PlayOn: July 2005 Archives

« June 2005 | Main | August 2005 »

July 28, 2005

PvP Server Difference (Part 1)

Server Sample: RP (High), PvE (Medium), PvE (High), PvP (High)
Sampling Period: 6/30/2005 2:00 am - 7/07/2005 2:00 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: 106,762 characters

We now have data from a PvP server so we wanted to list some of the PvP vs. non-PvP server findings. Overall, the differences were surprisingly small. The differences may be driven by the idiosyncrasies of the PvP server we chose. We are gathering data on a second PvP server so we'll know whether that is true soon.

Playing Time

Characters on the PvP server played about an hour more (~ 70 mins) over the one week period than characters from the RP and the two PvE servers. The average character play-time in this sample was 11.2 hours, so that's about a 10% difference of the mean.

Guild Involvement

While we might have expected that characters on PvP servers may be more inclined to be in a guild (for safety in numbers, etc.), the data didn't bear this out. Overall, guild involvement rates were comparable across server types. Characters on the RP server were slightly less likely to be involved in a guild at the lower levels.

Guild Size

We might have also expected that guilds in PvP servers would be larger in general given the demands on survival. This also didn't bear out with the data. We created a list of every guild that was observed and the total number of unique characters observed to have that guild tag. There were significant differences, but it was between the PvE (high) server and the RP (high) server.

Posted by nickyee at 04:07 PM | Comments (1) | TrackBack

July 25, 2005

Playtime and Class / Race

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

We compared the weekly play-times of the different classes. It was surprising to find that Rogues play about 2 hours more each week than most other classes.

We also ran the analysis split by race. Gnomes and Dwarves spend the least time each week playing - about 1-2 hours less than other races.

To get a better sense of these play-time differences, we include below the average play-times of all race and class combinations with level controlled for.

Levels 1-20 (average play-time in a week):

Levels 41-60 (average play-time in a week):

Overall, we found that the main variation in playing time was driven by class rather than race. For example, rogues play 5.5 hours (28%) more than Hunters.

Count of race/class combinations:

It seems like we're picking up on some farming activity ...

Posted by nickyee at 10:00 AM | Comments (5) | TrackBack

July 22, 2005

Zone Type

We calculated the ratio of time characters spent in instances, outdoor zones, and city zones. The following graph plots the time ratio spent in these 3 zone types. City ratio is stable across levels. Outdoor zone ratio drips post 40 and again post 55. Instance ratio rises post 55.

Posted by nickyee at 11:23 AM | Comments (1) | TrackBack

July 20, 2005

Leveling Time By Server

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 leveling event. We tabulate the time between a character's level and when we observe them at a new level. We exclude the first leveling event from every character because it doesn't constitute the total amount of time to make that level.
Data Filter: None
Sample Size: 81,887 leveling events

Is leveling time comparable across servers? We ran an ANCOVA with Server as the independent variable, Leveling Time as the dependent variable, and Level as the Covariate. The effect of Server was significant (p = .001). The PvE (Medium) server had lower leveling time than the PvE (High) and the RP (High) servers.

Between levels 1-20, the difference is about 15 minutes per level (13% faster). Between levels 21-40, it's about 87 minutes per level (19% faster). And between levels 41-60, it's 192 minutes per level (about 21% faster).

We're not sure whether we're looking at general variation across servers or a function of server size (medium vs. high population servers). The server size hypothesis seems to make sense. It may be easier to complete quests and gain access to mobs on a less populated server. Hopefully, future data will help clarify these findings.

Posted by nickyee at 10:28 AM | Comments (0) | TrackBack

July 17, 2005

Guild Size

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 guild. Each guild was tracked and every character with that guild tag was counted in the tabulation over the one week period.
Data Filter: None
Sample Size: 2,728 guilds

17.5% of characters were observed to be in a guild that no one else was observed to be in during the one week period - i.e., a one-person guild. The average guild size was 14.5 (16.8 if one-person guilds are excluded). The median was 6 (9 if one-person guilds are excluded). The largest observed guild had 257 members. The lower quartile is 2. The upper quartile is 16. The 90th percentile was 35. The 95th percentile was 54. The 99th percentile was 110.

The graphs below are presented as a complement to Christopher Allen's interesting article on Dunbar's number.

In this graph below, each guild is represented once and sorted in ascending order.

In this graph, the frequency of each guild size is shown.

Posted by nickyee at 02:22 PM | Comments (5) | TrackBack

Guild Affiliation and Leveling Time

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 leveling event. We tabulate the time between a character’s level and when we observe them at a new level. Only a character's online time is included. We exclude the first leveling event from every character because it doesn't constitute the total amount of time to make that level.
Data Filter: None
Sample Size: 81,887 leveling events

Do characters in guilds level faster or slower than characters not in guilds? Again, we won't be able to pinpoint the causality or take into account other variables, but it would be interesting to find out just how much difference there is in leveling times between characters in and not in a guild. Running an ANCOVA with Guilded (yes / no) as the independent variable and Leveling Time as the dependent variable, we found a significant, but small, effect where characters in a guild have a longer leveling time than characters not in a guild.

