5 ms·

Then what is the algorithm for determining where something should be ranked. I see submissions with 400 votes in two hours being ranked far below ones with 100

by kolbe 4y ago

Then what is the algorithm for determining where something should be ranked. I see submissions with 400 votes in two hours being ranked far below ones with 100 votes in 10 hours, and often the variable seems to be the article is critical of ChatGPT or a YC portfolio company. What else goes into the rankings?

dang 4y ago
"How are stories ranked" is near the top of https://news.ycombinator.com/newsfaq.html https://news.ycombinator.com/newsfaq.html. Does that answer your question? In terms of moderator action: we might downweight ChatGPT topics (for oar against) if they seem repetitive rather than significant new information (https://hn.algolia.com/?dateRange=all&page=0&prefix=false&sort=byDate&type=comment&query=%22significant%20new%20information%22%20by%3Adang https://hn.algolia.com/?dateRange=all&page=0&prefix=false&so...). But we don't downweight posts that are critical of YC companies—or rather, we do so less than we would downweight similar threads on other topics. See https://hn.algolia.com/?dateRange=all&page=0&prefix=false&query=by%3Adang%20moderate%20%22less%2C%20not%20more%22&sort=byDate&type=comment https://hn.algolia.com/?dateRange=all&page=0&prefix=false&qu....
- kolbe 4y ago
  That link doesn't really tell us. We're a very analytical bunch here, and when I see one story that should have a 10x higher rank according to the "basic algorithm" being ranked lower, that means these "other factors" are much more than a slight twiddling. And when you don't provide the full algorithm, just a hand waive, it makes it difficult to ascertain what's really happening. Are you sure there aren't abuses from your portfolio companies managers/employees to flag negative stories? I imagine Sam, for example, knows exactly what he has to do to get ChatGPT criticism guided off the stage. Edit: for example, do you know what happened with this story? https://news.ycombinator.com/item?id=35245626 https://news.ycombinator.com/item?id=35245626 This is a very interesting/important topic. This was a new topic. It was really hot in the first hour, and just got smashed off the front page.
  - dang 4y ago
    > Are you sure there aren't abuses from your portfolio companies managers/employees to flag negative stories? I imagine Sam, for example, knows exactly what he has to do to get ChatGPT criticism guided off the stage. Quite sure. That is, there may be managers/employers of $companies trying to flag things, but being a YC portfolio company doesn't make that any easier. And yes I'm sure that Sam can't do that. (I also know that he wouldn't try, but that's a separate point.) Re the FAQ: it doesn't give a detailed explanation (we can't do that without publishing our code) but it summarizes the factors comprehensively. If you want to know more I need to see a specific link. Speaking of which: Re https://news.ycombinator.com/item?id=35245626 https://news.ycombinator.com/item?id=35245626: it was on HN's front page for 4 hours, and at some point was downweighted by a mod. I haven't checked about why, but I think most likely it was just our general approach of downweighting opinion pieces on popular topics. Keep in mind that the LLM tsunami is an insanely popular topic—by far the biggest in years—and if we weren't downweighting follow-ups a la https://hn.algolia.com/?dateRange=all&page=0&prefix=true&query=follow-up%20downweight&sort=byDate&type=comment https://hn.algolia.com/?dateRange=all&page=0&prefix=true&que..., it would saturate the front page every day. Actually we tend to not do that moderation on randomwalker posts (https://news.ycombinator.com/user?id=randomwalker https://news.ycombinator.com/user?id=randomwalker) - because they're basically always excellent. But a certain amount of randomness is inescapable and randomwalker posts do great on HN most lot of the time. If we made the wrong call in this case, so much the worse for us and I'm genuinely sorry.
    - p-e-w 4y ago
      > it doesn't give a detailed explanation (we can't do that without publishing our code) Precisely why would publishing (the relevant part of) the code be a problem? Twitter did it just a few days ago, and they aren't even known as an information hub of the open source world, plus they face a lot more public scrutiny for everything they do, to put it mildly.
      - dang 4y ago
        I'm not sure you've got an apples-to-apples comparison there (between what Twitter published and what I was just talking about). Either way, though, I don't want to publish that part of our code for two reasons: I fear that it would make HN easier to game/mainpulate, and I fear that it would increase the number of objections we have to deal with. It's not that I mind dealing with objections in principle, but a 10x increase would bury me.

