3 ms·

Thanks for posting this interesting data set, but after thinking about it for some time, I don't think this data offers any insights into yc.news. The leader to

by Spyckie 19y ago

Thanks for posting this interesting data set, but after thinking about it for some time, I don't think this data offers any insights into yc.news. The leader to user ratio is too high, which skews all the numbers so that nothing really meaningful can be extracted. It's also hard to spot any trends from the point system alone - user identity, time of post relative to other news, and the controversy of the post content can play important factors in the post's point value, and all of these don't necessarily have any correlation with the user who posted it. I think in order to figure out something about the community, someone will have to get their hands dirty and actually look at post content.

matstc 19y ago
There is a lot of talk about what should be done in order to keep the community as it is, about what is a relevant post and what's not. But we have no metric to measure the cohesion, except for 'pleasure'. The analysis in the post is of course incomplete but I believe there is value in a metric that could tell us whether or not we are being steered away from the initial community.
cawel 19y ago
Those are nice stats, and I do think we can get some interesting characteristics about our community. I thought it is quite powerful to point to paul, palish, sharpshoot, or pg, and say that they are the community, since they have the best points per post measure. They are nothing less than the best representatives of our community (with regards to posts, not comments). I'd like to think that the factors you mentioned (traffic, time of the day, controversy of the post) iron out statistically (since all posts are affected by those, there is an argument to ignore their respective influence). This initiative makes me wish it is only the beginning: I want some more stats! For example: - It would be interesting to see what's the points per post (PPP) of the leaders, compared to the PPP for most of the others (I agree with you that in the dataset, the leader to user ratio is too high). Maybe one could have a plotted distribution of PPP's across all users. - I know HN has 8000 uniques/day, what is the proportion of those that are posting? e.g. how many of those users with less than 5 posts? - also interesting is the proportion of points between posts and comments, per user. Different profiles: some people discuss a lot, some people post a lot, as it was often repeated in discussions. But what's the proportion of those 2 groups?

by Spyckie 19y ago

matstc 19y ago

There is a lot of talk about what should be done in order to keep the community as it is, about what is a relevant post and what's not. But we have no metric to measure the cohesion, except for 'pleasure'. The analysis in the post is of course incomplete but I believe there is value in a metric that could tell us whether or not we are being steered away from the initial community.

cawel 19y ago

Those are nice stats, and I do think we can get some interesting characteristics about our community. I thought it is quite powerful to point to paul, palish, sharpshoot, or pg, and say that they are the community, since they have the best points per post measure. They are nothing less than the best representatives of our community (with regards to posts, not comments). I'd like to think that the factors you mentioned (traffic, time of the day, controversy of the post) iron out statistically (since all posts are affected by those, there is an argument to ignore their respective influence). This initiative makes me wish it is only the beginning: I want some more stats! For example: - It would be interesting to see what's the points per post (PPP) of the leaders, compared to the PPP for most of the others (I agree with you that in the dataset, the leader to user ratio is too high). Maybe one could have a plotted distribution of PPP's across all users. - I know HN has 8000 uniques/day, what is the proportion of those that are posting? e.g. how many of those users with less than 5 posts? - also interesting is the proportion of points between posts and comments, per user. Different profiles: some people discuss a lot, some people post a lot, as it was often repeated in discussions. But what's the proportion of those 2 groups?