hockey: NHL hockey data

Description

Every NHL goal from fall 2002 through the 2013 cup finals.

Arguments

Value

goalInfo about each goal scored.
oniceMatrix with entries for who was on the ice for each goal: +1 for a home team player, -1 for an away team player, zero otherwise.
configSpecial teams info. For example, S5v4 is a 5 on 4 powerplay, +1 if it is for the home-team and -1 for the away team.
playerPlayer info: names, most common position, and plus-minus.

Details

The data comprise of information about play configuration and the players on ice (including goalies) for every goal from 2002-03 to 2012-13 NHL seasons. Collected using A. C. Thomas's nlhscrapr package. See the Chicago hockey analytics project at github.com/mataddy/hockey.

References

Gramacy, Jensen, and Taddy (2013): "Estimating Player Contribution in Hockey with Regularized Logistic Regression." http://arxiv.org/abs/1209.5026.

Examples

Run this code

data(hockey)
x <- cBind(config,onice)
y <- as.numeric(goal$whoscored=="HOME")

## fit a lambda (gamma=0) model
plot(fit <- gamlr(x=x, y=y, family="binomial",  
      free=1:ncol(config), standardize=FALSE, 
      thresh=1e-8))

## BIC selected player coefficients
B <- coef(fit, s=which.min(BIC(fit)))[colnames(onice),] 
length(B <- B[B!=0])
## print nonzero player effects on who-scored log odds
round(B[order(-B)],3)

## predict the last goal of the 2013 playoffs (this time use AIC)
goal[fit$n,]
predict(fit,x[fit$n,,drop=FALSE],type="response",k=2)