Between levels 1 and 20, the difference was 15 minutes (105 vs. 115 minutes). Between levels 21 to 40, the difference was 24 minutes (417 vs. 441 minutes). And then between levels 41 to 60, the difference was 72 minutes (802 vs. 874 minutes).

Posted by nickyee at 02:20 PM | Comments (2) | TrackBack

July 13, 2005

Grouping and Leveling Time

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 leveling event. We tabulate the time between a character's level and when we observe them at a new level. Only a character's online time is included. We exclude the first leveling event from every character because it doesn't constitute the total amount of time to make that level.
Data Filter: None
Sample Size: 81,887 leveling events

How does grouping ratio affect leveling time? Does grouping more help or hurt leveling time? Of course, because we are looking at correlational data, we won't be able to pinpoint causality or directionality (or a third variable that isn't accounted for), but the data might lend support to either the "help" or "hurt" hypothesis and help us further explore the impact of grouping behavior on game-play.

We ran an ANOVA with Group Ratio and Level as the independent variables, and Leveling Time as the dependent variable. The effect of Group Ratio was significant (p < .001). The graph below shows that characters that never group level faster than those that do. Among characters that do group however, there is no difference in leveling time if they group a little or a lot.

To get a better sense of the faster leveling time of characters who never group, we calculated the average leveling time for characters who group more than 1% and compared that average with the leveling time of characters who never group. Below we plot the leveling time advantage of characters who never group over the levels.

Between levels 2-10, characters who never group make each level on average 23 minutes faster. Between levels 11-20, it's 40 mins per level. Between levels 21-30, it's 74 mins per level. Between levels 31-40, it's 130 mins per level. Between levels 41-50, it's 215 mins per level. And between levels 51-60, it's 293 mins per level.

We are probably looking at several variables interacting. Players who group more often probably are motivated to play for more social reasons than players who never group. At the higher levels, a lot of time may be spent helping fellow guild members. Another issue is that players who never group are probably playing more soloable classes. In other words, we're not making a pure comparison of grouping behavior.

Posted by nickyee at 09:32 AM | Comments (6) | TrackBack

July 10, 2005

Leveling Time By Class

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 leveling event. We tabulate the time between a character's level and when we observe them at a new level. Only a player's online time is counted. We exclude the first leveling event from every character because it doesn't constitute the total amount of time to make that level.
Data Filter: None
Sample Size: 81,887 leveling events

Below, we plot the leveling time by class split for 3 level ranges: 1-20, 21-40, and 41-60. In all three plots, the starting level is controlled for. Controlling for group ratio produces similar plots. Druids are consistently the slowest levelers. Priests, mages, and shamans are consistently the fastest levelers. Between levels 1-20, mages level 10% faster than druids. Between levels 21-40, mages level 14% faster than druids. Between levels 41-60, mages level 20% faster than druids. The data shows that between level 41 and 60, mages make each level about 2-3 hours faster than druids. Even when controlled for group ratio, the trend is the same.

In fact, we know that this effect is not being driven by group ratio because Druids group almost just as much as Priests. At first we thought this may be a difference in DPS and downtime (which could make sense for shamans and mages compared with druids), but the priests throw us off. So it may be a combination of efficiency (DPS) and demand in groups that yield high XP.

Levels 1-20:

Levels 21-40:

Levels 41-60:

Posted by nickyee at 04:17 PM | Comments (57) | TrackBack

July 08, 2005

Leveling Time

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 leveling event. We tabulate the time between a character's level and when we observe them at a new level. Only a player's online time is counted. We exclude the first leveling event from every character because it doesn't constitute the total amount of time to make that level.
Data Filter: None
Sample Size: 81,887 leveling events

We calculated the amount of time for every character to make a level within the sampling period of one week. The sampling unit was a leveling event. A leveling event was defined as the time it took for a character to be observed at a new level from their previous level. The first leveling event of every character is excluded because it doesn't constitute the total time it took to make that level (since presumably we started sampling when most characters were in the middle of their current level).

Below we plot the average amount of time it took for characters to make the next level from a given level. The curve is roughly linear with a mild dip at level 39 and a large spike at level 40. In other words, it takes people slightly less time to get to level 40 (playing harder) but then much longer to get to level 41. The curve also becomes steeper after level 53. Also of interest is the mild "step" effect throughout - leveling time is shorter at odd levels (close to new skills) and longer at even levels. All of these findings make sense given the new skills granted at even levels and the mount becoming available at level 40.

The time it takes to make the next level is roughly linear. The equation is approximately:

Leveling Time (in mins) = (Current Level x 17.8) - 49

If we assume that current level 60s spent these amounts of time while reaching level 60, then the average level 60 character has an accumulated play time of 20.3 days - a total of 61 8-hour work days, or roughly two full months of work days. 10% of all characters in WoW are currently level 60 and have spent this amount of time in the game.

