Being a social game Survivor lends itself well to social network analytics. Over the season I'll be tracking the alliances of the castaways as they evolve each episode. Typically we see strong alliances start from the beginning, weaker alliances get picked off and then there are power plays. Visualising the network as a graph will show the strength of the alliances, those on the chopping block and potential swing votes. I'll also use the graph to keep track of who has hidden immunity idols and other advantages.
Come the merge things get more interesting. The game opens up and big plays are made. At this stage of the game the graph can be used to make predictions of the most likely to be named sole survivor, those most likely to make it to the final three, those that will make it to the jury and who are the big players staying under the radar ready to make their move. The graph will also help to measure who played a strong game based on the data, including who beat the odds. This is just the beginning and best will be coming in the later weeks.
Alliance network graph
The graph has the following features:
- Each node is a castaway
- Each edge represents those two castaways voted together (not for the other person)
- The closer the node, the stronger the alliance
- The width of the edge, the stronger the alliance
- The colour will remain as their original tribe colour (for simplicity)
- In the case of a tie, only the final deciding votes will be used
Who votes with who isn't the ultimate indication of the alliance or strength of the alliance. There's a lot more to Survivor than just the votes. Castaways can vote with each other to vote out a common enemy or split the vote to flush an idol. However, in general closer nodes will indicate stronger bonds, for example the strongest 2 in an alliance of 5.
Machine learning algorithms mathematically determine alliances by estimating communities or clusters. This representation gives us a slightly different picture of the true story but establishes a good base important for advanced prediction. It will be interesting to see how closely they fit to the true alliances as the season evolves.
Current state of play
In episode 1 we unfortunately saw Pat leave the island due to injury, which isn't a way anyone wants to go. In episode 2 we saw 2 alliances emerge, one voting for Lyrsa and one voting for Jessica with the majority voting for Jessica. It was a blindside that took Carl, Davie and Bi by surprise.
In episode 3 it was time for the Goliath tribe to face their first tribal council. There was confrontation between Jeremy and Natalie. Jeremy was rightfully concerned that his tribe mates were plotting against him. He attempted to convince them to stop plotting and scheming and remain unified. In a sense he got what he wanted, a unified tribe. Unfortunately for him the tribe was unified against him and received all the votes.
After 3 episodes here's how the alliances have evolved (hover, click and drag nodes to see some stats! If you can’t see the graph try Chrome or unblocking the content.).
The same graph again with the eliminated players removed.
Understandably the alliances are completely disjoint at this point since there has not yet been a tribe swap and each tribe has only visited tribal council once. Over the coming episodes we may see things get shaken up. According to the Survivor Wiki a tribe swap will happen next episode which should be fun!Follow me on social media: