Hacker News new | past | comments | ask | show | jobs | submit login

You're right, the problem is especially acute for story submission.

Once an article is on the frontpage, it is likely to stay there. People are much more likely to upvote a frontpage article than something in new, simply because they are far more likely to look at the frontpage.

What you really want is to estimate the probably that a reader will click and/or upvote a link, not the total number of clicks/upvotes. You might also want to discount this by time, but this discounting might not be necessary---there might be a natural decay in the probability, particularly if you compute a moving average.

If you'd like my feedback about how to model this, feel free to email me. I'm happy to provide you with a more concrete methodology.




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: