@thomasnield well bayes is about inverse probabilities, so basically the likelihood of something being spam is the prior times the product of the likelihood ratios of each feature