by kolbe 4y ago

dang 4y ago

"How are stories ranked" is near the top of https://news.ycombinator.com/newsfaq.html https://news.ycombinator.com/newsfaq.html. Does that answer your question? In terms of moderator action: we might downweight ChatGPT topics (for oar against) if they seem repetitive rather than significant new information (https://hn.algolia.com/?dateRange=all&page=0&prefix=false&sort=byDate&type=comment&query=%22significant%20new%20information%22%20by%3Adang https://hn.algolia.com/?dateRange=all&page=0&prefix=false&so...). But we don't downweight posts that are critical of YC companies—or rather, we do so less than we would downweight similar threads on other topics. See https://hn.algolia.com/?dateRange=all&page=0&prefix=false&query=by%3Adang%20moderate%20%22less%2C%20not%20more%22&sort=byDate&type=comment https://hn.algolia.com/?dateRange=all&page=0&prefix=false&qu....

kolbe 4y ago
That link doesn't really tell us. We're a very analytical bunch here, and when I see one story that should have a 10x higher rank according to the "basic algorithm" being ranked lower, that means these "other factors" are much more than a slight twiddling. And when you don't provide the full algorithm, just a hand waive, it makes it difficult to ascertain what's really happening. Are you sure there aren't abuses from your portfolio companies managers/employees to flag negative stories? I imagine Sam, for example, knows exactly what he has to do to get ChatGPT criticism guided off the stage. Edit: for example, do you know what happened with this story? https://news.ycombinator.com/item?id=35245626 https://news.ycombinator.com/item?id=35245626 This is a very interesting/important topic. This was a new topic. It was really hot in the first hour, and just got smashed off the front page.
- dang 4y ago
  > Are you sure there aren't abuses from your portfolio companies managers/employees to flag negative stories? I imagine Sam, for example, knows exactly what he has to do to get ChatGPT criticism guided off the stage. Quite sure. That is, there may be managers/employers of $companies trying to flag things, but being a YC portfolio company doesn't make that any easier. And yes I'm sure that Sam can't do that. (I also know that he wouldn't try, but that's a separate point.) Re the FAQ: it doesn't give a detailed explanation (we can't do that without publishing our code) but it summarizes the factors comprehensively. If you want to know more I need to see a specific link. Speaking of which: Re https://news.ycombinator.com/item?id=35245626 https://news.ycombinator.com/item?id=35245626: it was on HN's front page for 4 hours, and at some point was downweighted by a mod. I haven't checked about why, but I think most likely it was just our general approach of downweighting opinion pieces on popular topics. Keep in mind that the LLM tsunami is an insanely popular topic—by far the biggest in years—and if we weren't downweighting follow-ups a la https://hn.algolia.com/?dateRange=all&page=0&prefix=true&query=follow-up%20downweight&sort=byDate&type=comment https://hn.algolia.com/?dateRange=all&page=0&prefix=true&que..., it would saturate the front page every day. Actually we tend to not do that moderation on randomwalker posts (https://news.ycombinator.com/user?id=randomwalker https://news.ycombinator.com/user?id=randomwalker) - because they're basically always excellent. But a certain amount of randomness is inescapable and randomwalker posts do great on HN most lot of the time. If we made the wrong call in this case, so much the worse for us and I'm genuinely sorry.
  - p-e-w 4y ago
    > it doesn't give a detailed explanation (we can't do that without publishing our code) Precisely why would publishing (the relevant part of) the code be a problem? Twitter did it just a few days ago, and they aren't even known as an information hub of the open source world, plus they face a lot more public scrutiny for everything they do, to put it mildly.
    - dang 4y ago
      I'm not sure you've got an apples-to-apples comparison there (between what Twitter published and what I was just talking about). Either way, though, I don't want to publish that part of our code for two reasons: I fear that it would make HN easier to game/mainpulate, and I fear that it would increase the number of objections we have to deal with. It's not that I mind dealing with objections in principle, but a 10x increase would bury me.