Vox Charta has over last few years become one of more prominent tools in every astronomers arsenal. For those who might be unfamiliar with the concepts like Vox Charta and arXiv, very shortly, on Vox Charta website members of the participating academic institutions can “upvote” or “downvote” papers that have appeared on the Internet (arXiv). Idea is that people will upvote papers that they found interesting and want to talk about on the next discussion session in the department. Everybody can see how many votes a paper has received and one can easily see which papers are “hottest” i.e. which have spurred most interest in the astro community. Let’s see how does the number of votes on Vox Charta in the 2014 correlate with some other parameters!
Above we see that publication position of the paper strongly correlates with the number of votes above position 20 on the arXiv list (Lines show poor broken power-law fit to the data, done with “eyeballing” method). Below position 20 trends seems to stabilize. Scatter increases at very high numbers simply because there are very few days when 60+ papers are published. Interestingly, first position does not mean also the largest number of votes. It is important to note that there is significant number of papers that tend to be first on the list but were not actually first ones to be submitted after the deadline; they were usually submitted day or so before and I assume that there was some problem which caused them to be published with delay through moderator action.
Different days of the week spur different number of votes. Day with most activity seems to be Wednesday and the slowest day is Monday. It also seems that astrophysicists like to upvote papers more in the middle of the week. Even thought there is some difference it is only at about 20% level.
This difference is largely driven by the number of papers that are published each day. Papers published on Tuesday seem to be having lowest number of votes and Tuesday also seems to be only significant outlier.
Distribution of votes is highly non-uniform. In plot above, we show cumulative distribution of votes that papers receive. So, for instance one can see that almost 40% of papers receive no votes, and around 80% of papers receive 5 or less. Having 10 votes is already being in the top 10%, while cca 18 votes are needed to break top 5%.
Ok, so if one wants to be on the top of the arXiv list and (perhaps) have a better chance of getting more votes, how quickly should should the paper be submitted?
We show three lines which show different speeds of filling up. In blue, results are shown for 10% days which have reached 20 papers submitted the quickest. In orange mean is shown and in green we show results for slowest 10% of days.
On average, submitting in around 20 seconds after deadline will secure one of first five positions. After initial rush is over in cca 1 minute, things slow down considerably.
Ok, so you want to be first on the list. How quick do you have to be to succeed in that mission? Data shows that in order to have 50% probability of success paper has to be registered by arXive in the first second and this has no strong dependence on the day of the week when the paper is published. This does not take into account the before mentioned effect, that even if you submit first you might not get first place, because of moderator’s action.
Being in top 5 is somewhat easier and shows stronger day dependence As one can see above, submitting within first 20 seconds should place the paper in the first 5 positions. Competition is much weaker for Monday and Tuesday submissions then for other days of the week.