4 ms·
Wow you put a lot of work into this. What motivated you? It's interesting that the submitter based model wasn't more successful. You'd think people would center
by aliqot 4y ago
Wow you put a lot of work into this. What motivated you? It's interesting that the submitter based model wasn't more successful. You'd think people would center more around niches.
- PaulHoule 4y agoI have a hard time ignoring things so I want good filters. More of my motivations are written up here https://ontology2.com/essays/HackerNewsForHackers/ https://ontology2.com/essays/HackerNewsForHackers/ but a lot of that turned out to be factually wrong. For instance that "Apple overload" was pretty temporary because Apple had made a wave of product announcements. Today people might claim about "Twitter overload" but it will be something else next month. As for submitters I think that for what I am measuring the people who submit a lot of things and have high karma tend to be pretty average in what they submit since they've internalized some model of what does well on HN. I bet there are people with a very different profile but they would have fewer submissions so the model wouldn't be able to make strong conclusions... That logit model I used is well-calibrated and not prone to jump to conclusions if it has just a little bit of data. With the right model and tools you might store something out. Overall I think my submissions are pretty average but I have a list of web sites that I scan and every day you'll notice I post something from The Cornell Chronicle, World Nuclear News, phys.org, arstechnica, arxiv, etc... Most superposters have an efficient strategy for posting a diversified selection of articles that have broad appeal.