The 10 day accumulated play time occurs around level 43. 20% of characters in WoW are at least level 43 and have spent 30 8-hour work days, or roughly one full month of work days in the game. Please note that we are counting characters and not players. Actual play-time will be higher for players who have multiple characters. Also, we only have non-PvP servers in this sample. Leveling times on PvP servers may be significantly higher. We'll try to get this data soon.

Here is the plot of the average accumulated play time by level.

The integral of the "time to next level" equation would give us the accumulated time on average needed to get to a certain level:

Accumulated Leveling Time (in mins) = 8.9 x Current Level (Current Level - 5.5)

Below are the tables for:
1) The average time it takes for a character to reach the next level from a given level
2) The average accumulated play-time of a character at a given level

Posted by nickyee at 10:49 AM | Comments (40) | TrackBack

July 06, 2005

Economics and Sociability

I like playing the economic game. Thinking about how I played both SWG and WoW, it's interesting to see the ways that the implementation of economic transactions can have a large impact on the formation of long-term social relationships.

Commoditization Crafted items in WoW are commodities, in the sense that a Mooncloth Robe made by one tailor is essentially identical to a Mooncloth Robe made by anyone else. There is no customization available. As a result, price becomes the only differentiator in products that are widely crafted, and availability the differentiator for those that are more rare. In SWG, the situation was different. (I haven't played since December, 2004.) Raw materials themselves had varying qualities (along multiple dimensions) which were reflected in crafted objects and subcomponents. Crafters had to put quite a bit of effort into collecting good quality materials, and more effort into learning to combine them to produce marketable goods. In my case, I produced a type of armor favored by RPers and the fashion-sensitive. Each suit was custom-made, both in terms of color, and in terms of the additives the customer was willing to pay for. It usually took several days to negotiate an order, and a few more to fill it. Of course, I kept track of past customers. I had a clientele.

The fact that resources were not commodities meant that I had a competitive/cooperative relationship with a few other armorsmiths. Armorsmiths tended to fill niches, and I could cooperate more freely with smiths in different niches, but even with direct competitors, I would occasionally trade favors back and forth, knowing that we each had stockpiles the other could use from time to time.

In WoW, I don't keep track of clients. First off, they probably bought from me because I had the lowest price. But secondly, I don't even know who they are...

Market Visibility SWG and WoW both provide information about the seller to the potential buyer. When there is a sale, however, the two games differ. SWG sends email to the seller which contains the money, but also tells the name of the buyer. In fact, the in-game email appears to come from the buyer. In WoW, the money is attached to an in-game email from the auction house, and the name of the buyer is unknown.

Several sellers in SWG, including me, had a habit of sending a post-sale "thank you" note to the buyer. Several other practices were possible. Although SWG limited the value that could be sold through the main bazaar system, some craftsmen would sell certain items on the bazaar to advertise their vendor location, or to collect names of players buying certain types of wares. (Since crafted items could bear a crafter-supplied name, some tailors would place items on the bazaar whose name contain the address of their store.)

If we assume that crafted items in these games are likely to be purchased by a larger population than those who crafted them, then it stands to reason that sellers may have more of a vested interest in identifying and "locking in" high-volume or high-profit buyers. Thus, sellers may be more interested in collecting names of buyers than vice versa. But in WoW, it is only the buyers who can initiate contact. As such, the only contacts I'm aware of between a buyer and a seller otherwise unknown to each other involved a long-term contract for enchanting reagents and contracts for morrowgrain, needed in very large supply by people doing a rare quest.

All in all, I reach the conclusion that even minor facets of the crafting and economic can have a large impact in the social relationships formed among crafters, between crafters and consumers, and between buyers and sellers.

Posted by at 11:20 AM | Comments (9) | TrackBack

July 01, 2005

Guild Affiliation and Group Ratio

Even though characters in guilds do not play more than characters not in guilds, they do group more often and this effect becomes more pronounced over the levels. We wanted to check this in case being in a guild did not in fact provide a grouping advantage given the ease of finding an ad-hoc group. The following graph plots average percentage of time observed in a group with playing time controlled for.

In an ANCOVA with Guilded (yes / no) as the independent variable, average group ratio as the dependent variable, and Level (lowest observed level) and Playing Time (minutes observed over sampling period) as the covariates, the effect of Guilded was significant (p < .001) overall and for individual analyses within levels 1-20, 21-40, and 41-60.

Between levels 1 and 20, characters not in a guild are observed in a group on average 17% of the times compared with 20% for characters in a guild (note that these and the following group ratios are after playing time and level have been controlled for). Between levels 21 and 40, it is 26% and 31%. And between levels 41 and 60, it is 34% and 46%. In other words, between level 41-60, characters in a guild are in a group about 35% more often than characters not in a guild after playing time has been controlled for.

Posted by nickyee at 03:51 PM | Comments (0) | TrackBack