WarcraftRealms.com

Hmmm, the activity ratio given on each server may need a bit of tweaking, looking at Bleeding Hollow it says 1.5:1 but if you bring up the activity graph there is no single time that has more than 1.4:1. The average can't really be higher than the highest ratio...

I suspect it's just an average of all samples and since BH alliance samples are primarily taken by a player on EST times and the Horde samples primarily on australian time you get this. I would suggest playing around a little with this.

The total ratio might not necessarily be the most interesting either as all servers pretty much have times where there are no samples, usually not the same times on both sides. What is probably interesting though is prime time ratios, I realize that would be somewhat skewed for servers with high populations from timezones outside the US but there are in general a higher sample density at these hours for the most part giving better numbers. Taking the average number of all samples 20-23 on alliance and comparing to the average of the same hours on horde side or even doing it for each hour and then averaging the resulting ratios for example.

Yeah, I had noticed the same thing and agree that it needs some tweaking. Good suggestions on showing prime time averages instead of overall. The question then becomes, what exactly is prime time =)

I would say prime time would be 6 - 10 server time seems to be the most popular 'prime time' for play.

I would have gone with 7 - 11 but 6 - 10 works too, I doubt there will be any difference which exact span used there within the 6 - 11 range really.

I honestly don't know that "server primetime" is a good time to use. If Blizzard had left the timezones on the server select screen, this would make sense, but since they are long gone, people from every time zone are playing on every server. Also if you look at the server transfers, almost every server on the list is swapping timezones, IE PST to CST. This makes it even more evident that large amounts of people are playing on servers other than their time zone. Another point is that it would make more sense to have a "universal primetime" so the time is uniform when you compare servers.

You just need to average across the 24 hour period counting each time period as one 'sample' rather than each census result as its own sample.

So you are averaging the ratios for each hour rather than the average of every single sample submitted.

The problem with that is that a lot of times are missing on one side for some servers, if you want to do that you'd first have to check which hours actually have samples on both sides for the data to have any meaning.

Using prime time samples is a way to try and use data where there are generally a good many samples and consequently data that is more reliable in a statistical sense.

I did change it to average the time averages, and am playing with some numbers to check the 'prime time' averages as well. I am looking at 4 - 10 PM server time as prime time.