I think it's funny because last night I built models to predict: (1) the probability that a headline gets more than 5 comments on HN, and (2) the probability that a headline gets more than 16 votes. Both of these have a similar probability of happening.<p>The model is a very simple logistic regression/bag of words model and it gets what I would usually call an awful ROC-AUC of 60%, off-the-cuff I'd say 1 article in 500 has a predicted probability of 30% of hitting it big.<p>Anyway three of the keywords that most strongly predict than article <i>will not</i> hit it big on HN are "Biden", "Elon" and "Musk". (But it says that "Richard Stallman has died" has a 70% or so chance of success... and those of you who are concerned about politically bias should know that my crawler has sucked down 2020-mid 2022 data so far so there isn't as much "Trump" in there as "Biden".)<p>I'm not so sure what my model thinks you think about sex, I'm going to have to go home and ask it.