Penalty Bias

Predicting which NHL team will get the next powerplay opportunity.

About penaltybias.com

This website is based entirely on the paper Biased Penalty Calls in the National Hockey League by David Beaudoin, Oliver Schulte, and Tim B. Swartz.

The authors created a fitted logistic regression model according to whether the next ( $i$ th) penalty was called against the home team. This model is based on four covariates:

x_{1 i}

\equiv

total road penalties minus total home penalties in the particular match when the $i$ th penalty was called

x_{2 i}

\equiv

total road goals minus total home goals in the particular match when the $i$ th penalty was called

x_{3 i}

\equiv

the time in the match when the $i$ th penalty was called; $x_{3 i}$ ranges from the $0$ th minute to the $65$ th minute which is the end of overtime

x_{4 i}

\equiv

team strength parameter where values 1/0/-1 correspond to the stronger home team, evenly matched teams and a stronger road team based on regular season points

Using these covariates the fitted logistic regression model can be solved for the predictive value. This predictive value is used to find the chance of getting the next powerplay for a given team.

All things being equal, penalties are called on the road team in a 11:10 ratio.[1]

Code for this project can be found on Github.