Have you ever heard about the hypergeometric distribution? I haven't up to at least a few weeks ago. It is related to the binomial distribution in a sense that both of these distributions describe the probability to have certain number of successes after a given number of experiments. The difference between them being that binomial distribution assumes experiments to be independent (drawing the balls from the box with replacement), while hypergeometric distribution assumes dependence (the balls are drawn without replacement).

Let us construct a simple model for the hypergeometric distribution, and run simulations!

Presidential elections in the United States are quickly approaching. For this special occasion Financial Times has created an interactive game, which allows you to become party leader and try to win electoral college in 10 swing states. You will compete against other readers of the Financial Times. At my best attempt I have beaten 75% of the other readers. How many will you beat?

Fig. 1:Screenshot of the game.

Play the US election game »

The game itself is inspired by game theory and uses Colonel Blotto game as its basis.

This summer I had couple of experiences which ruined my opinion of open access. More than ten years ago I have felt that openness is crucial for future fundamental developments, but now I clearly see perverse practices even in top journals and publishers.

ritvikmath has already explained to us what the Beta distribution is. There is another video by ritvikmath, which explains multidimensional generalization of it. This generalization is known as Dirichlet distribution. Watch it for the data science perspective.

I myself understand Dirichlet distribution from the perspective of noisy voter model (or Kirman's model. Although I haven't explicitly mentioned it earlier, Dirichlet distribution arises from the multistate generalization of the voter model.

Another summer has gone by. This year my main summer activity was supervising summer internship project. The intern was rather capable, so the project itself was not very eventful and comparably effortless on my part. Also the final conclusion on the research hypothesis was negative (i.e., state interaction network doesn't seem to have any effect on stationary distribution), so there wasn't much to analyze.

From a technical side, once again, my institution's IT admins have irritated my a bit. In mid-June I have received a notification about outdated software on a server I supervise (which is not related to Physics of Risk in any way). I promptly connected to the server to immediately apply the update (as I was warned about severe security risks of not updating), just to find that no update is available for the Linux server. After quick search I have found that a newer version of the software in question is in fact directly available from its developers, but hasn't yet landed in the official repositories. It is not a good idea to update from outside the official repositories, so I felt somewhat annoyed due to being prodded and lectured for no real reason.

Summer experiences aside, what is coming for Physics of Risk? At the time of writing I do not know. I do have material for a rant about my summer experiences with open access publishing, but really not much else. I expect that the first-party content might become much rarer in the foreseeable future.