The PollsPosition project is an open-source endeavor with three main goals:

  1. Exploring new statistical methods and applying them to real-life issues and datasets. This is a way for me to gain new skills and stay up to date with the field of Bayesian statistics, where the fun part is that you're never done learning! As such, I like to call PollsPosition my "nerdy sandbox" 🤓 Understanding how and why people make decisions when they don’t have all the facts is fascinating to me. That’s why I like, among others, electoral forecasting, and I use it as a sparring partner.
  2. Popularizing and illustrating the power of Bayesian statistics. These are especially helpful in contexts where data are sparse and imprecise, domain knowledge is important, and estimating uncertainty is one of the main interest of the analysis. Well look at that: democratic elections check all these boxes ✅ Icing on the cake, French political parties are numerous and change in nature -- compared to the US for instance -- which makes the models all the more complicated, but also... interesting! If you're curious, I presented a talk in 2020 about a version of the model:
  3. A last, broader goal is to try to counteract our natural tendencies (especially during electoral campaigns!) to cherry-pick data, overreact to the latest poll, and, most importantly, completely misinterpret uncertainties and probabilities. I was invited on a podcast in 2020 to talk about just that, if that's of interest to you 📻 I'm perfectly aware that it's a lofty and probably unattainable goal. I do believe that spreading the methods of rational and critical thinking is essential though, so I can at least try 🤷‍♂ For sure, I can't do it alone, so if you like what we do at PollsPosition, feel free to share our content with your friends and colleagues -- or if you don't like it, share it with people you don't like, that works too!

If this all sounds fun to you and you're looking for a project to improve your Python and Bayesian chops, feel free to contribute pull requests -- there is always something to do!

My name is Alexandre Andorra by the way. By day, I'm a Bayesian modeler at the PyMC Labs consultancy and host the most popular podcast dedicated to Bayesian inference out there -- aka Learning Bayesian Statistics . By night, I don't (yet) fight crime, but I'm an open-source enthusiast and core contributor to the awesome Python packages PyMC and ArviZ .

An always-learning statistician, I love building models and studying elections and human behavior. I also love Nutella a bit too much, but I don't like talking about it – I prefer eating it 😋

I can't finish without acknowledging the people who help me in this nerdy adventure, most notably the brilliant Alexis Bergès who devises with me for hours about ways to best model elections, as well as the wonderful core-developers of ArviZ and PyMC, who often indulge my unrelenting stats questions 🤩

Feel free to reach out on Twitter if you want to talk about chocolate, statistical modeling under certainty, or how "polls are useless now because they missed two elections in a row!" -- yeah, I'm a bit sarcastic.