It seems that my statistics report is wrong. I continued to gather data and I’ve made a few discoveries. Now, the second part of that report, about the Horde/Alliance ratio, about the races and about the classes, is still correct as much as possible (since it revealed just the trend and there’s a considerable variation on each server). Instead is wrong my schema about the servers flags: high, medium, low.

To figure out that data I simply collected the total number of players online on the servers that were on the margin between different flags. For example I looked at the lowest server marked “medium” and got the total of players: 590 (as an example). Then I took the highest “low” flagged server and took its total: 412. So I established that servers with more than 500 players were marked as “medium” and servers with less then 500 marked as “low”. This because I started from the premise that those flags define the current load on that server.

Instead it seems (I’m still collecting data) that my premise is false. The servers flags don’t show the load of each servers based on fixed values. What the system seems to do is to count the TOTAL of players logged in on EVERY server. And then distribute that number to establish in which of the three groups (high, medium, low) a server fits. So these flags represent a proportion between the servers and the overall logged in population.

This means that there will ALWAYS be about 30% of the servers in each group. Even during the morning and the day when the number of players decreases sensibly (off-peak).

As an example there could be in total just 400 players logged in and the system would mark as “high” those servers with more than 4 players.

So the system doesn’t show at all the active load on the servers. It just shows the distribution of the players throughout all the servers. The only fixed data we have is that the server cap is set around 3200, 3500 players. After that limit you get a queue. But about the distinction between the other two flags we cannot know.

As another example, today during the peak the flags worked this way:

3400-2500 -> high
2500-1700 -> medium
1700-0 -> low

You can see how far from my first report. So there are still two possibilities (I’ll know tomorrow when I finish to gather the data again):
1- They changed the old values and adopted those I collected here.
2- Those flags work on a proportional schema that cannot be fixed, as I explained above.
And in the second case they did something extremely smart.