Model Madness 2019: Sweet 16 Update

My quest for a prefect bracket ended after one game. All four models picked the wrong team for the first game of the day. Well, that was fun while it lasted.

Of course, even though all brackets were broken before the tournament really started, we can still see whether or not the models do a good job against other brackets and which model performs best against other models.

To see which models are performing the best in bracket pools, first it is a good idea to understand how brackets are scored. The traditional system doubles the points from round to round. This gives more weight to picking the correct champion. A correct championship pick is worth as much as 32 first round picks. In this system, each round can award the same number of points, since the number of teams halves each round until there is a champion.

Here's how the points look like for the four models below using the traditional method of scoring. Provided are the current points as well as the points still available to each model.

Given that the models are kind of chalky (tends to pick higher seed), a lot of them have near max points available. The EDLM which picked Nevada to win the whole tournament lacks these points as their champion is knocked out. Most models did a decent job of picking the first round, but struggled in games between nearby seeds. The MASPM currently has the lead in terms of points, but were lucky to survive yesterday where Duke escaped yesterday with a one point win.

Given that the competition was top-heavy, seeing so few upsets is not too much of a surprise. A few models picked up on a few upsets (Murray State, Oregon) while missing out on others (Ohio State, UC Irvine (projected close), Liberty(projected close)).

It will be interesting to see which model wins out among the different models and we'll see which of the three models are still competitive when the Final Four is decided this weekend. I'll see you then.