Select Language

A preliminary numbers study


Okay, so one of the big discussions that has floated around about the census sites is that they are inaccurate or not useful. Blizzard has even gone so far as to say the numbers are incorrect. Of course they have also not provided any numbers to back this up so we have no way to know what the definitive 'official' numbers would be.

So we've just plodded along in spite of other's notions that the numbers are incorrect or worthless.

In an attempt to gather even more data to back up the data we have gathered thus far, I started also tracking player activity. The player activity numbers are simply the total counts of characters per realm split up into hours.

For example, if you take a census at 8:15PM (server time) and see 875 players, it would keep that 875 as an entry for 8. And if you again took a census at 9:35 and saw 965 players, then it would keep 965 for 9.

When you upload your data, those figures are averaged together from everyone else who has submitted time info. Those averages have then been plotted on a line graph to display average player activity per realm/faction over the course of a day (no distinction between weekend and weekday).

So we have been saving this data now for a few days and have enough data in a few areas to make some preliminary observations.

First, let's take a look at some overall numbers.

Here are the character counts for characters seen over the previous 30 days who are level 10 and up.

Total Characters: 1,060,382
Total Alliance: 662,912 - 63%
Total Horde: 397,468 - 38%
A to H Ratio: 1.7 : 1

Now, people say, well, this is only characters, and is not a good indicator of the number of people actually playing at that time, and they would be correct in that assumption.

So now, let's look at the TIMES data that has been submitted for all servers.

Average Alliance online per hour: 1137
Number of Alliance submissions: 437
Average Horde online per hour: 667
Number of Horde submissions: 411
A to H Ratio: 1.7 : 1

That 1.7 to 1 look familiar? That's right, it's the exact same ratio we got when looking at total characters seen.

Okay, maybe it's just coincidence or an anomaly. So let's look at more data.

There are particularly 2 realms that have submitted a fair amount of Times information, Bleeding Hollow, and Mannoroth.

Total character data for Mannoroth:

Total Characters: 18,097
Total Alliance: 10,517 - 58%
Total Horde: 7,580 - 42%
A to H Ratio: 1.4 : 1

Times data for Mannoroth:

Average Alliance online per hour: 1516
Number of Alliance submissions: 47
Average Horde online per hour: 1099
Number of Horde submissions: 39
A to H Ratio: 1.4 : 1

See a pattern here? Again, the ratios match up.

Let's look at Bleeding Hollow:

Total Characters: 20,325
Total Alliance: 10,709 - 53%
Total Horde: 9,616 - 47%
A to H Ratio: 1.1 : 1

Times data for Bleeding Hollow:

Average Alliance online per hour: 1168
Number of Alliance submissions: 19
Average Horde online per hour: 1052
Number of Horde submissions: 50
A to H Ratio: 1.1 : 1

Wow, yet again the ratios match.

What does this mean? From a preliminary point of view, it would appear that the times tracking backs up the total characters ratios. What this would mean is that these numbers are indeed correct and reflect the current population trends within the game. Generally speaking, Horde ARE outnumbered.

Proportionally speaking, why do the times numbers match up with the character counts? The reason is because the rule of big numbers is in effect here. When you are dealing with such a large sample size, the numbers should accurately portray the information we are looking for.

In the end, the only way to prove beyond a shadow of a doubt would be for Blizzard to publish these numbers. One way to further verify the numbers shows so far is to collect more data.

I do not see how Blizz can dispute these numbers without providing data to the contrary, and I encourage Blizzard to do so as I feel their numbers will only affirm our theories.

In the interim, I encourage you to continue to take census data with the latest version of Census+ and continue to upload the information to

I am thinking of posting this over on the general forums, what do you guys think?
Anonymous says:
January 26,2005 at 16:22
I definately agree with the conclusions, the only thing that will really skew this in my mind is changes to the server caps. From experience with other similar games such as daoc I expected the average number of characters per account to differ very little between Horde and Alliance, especially considering there is only one unique class per side.

I expect Blizzard to continue to claim that the figures are wrong, and looking at specific servers I'm sure they can make that argument fly but not over all the servers, nor over pvp, normal or even rp servers. They have very much chosen the mushroom approach dealing with customers regarding any matter that may make the game look less than perfect, consequently they have damaged their own credibility severely.

I would post this on both official boards and VNboards (VNboards tend to be up and threads don't mysteriously disappear or crash that often).

I would add a little more about the difference between pvp and normal/rp servers in ratio. Pvp servers that actually have bad ratios will be the ones hardest to deal with, luckily they aren't that many.

