This is a Leopards team’s submission for the Propaganda Detection datathon. Key findings: the best performing classifier is logistic regression, operating on Word2vec representations of the sentences plus several designed features like the proportion of sentiment-bearing words in the sentence.