Modelling Individual Games   <- Talks   <-  Sean Forman   <-  You Are Here

   / /

Fitting to a Logistic Model

• A logistic model, $f(T_1, T_2, loc) = T_1$'s probability of winning.

• $spread = T_1 - T_2 - 4.1*loc$, where $loc = {-1,1, 0}$, for $T_1$ on road, at home, or at a neutral site.

• $${ f(spread) = \frac{1}{1 + c \times exp (a \times spread))}}$$

• Using a change of variable, we can use linear regression to fit a curve to this data, i.e. find $c$ and $a$.

• The best fit was $${f(spread) \frac{1}{1 + exp (-0.198 \times spread)}}$$ Graph of fit

• We will use this to provide the probability of teams winning individual games.

/