4 ms·

I love this. The search box had an option to autocomplete terms I had recently searched, I clicked on one (a term I was interested in, obviously) and immediate

I love this. The search box had an option to autocomplete terms I had recently searched, I clicked on one (a term I was interested in, obviously) and immediately found an interesting HackerNews post. Nice! I think HackerNews and other link aggregators (e.g. reddit) have a kind of recency problem, where there is a lot of great content, but people only see the recent stuff. This seems like a great way to uncover some of the latent value of old HackerNews content. Is there a way to suggest content too? e.g. If you liked that, you'd probably also like X, Y, and Z.

MattGaiser 6y ago
>I think HackerNews and other link aggregators (e.g. reddit) have a kind of recency problem, where there is a lot of great content, but people only see the recent stuff. With Reddit and HackerNews, I want a relative ranking index. I can search by the top content, but something 5th today could have more votes than the top submission of 2015 because of forum/subreddit growth. I want them ranked by something like the ratio of views to upvotes or upvotes compared to total upvotes for that day.
- rhaksw 6y ago
  > I want them ranked by something like the ratio of views to upvotes or upvotes compared to total upvotes for that day. Yeah, this is definitely possible with public data. I did something similar on reveddit [1] for removed reddit content. Hovering over the graph shows the item with the highest vote ratio [3], and clicking skips to that point in time. Code is here [2] for anyone interested and I apologize in advance.. [1] https://i.imgur.com/p3Bi5IS.png https://i.imgur.com/p3Bi5IS.png [2] https://github.com/reveddit/ragger/blob/master/revddit_aggregator.py https://github.com/reveddit/ragger/blob/master/revddit_aggre... [3] https://www.reveddit.com/r/worldnews/?showFilters=true https://www.reveddit.com/r/worldnews/?showFilters=true
- nwienert 6y ago
  I want a map-reduce over their content that turns them into a canonical “master guide” of their topics.
- runnerup 6y ago
  Absolutely. HN user 'EvanMiller had a lot to say about this, 11 years ago. His tl;dr is that the ranking score should be "the Lower bound of Wilson score confidence interval for a Bernoulli parameter." I believe HN's ranking system is extremely creative and works great for the main page and day-to-day use (my understanding is that it additionally makes use of time-decay terms for comments and stories). Like you said, it's really just historical search (i.e. algolia) that seems broken. https://www.evanmiller.org/how-not-to-sort-by-average-rating.html https://www.evanmiller.org/how-not-to-sort-by-average-rating... https://news.ycombinator.com/item?id=15709405 https://news.ycombinator.com/item?id=15709405 https://www.evanmiller.org/bayesian-average-ratings.html https://www.evanmiller.org/bayesian-average-ratings.html https://www.evanmiller.org/ranking-items-with-star-ratings.html https://www.evanmiller.org/ranking-items-with-star-ratings.h... https://redditblog.com/2009/10/15/reddits-new-comment-sorting-system/ https://redditblog.com/2009/10/15/reddits-new-comment-sortin...
- manx 6y ago
  Agreed! The problem I see is that the score does not correlate well with the perceived quality of the community. I'm researching that for some time now and am in the process of preparing a blog article with data analysis and solution approaches. Work in progress: https://github.com/fdietze/downvote-scoring https://github.com/fdietze/downvote-scoring https://felix.unote.io/hacker-news-scores https://felix.unote.io/hacker-news-scores

MattGaiser 6y ago

>I think HackerNews and other link aggregators (e.g. reddit) have a kind of recency problem, where there is a lot of great content, but people only see the recent stuff. With Reddit and HackerNews, I want a relative ranking index. I can search by the top content, but something 5th today could have more votes than the top submission of 2015 because of forum/subreddit growth. I want them ranked by something like the ratio of views to upvotes or upvotes compared to total upvotes for that day.

rhaksw 6y ago
> I want them ranked by something like the ratio of views to upvotes or upvotes compared to total upvotes for that day. Yeah, this is definitely possible with public data. I did something similar on reveddit [1] for removed reddit content. Hovering over the graph shows the item with the highest vote ratio [3], and clicking skips to that point in time. Code is here [2] for anyone interested and I apologize in advance.. [1] https://i.imgur.com/p3Bi5IS.png https://i.imgur.com/p3Bi5IS.png [2] https://github.com/reveddit/ragger/blob/master/revddit_aggregator.py https://github.com/reveddit/ragger/blob/master/revddit_aggre... [3] https://www.reveddit.com/r/worldnews/?showFilters=true https://www.reveddit.com/r/worldnews/?showFilters=true
nwienert 6y ago
I want a map-reduce over their content that turns them into a canonical “master guide” of their topics.
runnerup 6y ago
Absolutely. HN user 'EvanMiller had a lot to say about this, 11 years ago. His tl;dr is that the ranking score should be "the Lower bound of Wilson score confidence interval for a Bernoulli parameter." I believe HN's ranking system is extremely creative and works great for the main page and day-to-day use (my understanding is that it additionally makes use of time-decay terms for comments and stories). Like you said, it's really just historical search (i.e. algolia) that seems broken. https://www.evanmiller.org/how-not-to-sort-by-average-rating.html https://www.evanmiller.org/how-not-to-sort-by-average-rating... https://news.ycombinator.com/item?id=15709405 https://news.ycombinator.com/item?id=15709405 https://www.evanmiller.org/bayesian-average-ratings.html https://www.evanmiller.org/bayesian-average-ratings.html https://www.evanmiller.org/ranking-items-with-star-ratings.html https://www.evanmiller.org/ranking-items-with-star-ratings.h... https://redditblog.com/2009/10/15/reddits-new-comment-sorting-system/ https://redditblog.com/2009/10/15/reddits-new-comment-sortin...
manx 6y ago
Agreed! The problem I see is that the score does not correlate well with the perceived quality of the community. I'm researching that for some time now and am in the process of preparing a blog article with data analysis and solution approaches. Work in progress: https://github.com/fdietze/downvote-scoring https://github.com/fdietze/downvote-scoring https://felix.unote.io/hacker-news-scores https://felix.unote.io/hacker-news-scores