> For each topic I counted how many times it had been mentioned in a post title, then calculated a 0-100 index, where 100 is the month with the most mentions.

> For each topic I counted how many times it had been mentioned in a post title, then calculated a 0-100 index, where 100 is the month with the most mentions. This is not really useful data then. A time series with a big spike will "lose" against an identical one without the spike. Maybe the amount of days when a certain word was in the front page would be a better comparison metric on the HN side?