3 ms·

How does your model compare to using: (Users who upvoted a given submission / Users who saw a page that includes the submission and its vote icon) This would

How does your model compare to using: (Users who upvoted a given submission / Users who saw a page that includes the submission and its vote icon) This would be a percent between 0 (no one who saw a page containing a given submission upvoted it) and 100 (everyone who saw it upvoted it). Receiving more impressions wouldn’t change that percentage. Weaknesses: It can only be calculated by HN itself. On pages that list lots of submissions (like the home page), it need may to compensate for relative position on the page. These pages may already randomize position enough for this not to be an issue, or to only be an issue for the first 3-5 items on the home page.

jwarden 4y ago
Interesting question. One difference is that upvoteRate formula adjusts for where the submission appears on the page (the rank). It also adjusts for how many site-wide upvotes occurred during that time period. You are right, since we don't know the number of users who saw the page + the vote icon, so we can't calculate the probability Pr(upvote|saw submission with upvote button). But the upvoteRate formula would be proportional to this probability, times additional factors for rank and time. We talked about this in our original blog article here: https://felx.me/2021/08/29/improving-the-hacker-news-ranking-algorithm.html https://felx.me/2021/08/29/improving-the-hacker-news-ranking...