Anonymous says:
January 26,2005 at 20:9
I'm just glad all those census surveys I do on Bleeding Hollow do the trick :)

Though I would appreciate if a US player could help with the horde side, being an aussie I can get lots of off-peak player numbers but the peak time ones are a pain.
Anonymous says:
January 26,2005 at 21:43
Hehe, I could use an aussie to fill in the off peak hours for the alliance side on BH ;) early morning hours especially could use some more data.

Anonymous says:
January 26,2005 at 22:16
Yeah noticed that, also seems the little data that is offpeak alliance on BH seems alittle out of whack, over a 1000 players at 5am EST?

Rest of the data seem solid though.

Actually peak isn't too bad, but I can only run it on weekends - which tends to squew the numbers (for the horde) but daytime US is definately a problem.
Anonymous says:
January 26,2005 at 23:30
Yeah, that 5 am entry must be wrong but we'll see after a couple of weeks how good the data is. I'm sure if things look out of whack Rollie will look into what the problem is, a few entries being wrong here and there isn't a big deal and since the process isn't exactly automatic I expect to see some errors. When enough data is collected they shouldn't make much difference and if they do it's not that hard to make the display drop numbers that deviate too much, standard techniques for displaying statistics :)
Rollie says:
January 27,2005 at 1:51
Well, it depends. If there is a heavy Aussie population on BH, then 5AM EST would be normal time for Aussies.
Anonymous says:
January 27,2005 at 7:3
BH is a East server, the only reason I'm there is due to my guild, the 'unoffical' aussie/kiwi server is Blackrock and Proudmore (decide in the first week by a bunch of aussie wow sites).
Anonymous says:
January 27,2005 at 7:56
I've also kept track of people logged on to BH/Alliance for quite a while and even on weekends the numbers start to drop after around 1 am EST, at 6-7 they are at or under 400, that would make the 5 am number highly impropable.

Rollie says:
January 27,2005 at 9:43

Well, in that case I dunno. Unless someone modified thier variables files, that would be what it tracked for that hour.

When we get enough data, I will probably start throwing out entries that are outside of a set deviation to drop anomallies.
Anonymous says:
January 27,2005 at 10:41
Well, my files were screwy there for a while so it might be something from them I guess? I wouldn't have done census at 3 and 5 am (wife would kill me...)... just a thought...

Rollie says:
January 27,2005 at 10:54
I can poke around and see if I can come up with the responsible entries.
Anonymous says:
February 1,2005 at 4:46
Why do people waste their time arguing that the numbers are inaccurate? If you think the numbers aren't right, then fix them! When I started playing on boulderfist, said the ratio was 3.6:1.0 or something ridiculous like that. I knew it was outdated, so rather than whine about it, I took a crap load of censuses on the horde side, and now the ratio is at 1:1.1 If I cared enough, I'd make a character on alliance and do the same thing, since I know that is inaccurate. But hey, all I want are the horde stats.

Lets think realistically though, censusplus is an excellent program for keeping your own database for your realm. However, the submitted data on the website is just untrustworthy. Even when the "last updated" date says only 1 day before, there's no way of knowing if that person didn't take only one census at 3:00am. From that retrospective, I can understand why people are making a fuss about it. If you really want your website to be more successful, you'll have to find some kind of system to show how accurate the data is, rather than just the last-update field. If a person takes a census every 2 hours over the course of a day, you want that data to be more credible than the person that just takes 1 before they log off and go to bed. Or is that sort of like the time-logging feature you talked about in the news post? Whatever, I'm really tired and thats my 2 cents.

Feel free to flame me, I don't know what I just said either.
Anonymous says:
February 1,2005 at 8:40
Ok, I'll flame you, if you actually read any of the the above you'd know we have problems with a 5 am entry. Wife and daughter would kill me if I was up at that hour messing with WoW and I don't blame them for that.

Now, the reason we brought it up at all is for Rollie to be aware of it just in case this isn't just some kinda one time fluke but actually a bug of some kind. He knows about it now and if something similar pops up someplace else he might be able to fix the problem.

I did stay up 'til 3 am on weekend and as expected the value I got there was much lower than the existing one, these erroneous values will disappear in a month so lets hope we'll see good data until then.
Anonymous says:
February 1,2005 at 9:18
Oh, and just to make it clear, since the time stamp is the server time and not actually entered by people doing the census the worry is this.

- Server time is inaccurate, we are actually talking about several servers running a world, each doing certain aspects of it. If the time comes from for example an instance server sometimes it may differ from the time on the "main" server.

- Something may be screwy with a file and the checks in place missed it.

Now all this is to make the data as accurate as possible despite Blizzards mushroom approach.

Anonymous says:
February 2,2005 at 16:56
I happened across references to your census, and the point I anticipate you trying to make is that the Alliance has or will have an unfair advantage in raids and battlegrounds. Is this the case?

Rollie says:
February 2,2005 at 21:7
Personally, I am not really trying to make any point at all other than tracking for curiosity and historical sake.

We'll see if the differences in population make any real difference when battlegrounds come out, but it remains to really be seen one way or another at this point.
daedalus says:
February 4,2005 at 16:28
Hi I just registered. Great job with the census work and posting your detailed methodology.

I only wish I could actually log into the mainboards to slam one poster who claims the census sites do not post methodology. They really need to fix that login bug.

To answer your question, yes I definitely would like you to post your methods, assumptions and limitations on the boards.

Sorry to post here, but like I said, no login to the main boards after about 12PM pst for me.

When you post you might want to consider the following.

There are really two arguments that seem on the surface opposites of each other, but in fact are not even relevant to each other.

One group vigourously denies that the census sites have any validity because their cannot offer accurate counts of players. Some point out that their guild is reported inacurately for example, or their character has the wrong level. In this they are right, after all the information about any one player, or about any "count" of players can be skewed by holes in data collection or other human factors.

The other group is the one that says look, if you take into account moment in time numbers for relative horde vs alliance populations and plot them as a function of time you will get with a high degree of accuracy a trend for horde vs alliance player numbers. Again this is instant in time data based on accurate /who. If you assume /who is accurate then the data plot is correct. Even if /who shorted the number of players by 10%, as long as it shorts consistently for both sides you are still good. A sample is all that is required to plot this kind of independent data.

It is infuriating to see that these two sides are both right, yet not be able to comment on it on the main boards. :) In fact Blizzard is also right, insofar as they take on the position of total player counts, or specific player data being accurate. However blizzard is simply mistaken if they think that the point in time data for ratios is wrong simply becase the data is not exactly accurate on total counts. Even aggregate data taken over time which, if sufficiently large still shows trends fairly accurate as to horde vs alliance ratios.

My point is that if you post you may want to include a note that acknowledges that yes, the numbers for counts of guild players, player levels and total players are not accurate due to procedural issues like people not collecting data on certain servers, or certain time s etc. However the point in time methodology used can provide accurate estimates on ratios because of it only relies on data at the time of the census to establish a count.

Ratios of classes
Ratios of Horde vs Alliance
Ratios of levels

and any other relationship information for which instant snapshots can provide useful data.

Basically if you post make sure you state up front what you are claiming and what you are not claiming. I think it might help to deflect the instant replies from those that attack sites like yours because they cannot accurately display player total counts and focus on what the census today can tell us.

Of course I hope in the future more data points will allow better estimates of total counts.
Anonymous says:
February 18,2005 at 8:41
[quote:13710421ad="Rollie"]There are particularly 2 realms that have submitted a fair amount of Times information, Bleeding Hollow, and Mannoroth.

Total character data for Mannoroth:

Total Characters: 18,097
Total Alliance: 10,517 - 58%
Total Horde: 7,580 - 42%
A to H Ratio: 1.4 : 1

Times data for Mannoroth:

Average Alliance online per hour: 1516
Number of Alliance submissions: 47
Average Horde online per hour: 1099
Number of Horde submissions: 39
A to H Ratio: 1.4 : 1[/quote:13710421ad]

As Mannoroth shows it now:

Total character data:
Total Alliance: 11,649 - 62%
Total Horde: 7,280 - 39%
A to H Ratio: 1.6 : 1

Average Alliance online per hour: 1452
Number of Alliance submissions: 122
Average Horde online per hour: 1091
Number of Horde submissions: 68
A to H Ratio: 1.4 : 1

But then looking at it closer, last horde submission was on the 15th, and they really seem to be dropping the ball on getting hourly submissions into you yet again. (68 to the Alliance's 122) I guess the presence of a Euro Mannoroth server(is there one?) could also help contribute to causing some other numerical chaos that will take a while to sort out on the total player count side of things, like uh, 30 days. Though the alliance number doesn't look too far off. I'm not in game running census for most of the AM hours, and a fair sized portion of the PM ones as well... And I have a census file on Mannoroth that is creeping ever closer to 19,000 chars. :lol:
Anonymous says:
February 19,2005 at 21:58
I have pretty much the same situation with BH alliance but since he's decided to mix EU and US data whatever is displayed there will not be of much use anyway unfortunately :/


Leave a Reply

You must be Logged In in order to post comments.

Advertise Here!