I mentioned in my previous blog entries how I used R and Python to generate predictions for previous NCAA basketball tournaments. For the upcoming tournament starting tomorrow, see "ncaa17.ipynb" for the IPython notebook and "ncaa17.py" at https://github.com/jk34/Kaggle_NCAA_logistic_trees_SVM for the Python script I used to generate the predictions for the upcoming 2017 tournament. The README.md file at the following link provides a report of how I conducted my analyses: https://github.com/jk34/Kaggle_NCAA_logistic_trees_SVM/blob/master/README.md I used getBPI.r to scrape the BPI rankings for each team. The predictions are in predictions.csv A few interesting predictions for the first round: -For the 1-16 and 2-15 seed matchups, most of the predictions predict the favorite to win with about 99% probability -11-seeded Xavier has a 64% chance to beat 6-seeded Maryland -6-seeded Creighton has only a 55% chance to beat 11-seeded Rhode Island -5-seeded Minnesota has only a 52% chance to beat 12-seeded Mid Tennessee -Other than the 8-9 seed matchups and 7-10 seed matchups, for the first round, the higher-seeded team is expected to win with probabilities around 60-90% Upsets are more likely to occur in the 2nd and 3rd rounds: -3-seeded Baylor and 6-seeded SMU is roughly a 50-50 game -1-seeded Villanova only has a 55% to beat each of 4-seeded Florida and 5-seeded Virginia -In fact, my predictions predict nearly an equal probability of winner of that bracket being either Villanova, Duke, Florida, or Virginia -7-seeded Saint Mary's has a 58% chance to upset 2-seeded Arizona. In the following round, it was a 55% chance to upset 3-seeded Florida State -1-seeded Gonzaga has only a 56% chance to beat 4-seeded West Virginia -1-seeded Kansas has only a 52% chance to beat 4-seeded Purdue -2-seeded Kentucky has only a 59% chance to beat 10-seeded Wichita State -3-seeded UCLA has only a 55% chance to beat 6-seeded Cincinnati Below are a few bar plots showing the probabilities of teams winning in their matchups:
0 Comments
Leave a Reply. |
AuthorHello world, my name is Jerry Kim. I have a Master's Degree in Physics and years of work experience in Image Processing, Machine Learning, and Deep Learning. I mostly have used C++, Matlab, and Python. I created this website to showcase a small sample of the things that I have worked on Archives
March 2017
Categories |