February Development Update

After the election, where we met most of our accuracy benchmarks, I didn't really know what I wanted to do with this site next. The midterm elections were two years away and it would be hard to create accurate predictions this early. I considered studying the data around other elections abroad, such as the upcoming German federal elections, but the way I built my model made it difficult to adapt to other electoral systems.

I decided to turn to sports, specifically college basketball. I've always loved the sport and I personally wanted the challenge of predicting such a vast league. The process was fairly simple, all I needed was to grab some college basketball data (which ended up being far easier to find then election data) and edit a few lines of code. All in all, it took me around two days.

I thought about publishing immediately to the site but I held off to try to see how it would do predicting actual games. To be frank, it was pretty terrible at first, but after solving some of the issues (expanding data sample, factoring in recent form, etc.), it started performing pretty well. We expect that over the course of a season it would predict about 72% of games correctly and finish with a MAE of 9.2.

Yesterday, I finally felt like it was ready to publish, and after working out some kinks with the site we now have a page ranking all 364 teams and another page that goes deep into the possibilities for the rest of the Big West season. These will be updated either late at night after all games are played or in the morning after. I'll try to publish further articles going more in depth into the world of college basketball in the future as well. While I generally did not make many updates on the elections side, I did run our model for the upcoming 2025 elections. The ratings for those can be found on the front page and I'll publish a forecast on it as we get closer to Election Day.

That concludes this update on this major development to the site. I'll try to go more into detail about some of the projects I've added, as I know that the tables may be confusing. I also realize that I don't have anything up that actually predicts games, I don't have the ability to predict all games, but I can pretty easily get like a "Games of the Week" prediction table up or something like that when I can.