News

Can advanced stats predict the Stanley Cup winner?

Benjamin Wendorf
By:

Stanley Cup (Getty Images) Author: The Hockey News

News

Can advanced stats predict the Stanley Cup winner?

Benjamin Wendorf
By:

Advanced stats aren’t perfect predictors, but they make playoff crystal balls clearer – and they can even help show how close a series will be.

When it comes to building a Stanley Cup winner, there’s a lot to be said for analytics and advanced statistics. When it comes to predicting a Stanley Cup winner, well, there’s a lot to be said for tarot cards, crystal balls, coin flips and a game of rock, paper, scissors. That’s due to the unpredictability of the playoffs. It’s not that advanced stats are useless for those trying to project the conference finalists, but we wouldn’t recommend taking them to the proverbial bank.

Part of the reason for that is the playoffs essentially cut the league in half, and that half consists of 16 teams in which the differences in quality are far more minute. Smaller differences make it more difficult to predict who’s going to win. Don’t trust us? Just look at the numbers. Over the past decade, conference seedings have been a successful harbinger of a series winner only 52.5 percent of the time – basically a coin flip. There are some indicators that can help you out, but even the best of those has just a 65.9 percent accuracy rate in predicting a series winner. And let’s face it: it’s just not sexy to only predict the series winner. To truly be a playoff prophet, you have to nail the series in the right number of games. So, let’s look at team 5-on-5 scoring chances-for percentage when the score’s close, our strongest testing metric in determining series winners.
Chart 1 Since the difference between the league-high and league-low frequently ranges about 16 percent and is fairly evenly distributed, we can divide that into four equal outcomes. A difference of 13 percent or more between teams in a series translates into a four-game sweep prediction. A nine- to 12-percent variance suggests a five-game series, a five- to eight-percent difference a six-gamer and anything under four percent should go seven.
Chart 2 If you’d compared the team scoring-chance metrics before the 2014 Cup final, you would have been right had you predicted the Kings would topple the Rangers. L.A. commanded 55.6 percent of 5-on-5 scoring chances in the regular season compared to New York’s 52.4%. (And those numbers remained true in the final, with the Kings getting 120 chances to the Rangers 96.) But the difference, just 3.2%, would have led you to believe the Rangers would extend the series to seven games. It went five, but three of the Kings wins were in overtime (with two in double OT), so perhaps there was less to choose from between the two teams than the result indicated. Close matchup, close results, but a 4-1 outcome – exactly the kind of numbers you’d expect from a 65.9 percent accuracy – and a competitive NHL playoff.
Chart 3 So, it’s nice to know that the teams that get the lion’s share of the scoring chances have the better chance of winning the series. That stands to reason. You can make some pretty reasonable assumptions about the playoffs using analytics, but you might want to keep that crystal ball handy, just in case.
– WITH KEN CAMPBELL
This is feature appears in the Playoff Preview 2015 edition of The Hockey News magazine. Get in-depth features like this one, and much more, by subscribing now.

Comments
Share X
News

Can advanced stats predict the Stanley Cup winner?