The definition of what constitutes a conflict, in terms of both time and events, has been consistently debated in the historical study of social phenomena. What has become increasingly clear, however, is that the methods used to define it are generally biased as they are highly dependent on the perceptions of the researchers who establish them. Using Mexico as a case study, we tested a data-driven method for periodizing long-term homicide trends. We found structural break analysis and Markov-switching regression modelling to be useful methods for periodizing long-term time series data of homicide rates, without an "a priori" or skewed definition. We also found that the current trend of increasing homicidal violence may prove difficult to reverse given that historical structural shifts in Mexican homicidal violence are unlikely events in the long term.