« On the internet everyone knows you're not actually an elf. | Main | PvP Ranks (by Hours Played) »
October 23, 2006
PvP Ranks (Basic)
Eric had recently included a PvP rank scraper into the census script. Because PvP rank is not one of the variables that is returned via the /who list, this data was collected using the method used to scrape character gender. We gathered PvP rank data by moving collection characters to the faction capitals (Ogrimmar & Ironforge). As the character census occurs, the collection character tries to target each character seen in the census. If they happen to be nearby, we note down their census rank. As discussed in the character gender thread, this method has several biases. We're more likely to find the PvP ranks of players who: 1) play a lot, and 2) spend a lot of time in the main cities. On the other hand, given the way that PvP is currently structured (via queues originating in the main cities), the sampling bias may dovetail with the practice of PvP.
We began analyzing the data by trying to get a sense of how well or how poorly the scraper managed to get the PvP rank for all characters on the server. We choose a one week time-frame. Because PvP ranks are updated once each week on Tuesday, we choose a Tuesday to Monday period to analyze.
Overall, the scraper got 50% of the PvP ranks of all characters on the 5 servers. But this percentage is actually deceptive because PvP ranking doesn't begin till the upper levels. The following graph shows the average character level by PvP rank of the character. If we only look at characters above level 45, the scraper found 72% of their PvP ranks.
The distribution of PvP ranks looks like this:
While the scraper did not find the PvP ranks of all characters, there is probably enough to explore the data a little and get a sense of underlying differences.
Server Sample: RP (High), PvE (High), PvE (High), PvP (High), PvP (High)
Sampling Period: One Week in October 2006
Sampling Resolution: ~12 minutes
Parsing Method: The sample unit is each unique character in each hour of the day.
Data Filter: None
Sample Size: 128,477 characters
Posted at October 23, 2006 01:28 PM
Trackback Pings
TrackBack URL for this entry:
http://blogs.parc.com/cgi-bin/mt-tb.cgi/89
Comments
This study will skew towards displaying the distribution of ranks of players who use Battlegrounds to rank up--and possibly away from players who rely on other sources of honor points.
But nice job on the sampling.
Posted by: Tyler Cunningham at October 24, 2006 09:20 AM
Tyler: Why do you think there is a skew here?
Posted by: Eric Nickell at October 27, 2006 09:48 AM
Its very hard to rank up other than the BG's given the amount of honor available that way compared to any other method. I suspect a good portion of the missing ranks may be players who do not engage in PvP.
Posted by: Tom at November 1, 2006 02:45 PM