NCAAursinus <-
Talks
<-
Sean Forman <-
You Are Here
Next: Modelling Individual Games /
Previous: Monte Carlo Simulation
/
Up: NCAAursinus
- Before we can predict anything about the bracket, we need to be
able to model individual games.
- NOTE: Examples are taken from the magical 2004 SJU
basketball season
- One possibility: Flip a coin, Heads SJU wins,
Tails Liberty wins.
- Should each team win 50% of the time.
- Need to factor in team ability.
- AP Polls, Sagarin ratings, RPI, BCS, ELO are different ways to
do this.
- I used the Sagarin Ratings'
Predictor
rating
for the team quality.
- This combines all of the team's games and outcomes into a single
number.
- Assumes:
- Team ability level is consistent for entire season. Injuries,
hot streaks, are only relevant as to how they affected past outcomes.
- Home court affects teams equally.
- The probability of a team beating another is proportional to the
difference in their ratings.
- All NCAA tourney games are on a neutral court.
Next: Modelling Individual Games /
Previous: Monte Carlo Simulation
Copyright© 2005, Saint Joseph's University and Sean Forman