In the context of quantitative literary analysis studies, this research aims to verify the existence of a predictive model of characters' death in literary works. Using George R. R. Martin's book of A Game of Thrones as a model, the weighted degree, eigenvector centrality, betweenness centrality and local clustering coefficient for each node were calculated. After that, a monadic level test of logistic regression took place between these measures and the life and death condition of the characters at the end of the book. The outcome of the test did not lead to rejecting the null hypothesis, since no statistically significant relationship was found, but it did lead to interesting reflections on the work and poetics of George Martin, whose search for realism leads him intentionally to decree the death of his characters regardless of their centrality.