The COVID-19 pandemic has affected our society in various ways and has changed numerous events on the schedule for 2020. One such event that we were looking forward to was the end of the 2019-20 NBA season as well as the 2020 NBA Playoffs. Since the season was suspended for several months due to the pandemic, we were inspired to create a model that would simulate the playoff matchups had the season not been suspended. This model can be extended to predict future game matchup outcomes.
What it does
This model allows users to run simulations of NBA matchups and predict winners of future matchups, including playoff series matchups as well as end-of-season stat leaders in the major statistical categories.
How we built it
We utilized the data collected from the games played in the 2019-20 season to run regression prediction models and calculate Elo ratings for each team in order to predict the standings for 2019-20 as well as the matchups and results of the Playoffs and the NBA season awards.
Accomplishments that we're proud of
Utilizing the Elo ratings and simulating the remaining matchups was an interesting task. Our model was able to use several types of regression models to identify the best fit to best predict future outcomes.
What we learned
We learned how different types of regression models can be better suited for different situations. Such models include multiple linear regression models, logistic regression models, and lasso regression models.
What's next for NBA Matchup Winner Predictor
This model can be extended to predict outcomes in any future seasons. Additionally, this idea can be applied to predict outcomes in other sports, such as football (NFL) or baseball (MLB).
Log in or sign up for Devpost to join the conversation.