Do you do some deduplication. Some classics are resubmitted a few times successfully. Perhaps add a third sort criteria, that is sorted by the number of big reposts.
Do you do some deduplication. Some classics are resubmitted a few times successfully. Perhaps add a third sort criteria, that is sorted by the number of big reposts.
I created this for my own use but sharing it here because others may find it useful (hope this isn't breaking any HN etiquette and if you're the original author and want me to take it down please let me know).
[1] https://docs.google.com/spreadsheets/d/16MNPM9fhpglC1s1OV-tp...
[2] https://gist.github.com/san-kumar/0b7d218a08d4d8b4a7d9c07db5...
For example say that several years ago a post might hit front page and 500 people would vote on it. If 300 people voted up and 200 down then it gets 100 points.
Fast forward and now there are more people on the site. That means greater confidence in the percentage but because only points and not percentage of up/down is revealed, a post with similar percentage up- vs downvotes appears to be better. 600 up and 400 down is still 60% up vs 40% down but it results in 200 points total.
Like I said though maybe HN does account for this and 100 points today means the same percentage of people upvoted it, I don’t know.
https://news.ycombinator.com/item?id=243417
As a result, its score did not increase, although upvoting "worked".
for (var i in data[yr]) {
var story = data[yr][i];
if (seen_titles[story.title]) continue;
seen_titles[story.title] = 1;
// ...add story to list of stories
};