What happened?Hockey is a maelstrom of movement and almost every facet of it is a threefold-mixture of two extremely skilled teams and a great deal of happenstance and puck-luck. People have long sought to sift the signal of talent from the noise of randomness with statistics---describing just how unusual and how remarkable are the performances we watch every day. We understand the amazing by comparing it to the ordinary, and our visual intuition for making these comparisons is extremely strong---much stronger than our intuition for numbers in tables.
For instance, here's a graph showing where the Ottawa Senators took shots when Erik Karlsson, their star defender, was on the ice:
You can quickly see the kind of impact he has: the big red patch of shots that he takes personally from his usual spot at the right point; the way the team gets great shots from right in front, the way the team avoids he high slot and instead attacks along the goal line. I make many kinds of descriptive graphics that help people understand what happened -- who did what, who was amazing, who hurt their team, who was a non-factor.
I make a raft of different descriptive visualizations, simple and complex, covering single games, seasons, teams, and whole years, published every day of the hockey season on my website, hockeyviz.com.
What happens next?At the same time, we want statistics to help us separate the lucky from the good, to understand whose fortunes are about to turn for the better. Prediction requires us to isolate which features of hockey arise from the skill of the players, which of these skills can be repeated in the future, and which of these matter for winning. I make predictive models which I use to simulate seasons and estimate the probabilities of making the playoffs, of meeting particular teams, or of getting a high draft pick. For instance, here is a graph showing Minnesota's likely first-round opponent from last year:
They salvaged their season with a goalie trade and then maneuvered their way into a match-up with St Louis after seeming sure to meet Anaheim or Nashville. There are various outlets which produce predictions, but their methods are either shrouded in secrecy or known to be weak. I've written extensively about my methods and I publish all the details of models for anybody to read or use for themselves. I make many predictive visualizations---for the Presidents' Trophy, for the lottery picks, for specific playoff match-ups, point projections for individual teams, and much else.