The average hours played per account in the above graph doesn't refer to the length of the play sessions of those who logged in.  The average includes all paying accounts.  So for example, in May 2006 all 120,354 accounts would have needed to log in for an average of 21.2 hours per week in order to register the 15,207 average concurrent users recorded that month.  In May 2012, that figure had fallen to 12.8 hours per week to generate an average concurrent user mark of 27,764.

So what happened?  Did people get older and have less time to play?  Did the number of alt accounts lower the average?  Did EVE just have more bots in 2006 than in 2012?  Or is the answer a combination of all three plus factors I've neglected?  I suspect the latter is the correct answer.

The numbers are interesting, but don't tell the whole story.  In fact, because CCP stopped releasing information to after May 2012, we don't even have complete information for the entire period the blog banter covers.   The lack of information for when the War on Bots really picked up is quite annoying.  From May 2012 onwards, analyzing Chribba's graph requires examining the events in EVE as well as some of the design decisions CCP made over the past 18 months.  But as this wall of text is too long already, I'll leave that explanation for another post.


1.  CCP uses the term "concurrent users" when referring to the number of active accounts logged onto any of their shards.  Concurrent users is not the same as concurrent number of unique players.

2.  When I use the term subscriptions, I am not referring to the number of actual players.  The number of subscriptions is the number of active paying accounts on the Tranquility server.


  1. Drop in average hours played per-account per-week? One reason could be a relative increase in the number of "alt" accounts that just login for short periods of time for very specific reasons (if at all) - compared to main actively-played accounts.

    Some things off the top of my head...

    You have to have a separate account for your cyno alt. So is this a side-effect of more and more players moving up into cap ships?

    More dedicated capital-ship accounts, that only log in for specific ops?

  2. Great post Noizy, I look forward to your subsequent analysis.

    "So what happened? Did people get older and have less time to play?"

    EVE has always attracted the older gamer, but it would be interesting to see how many veteran players have adapted their playstyle to suit the changing demands of their lives. I certainly identify with this phenomenon, having found marriage and parenthood impacting on my 10 year tenure as an EVE player. My time (and desire) to play have changed drastically, yet I'm still here. I wonder if this is widespread and exactly how that impacts EVE's figures.

    As you point out, EVE is an unusual case in that it has existed long enough and with a significant loyal user base for ageing to be a possible factor.

  3. How do you manage to go from accounts/PCU averages to average number of hours played per account per week? I have been interested in that number myself, but as far as I know you need more than just PCU and account numbers to get that. You'd need to know hour-by-hour the number of logged in users, which leads to graphs such as those on and

    Does anyone keep an archive of the number of logged-on users hour by hour going back over some period of time? If so, the area under the curve each day could be integrated giving an accurate number of player-hours per day. THAT number could be used to generate meaningful time-played per account statistics.

    It's also interesting to view the daily logged-on user graphs to pinpoint what timezones are more active than others. I seem to recall seeing somewhere that the majority of EVE's players are from North America, but never have any of the NA timezones ever coincided with the daily PCU number. That happens every day without fail in the European timezones. Maybe CCP should concentrate on getting new users outside of their traditional "peak" timezones. They could increase the playerbase substantially - outside of GMT+1 and GMT+2 - without significantly altering the maximum daily PCU number.

    1. To get the average logged on, I took the hourly average concurrent users that I calculated from the weekly totals from Chribba's chart and converted that into a monthly average. I then multiplied that total by 24 in order to get the number of hours played in an average day. I divided that total by the number of subscriptions for the month to give me the average hours played per account per day. I then multiplied the daily average by 7 to get the weekly average.

      Of course, if I were smart, I would have just developed this formula.

      168 * average hourly concurrent users in a month / subscribers for the month