PlayOn: June 2005 Archives

« May 2005 | Main | July 2005 »

June 29, 2005

Grouping and Level Progress

Server Sample: RP (High), PvE (Medium), PvE (High)
Sampling Period: 6/13/2005 9:00 am - 6/20/2005 9:00 am
Sampling Resolution: ~14 minutes
Parsing Method: The sample unit is each unique character. Each character was tracked across the server logs. Total playing time, lowest observed level, highest observed level, guild affiliation, and zones seen in were parsed.
Data Filter: None
Sample Size: 76,364 characters

Do characters who group more often level faster? We split characters into 4 bands of grouping ratio and then plotted their average progress over the levels. In other words, characters in the 0%-1% band were almoost never observed to be in a group. Average progress was calculated by subtracting the lowest observed level from the highest observed level. Characters who group more often appear to level faster but more so at lower levels than at higher levels.

We plotted the same graph but this time controlling for playing time. For characters between level 1-20, the correlation between group ratio and level progress is .13. For characters between level 21-40, the correlation is .04. For characters between level 41-60, the correlation is .03. In other words, characters who group more often do level faster at lower levels, but this is not the case for higher level characters. The graph below seems to show this break-even point occurs at around level 40.

Of course, grouping for instances increases post-40 and the decrease in levels may be offset by rare item acquisition from instances.

Posted by nickyee at 10:31 AM | Comments (3) | TrackBack

Server Population by Time of Day

Server Sample: RP (High)
Sampling Period: 6/13/2005 9:00 am - 6/20/2005 9:00 am
Sampling Resolution: ~14 minutes
Parsing Method: The sample unit is each unique character within each hour of the day.
Data Filter: None
Sample Size: 373,381 samples

For every hour of the day over the 1 week period, we tabulated the number of characters observed. The time shown on the x-axis in the graph below refers to Pacific Daylight Time. Server peak is 7pm and lowest at 3am.

Posted by nickyee at 10:30 AM | Comments (4) | TrackBack

June 27, 2005

Class Abandonment

By calculating the class distribution at early levels and then again at high levels, we can get an approximation of what classes are most likely to be abandoned. Of course, because we aren't tracking the same characters, part of what the data below shows may be responses to class changes over time rather than class abandonment itself.

Here are the two class distribution graphs - one for level 1-20 and the other for level 41-60. If we assume that class distribution at high levels compared with low levels reflects abandonment, then Warlocks are the most abandoned class. About 27% of who create Warlocks abandon their characters before they reach level 40-60. This is followed by Shamans (24%), Warriors (21%), Mages (17%), Priests (17%), and Druids (5%). Several classes had higher percentages at high levels than low levels and thus are seldom abandoned. These were the Paladin, Hunter and Rogue.

(I'm surprised by the abandonment of Shamans because Shamans seemed like a strong class from what I had read in the forums. Perhaps the Shaman drop is an artifact of it being a Horde-only class in some way?)

Also keep in mind that players choose character classes according to personality and attitudes, and that players who choose Paladins are likely to not be the kind of player who would choose to be Warlocks for example.

Class distribution for level 1-20:

Class distribution for level 41-60:

Addendum:

After reading over the reader comments, I realized it makes a lot more sense to split up Alliance and Horde for the analysis otherwise the lower Horde numbers get skewed by the large Alliance numbers. So here is data again but split for Alliance and Horde.

So there is no shift among Horde chartacters. And the drop in Shaman percentage goes away (thanks Jason). The interesting thing is that the shifts are driven entirely by Alliance characters. Compare the drop among Warriors and the increase among Paladins.

Posted by nickyee at 05:10 PM | Comments (11) | TrackBack

June 24, 2005

Race and Class

Humans are the most popular race (25%) while Orcs are the least popular race (7%). The Alliance outnumbers the Horde 2 to 1.

Class distribution among Alliance members ranges from 9% (Warlocks) to 15% (Warriors).

Class distribution among the Horde has a wider range - 7% (Druids) to 20% (Warriors).

Posted by nickyee at 10:57 AM | Comments (0) | TrackBack

June 22, 2005

Level Distribution

Because level distributions change over time, we wanted to capture the current level distribution as well as present it here on the blog. 14% of observed characters were level 1. 25% were level 5 and below. 12.4% were level 55 and above. 9.8% were level 60.

By plotting only those characters above level 5, the spiking effect at every 10 levels becomes a little more noticeable. There are more characters at every 10 even level than the level preceding it. This is most noticeable for the 39-40. There are about twice as many level 40 characters as there are level 39 characters.

Posted by nickyee at 01:54 PM | Comments (4) | TrackBack

June 21, 2005

Playing Time

What's quite striking about the plot of playing time over the 60 levels are the spikes right before every 10 levels and the sudden drop right after. This is most prominent right before and after level 40 (for the mount in addition to the new skills). The pre-40 spike has a ramp-up starting at level 37 that the other spikes don't have. Level 39 characters were played on average 17.2 hours over the sample period compared with 12.9 hours with level 40 characters. In other words, level 39 characters were played about 33% more than level 40 characters.

The gradual increase in play time over the levels might be due to several factors:
1) casual players become discouraged to continue leveling and stop playing.
2) some players play several low to mid-level characters at the same time, although it could be that high level characters are more likely to have alts than mid-level characters.
3) increasing commitment as characters grow in level.
4) a spreading effect - characters at the high end are those in habit of putting in enough time since December to reach the high levels.

