NHL analytics have exploded in popularity over the last decade.
Teams are investing in staff and building departments for the sole purpose of analyzing data and in the public sphere, it’s significantly enhanced the information and knowledge that fans and media have about players and teams.
The majority of the hockey world now agrees that data has
some value as a tool that can offer objective insights into player performance. The old days of intense analytics versus eye test debates are mostly gone because people understand that it’s not an either-or proposition, you should be using
both the eye test and data.
Hockey fans have a more sophisticated understanding of players around the NHL — especially players from teams they don’t watch — because of how accessible public analytics are.
That, of course, should be celebrated.
But if we’re going to rely on these tools, we should also be aware of their limitations and the cases where stats can be misleading.
I’ve been diving into analytical tools like shot shares, expected goals and PDO; types of microstats like zone exits, zone entries and passing data that Corey Sznajder tracks; powerful tools like
Dom Luszczyszyn’s excellent Net Rating model and more for years now. Today, I wanted to highlight examples where analytics were misleading about a player’s true value.
The purpose of this exercise isn’t to blame analytics. These are often situations where
I was personally wrong in putting too much stock in the data and not considering other factors.
Experience over time has made me handle analytics with more scrutiny.
Instead of drawing conclusions at face value because a model feels strongly about a player, I try and find the holes, blind spots and possible ways I could be wrong. When a player I need to offer deep analysis on is acquired,
I’ll watch hours of game film and dissect those observations to add a new perspective. I weigh a lot more qualitative factors now. And honestly, it all helps paint a clearer, more nuanced understanding of players.
Here are 10 players that analytics were misleading about, and some of the lessons I’ve learned as a result.