Creating a live win probability model based off of 4th quarter data from the 15-16 and 16-17 seasons
All cleaning of the data, making of the model, and looking at predictions is located in 'the-model'
Visualizations were created looking at the Oklahoma City Thunder vs Sacramento Kings game from December 6th, 2015.
In 'late-game' two different situations are examined.
First we look at situations where a team is down 1 with between 28 and 30 seconds left, and the opposing team has just gained possession of the ball. We analyze whether that team should foul or not. The data leads us to conclude that we should not foul.
Next we consider when we can rest players in the fourth quarter while having a small effect as possible on the outcome of the game. First we look at the beginning of the fourth. The data shows that if the team has a 15 point lead/deficit or more, then key players do not need to start the fourth quarter and likely do not need to go back into the game. Then we look at the 4-5 minute mark. If a team has about a 12 point lead/deficit, then their key players can be taken out of the game and the result of the game should be affected as little as possible.