Wikipedia:Reference desk/Archives/Mathematics/2014 November 12

= November 12 =

Craigslist Personals and Visibility Strategies
This has nothing to do with actual CL, just inspired by thinking about how it works. Essentially, as you post new ads, it drives other ones down the page (like a forum, but with no way to bump them back up). Suppose we have a function R(k) that is probability that a random viewer will read the first k posts, that there is N so R(1 + N) = 0.0, and that there are, at least, N posts. Suppose further that every person viewing the section has probability p of posting. Given that the number of viewers varies throughout the day, would you get more reads by posting during low traffic or high traffic, or does it not matter? My assumption is that since posting during low traffic extends the life, but lowers readers; while high traffic lessens life time, but has more readers per time unit, that, ultimately, it all balances out. But, I'm not very good with these types of problems and was curious what the actual answer is. Thank you for any help:-)Phoenixia1177 (talk) 10:58, 12 November 2014 (UTC)
 * It doesn't matter. It's easiest to see if you disregard the concept of time altogether. There are people visiting the forum, each has a probability to see your post and a probability to post and affect future people. If you put all visiting people in a sequence, these effects and probabilities have nothing at all to do with the time in which each person visited. "Low traffic" and "High traffic" talk about the temporal gaps between the people in the sequence, which don't matter. -- Meni Rosenfeld (talk) 12:19, 12 November 2014 (UTC)
 * It's easy to also see that the mean amount of views is $$\frac1p\sum_{k=1}^{\infty}R(k)$$. -- Meni Rosenfeld (talk) 12:29, 12 November 2014 (UTC)
 * Thank you very much:-) That is a much cleaner and reasonable account of what I was thinking - thank you, again, that is extremely helpful:-)Phoenixia1177 (talk) 13:56, 12 November 2014 (UTC)


 * As a practical point, posting takes more time than visiting the page (which determines what you see even if you take a while to read it), so there is a timescale present which introduces a dependence on the derivative of the traffic: you want to post as the traffic is increasing so that the new traffic sees your post while it is still being demoted only by the lighter traffic preceding. --Tardis (talk) 14:31, 12 November 2014 (UTC)


 * I didn't think of that, that is a very clever point (I would have never considered it) - thank you for pointing that out:-)Phoenixia1177 (talk) 14:37, 12 November 2014 (UTC)


 * I think that in the real world, traffic at time of post will definitely affect viewership. But the effect does not seem to be captured in your model. In real life, probabilities of posting are of course conditional, and will also depend on the content of previous posts (e.g. I have to respond to what he just said!) Also in a given community specific high-frequency posters might all be online at the same time, and that will affect things too. Meni is right about ignoring intervals between posts, but we can't ignore time when thinking of readers, because many can read at once. Also time a certain reader spends can affect things, because some posts will be too long to read in a given time period. Reddit would be a bad example for this question because of its voting, but there is a well-known joke there about how the Australians get up to hijinks when the Americans are sleeping. In part this works because total volume is lower when it is deep night in USA. Sites that don't have ways to move post include metafilter.com and 4chan, you might find some sympathizers on metafilter who could share data for further analysis. Anyway, my point is that if you add detail to your model, I think it might eventually display some differences in post veiwership based on total traffic flow at time of post. SemanticMantis (talk) 16:34, 12 November 2014 (UTC)