The average play-time per character over the period was 10.2 hours. Given that players are known to play multiple characters on their accounts, the average number of hours played per week per player is probably substantially higher. Also, 14% of observed characters were level 1, and 38% of these were never observed to advance beyond level 1.

Posted by nickyee at 02:50 PM | Comments (10) | TrackBack

June 19, 2005

Zone Count

Server Sample: RP (High), PvE (Medium), PvE (High)
Sampling Period: 6/2/2005 6:00 pm - 6/7/2005 3:00 am
Sampling Resolution: ~16 minutes
Parsing Method: The sample unit is each unique character. Each character was tracked across the server logs. Total playing time, lowest observed level, highest observed level, guild affiliation, and zones seen in were parsed.
Data Filter: None
Sample Size: 62,962 characters

Every log sample contains the zone that the character was observed in. The parsing script counted the number of different zones that each character was observed to be in during the sampling period (4 days and 9 hours).

Zone count increased gradually over level with a noticeable spike at level 39. This matches the spike in playing time. Also of interest is the dip at level 60. Our guesses are that most level 60 characters spend more time consistently in the same instance dungeons.

Because playing time correlates highly with zone count (r = .80), we present the same plot of level against average zone count but with effect of playing time controlled for. We see the same trends pre-40 and post-60 that we observed before. In fact, both become a little more clear.

Addendum:

After reading the comments, we redid the analysis and only counted a zone if a character was observed in it in for two contiguous snapshots. This effectively filters out zones counted before that were being traveled through. We see the same pattern as before - a a spike pre-40 and a drop at lvl 60.

The graph below has playing time controlled for.

Posted by nickyee at 11:56 AM | Comments (5) | TrackBack

June 18, 2005

Does guild size vary across server types?

From the parsed logs, we calculated the size of every observed guild. The size of a guild was defined as the number of different characters observed to have that guild tag over the sampling period (4 days and 9 hours). Thus, we were able to know the guild size of every character in the parsed data.

We ran an ANCOVA with Server as the independent variable, Guild Size as the dependent variable, and Lowest Observed Level as the covariate. The effect of Server was significant (p < .001, effect size = .05) and the Guild Size of each server was significantly different from each other (p's < .001).

See graph below for Guild Size by Server across Levels. Both PvE servers have larger guilds than the RP server. The high population PvE server has larger guilds than the medium population PvE server.

It is not clear however whether we're looking at a genuine difference between RP and PvE servers, or that guild size fluctuates among servers in general. Also, it could both be the case that server size affects guild size or that larger guilds prefer to migrate to certain serevr types. Hopefully, as we get more data from other servers, this will become more clear.

Posted by nickyee at 02:58 PM | Comments (1) | TrackBack

June 17, 2005

Grouping and Soloing

After getting the grouping ratios, we were also interested to find out the distribution of grouping ratios at different levels. To give a sense of the proportion of players who were never observed in a group, here's a histogram of grouping ratio for players who were observed to play at least 5 hours in the sampling period (N = 30,278).

Here is the same histogram but only for players under level 30 (N = 13,009). 16% of characters were never observed to be in a group.

And here is the same histogram but only for players above level 55 (N = 8,873). 3% of characters were never observed to be in a group.

Posted by nickyee at 03:04 PM | Comments (4) | TrackBack

June 16, 2005

Grouping Ratio by Class and Level

Grouping ratio was calculated for each character by counting the number of the times the character was observed to be grouped divided by the total number of times the character was observed.

The chart below confirms our intuitions about the soloability of the classes in WoW. Of course, players who choose to be warlocks are probably different from players who choose to be priests so the chart below conflates play style with character class.

The graph below plots grouping ratio against character level. We again see the spike right before level 40. There is also a strong increase in grouping starting after level 56. Starting at level 59, half of play time is spent in a group.

Addendum:

Here is the group ratio plots over level for the Warlock and Priest class (as per David's comment).

Posted by nickyee at 09:00 AM | Comments (11) | TrackBack

June 13, 2005

How does being in a guild affect playing time?

An ANCOVA with Guild Involvement (Yes/No) as the independent variable, Playing Time as the dependent variable and Lowest Observed Level as the covariate did not reveal a significant effect (p = .32). In other words, controlling for level, being in a guild does not increase playing time. Of course, we also have to keep in mind that these are probably two different kinds of players to begin with so this analysis is not a pure comparison of guilded and non-guilded players.

Plotting playing time against level by guild involvement shows several interesting trends. Playing time increases dramatically for guilded players right before level 40. Because of the new skills and mount granted at level 40, it makes sense that players may increase their playing time to achieve those goals. Perhaps being in a guild facilitates this because other guild members encourage and help players reach level 40 when they get close to it, thereby increasing playing time.

Also, playing time is more stable after level 40 for guilded players and fluctuates more for players who are not guilded. Our thoughts are that being close to new skills (at odd levels) increases playing time while being far away from new skills (at even levels) decreases playing time. More importantly, this effect would be stronger for non-guilded players who have more control over their playing time and weaker for guilded players who are also playing for social obligations. Interestingly, this trend is more true before level 50 than after level 50. After level 50, non-guilded players play more at even levels and less at odd levels.

Posted by nickyee at 02:50 PM | Comments (7) | TrackBack

Guild Involvement

Overall, 55% of characters are in a guild. If we exclude all level 1 characters, the overall average is 62%. As the graph below shows, about 90% of characters are in a guild after around level 43.

Posted by nickyee at 02:45 PM | Comments (0) | TrackBack