Introduction

This website will take you on a tour, analysing the Major characters in Warcraft . This will be done by accessing the wowpedia page for each of the characters using the wowpedia api , where we will get information about each characters gender, faction, race and whether the character is alive in the lore of the game, while also extracting their in-game quotes from the quotes section on the page. Additionally, we will extract user comments from the wowhead character pages using beautifulsoup , where the timestamp associated with each comment will also be used. The resulting raw and pre-processed json and txt files contains 1739 files (~50MB), which can be downloaded using the button above, including some pickled results etc.. Almost everything on this website is in some way, shape, or form interactive, so please remember to play around. You can use the Navbar in the top to auto scroll to different sections on the page.

In the Warcraft universe (primarily in World of Warcraft), the plot revolves around two competing factions, the Horde and the Alliance, which are in constant conflict with each other. From time to time, outside participants disturb this usual back and forth, e.g. when Arthas defected from the Alliance and became the Lich King, which both factions tried to see an end to (repeated again when Sylvanas Windrunner leaves the Horde and joins the Jailer in Shadowlands by shattering the Lich Kings mask). Some of this strife can be captured by creating a network where the characters have assigned a faction attribute, where neutral characters are mainly bosses in the game, or characters that aid both the Horde and Alliance in their fight against a common enemy.

We will analyse the Major characters by grouping them by different attributes such that we can compare the characteristics of these groups, e.g. comparing how verbal the Alliance are compared to the Horde and the sentiment of the different groups. In order to do this, we begin by creating a network where nodes are each of the major characters, and edges between them arise when one characters wowpedia page links to anothers wowpedia page. This resulted in a network consisting of 261 characters and 4009 edges connecting them, as seen on the Warcraft network visualization below.

Network Visualization

The nodes are colored according to their faction (red for Horde, blue for Alliance and gray for others, such as bosses in the game). Additionally, edges are colored according to the relationships of the two nodes, where gray links from any node to a neutral node, blue is between two Alliance nodes, red is between two Horde nodes and green is between a Horde and an Alliance node. (Interactive plot, reset button in the top right corner of the plot in case you get lost)

We can see from the network that there are some characters that are very well connected, taking a closer look we can see that the most connected Alliance characters include the current and late kings of the Alliance capital "Stormwind" (Anduin and Varian Wrynn). For the Horde characters, we notice the biggest nodes in the network, Thrall, which has been the warchief of the Horde capital "Orgrimmar" and is probably one of the most known character in Warcraft. Finally, for the Neutral characters we see Sylvanas Windrunner as one of the most connected, she is one of the newest boss characters who previously was a part of the Horde.

To get a quick overview of the different attributes in the network an attribute distribution is created for each of the four attributes in the graph. From this attribute plot we see that genders, faction and status is nicely divided into a few groupings, however the race distribution is populated with a lot of different races with a few occurrences apart from the main character races such as human, orc and blood elf.

Attribute Distribution

Gender Faction Status Race

Character Connections

In order investigate the characteristics of connections between the major characters of Warcraft, you can use degree distribution plots, along with stats for in-, out- and total-degree, and different centrality measures.

Below plots of the out- and in-degree distributions of the Warcraft network is shown along with the degree distribution of a random Erdős-Renyi network, which have the same number of nodes as the Warcraft network and the probability of connection is calculated from the average out- and in-degree respectively. From this it seems like the Warcraft network is not entirely random! Additionally, a scatter plot of the In- vs. out-degree for each character in the network is shown.

Out-degree In-degree Degree Correlation

In order to differentiate between the connectedness within the different communities of Alliance vs. Horde vs. Neutral or Male vs. Female characters, we created subgraphs for each of these and calculated some basic degree stats, that is min, max, mean, median and mode for the in-, out- and total-degree of these subgraphs as displayed in the table below. One thing to note from this is that e.g. the Horde characters are more disconnected from each other than the Alliance and Neutral characters (since the mean and median is lower for the Horde characters). Looking at the male vs. female characters, we see that the female characters seem to talk much less to each other than the male characters do! This could have something to do with the fact that there are far fewer female than male characters in the network (see attribute distribution plot).

We can dive even further into comparing the connectedness of characters, by inspecting the different centrality measures over different subgraph partitions of the network, as displayed in the table below. Comparing the centrality measures for the different factions, it is clear that the Horde characters and the Alliance characters have a much higher centrality measure in their subnetworks than the characters without faction do, which indicates that characters within the Horde and Alliance factions have more tight-knit communities than the characters without a faction.

In addition to partitioning by node attributes, it is possible to use graph partitioning algorithms. One of these algorithms is the Louvain algorithm, which when used on the Warcraft network splits the network into seven communities, where the top three characters of each community according to their node degree is:

From the top characters in the community partitions, it seems like the network is partitioned reasonably by the Louvain algorithm. Community 1 has a lot of connection through the lore that unfolded in the Legion expansion. Community 2 consists of end-game bosses. Sylvanas, Lich King and Varian of community 3 have all been faction leaders. Each character in community 4 has some kind of nature aspect. Community 5 consists of important orcs. Community 6 contains the race "Arakkoa". However, community 7 seems a bit random with respect to the lore and their races, factions etc. Below is an image of each of the most connected character overall for each community