20
\$\begingroup\$

Back when I was new to Code Golf SE (then PPCG), a lot of my questions didn't do too well, and I didn't have the experience to tell whether they were good challenged or not. As such, I would often blame it on anything I could, like post time.

A few years later, I've started to wonder about whether or not post time affects post popularity. Long story short: it doesn't.

At first, it sounds plausible. If you post during a time when few people are online, such as when major timezones would be asleep, maybe fewer people would answer, vote, or comment, and your question would be buried.

I ran a simple SEDE query on Code Golf (SELECT ViewCount, CreationDate FROM Posts WHERE PostTypeId = 1 or similar), and copied the ~10k results from the downloadable CSV to a JS script (JavaScript, as I found out, is not great at data analysis). I created this badly formatted graph, on a logarithmic scale:

Graph of results, basically just a random scatterplot

The X axis is time, from 00:00:00 to 23:59:59 (UTC). The Y axis is logarithmic (radix 10), with each longer line (every 5 short lines) being 1.

As you can see by the line of best fit (it looks a little high to me, but IDK), post time has pretty much no affect on popularity. The slope of said line is somewhere in the range of -21 views for each hour later in the day.

Being the nerd that I am, I noticed a seemingly higher density of dots to the right, indicating more posts, so I created another graph, which is oddly sinusoidal in appearance:

A bar graph showing low question rates, then high question rates

The X axis is hours (UTC, floored), and the Y axis is total number of questions, on a linear scale of 0 to 750. This graph is slightly more useful, showing that the highest number of questions are posted between 14:00:00 and 21:00:00 UTC. This is between morning and mid-afternoon in the US, and late afternoon to night in the UK.

I can't seem to find much of a purpose for this information, but do what you want with it. I guess I wasted my time so you don't have to, although it's entirely feasible that I'm the only one boring enough to even think about this (:.

Part two: The relationships between question length, views, and votes

||||||
\$\endgroup\$
  • 10
    \$\begingroup\$ The line of best fit is being pulled up by the outliers, which are much higher than they seem thanks to the log scale. :) \$\endgroup\$ – El'endia Starman Nov 10 '19 at 5:30
  • \$\begingroup\$ @El'endiaStarman Makes sense, I thought it might have something to do with it \$\endgroup\$ – Redwolf Programs Nov 10 '19 at 5:32
  • 5
    \$\begingroup\$ This is great and not at all boring. But I think you should look for features of questions which are correlated with number of votes. How about the number of characters in a question for example? Or the number of $ signs? \$\endgroup\$ – Anush Nov 10 '19 at 7:42
  • 5
    \$\begingroup\$ @Anush I think I'll do that; this was a lot of fun. Thanks for the suggestion! \$\endgroup\$ – Redwolf Programs Nov 10 '19 at 14:51
  • \$\begingroup\$ I can't speak for everyone, but when I browse CGCC for the first time in a day, I check the list of recently posted challenges to see if I've missed anything good. \$\endgroup\$ – Jo King Nov 10 '19 at 22:41
  • \$\begingroup\$ @JoKing That seems to be what most people do, judging by the R squared value of 0.00037 on that line of best fit \$\endgroup\$ – Redwolf Programs Nov 11 '19 at 1:55
  • \$\begingroup\$ @Anush I did some more analysis (post length and views, views/votes): codegolf.meta.stackexchange.com/questions/18262/… \$\endgroup\$ – Redwolf Programs Nov 11 '19 at 3:52
  • 2
    \$\begingroup\$ It kinda makes sense: more people are online to see the answer at peak times, but more people are posting answers, so fewer see yours. Interesting that the two factors cancel almost completely. \$\endgroup\$ – Dewi Morgan Nov 11 '19 at 17:56

You must log in to answer this question.

Browse other questions tagged .