May 07, 2014

Pucks On The 'Net: On Hockey Analytics

Hockey analytics - fancy stats to some - have certainly become all the rage in recent seasons.

Corsi. Fenwick Close. PDO. These are all a bunch of poorly named stats that do not exactly welcome newcomers. But do your homework because they are not going away. They are over-embraced by some members of the influential new media.

Make no mistake, these statistics all have some merit on their own. And the people working on these stats should be applauded for trying to help us all better understand the game. But the problem is you add them all together and they don't really tell us too much, regardless of what the new generation of bloggers believe.

Brian Burke's recent comments on hockey analytics were bang on in my opinion. Burke has been weirdly miscast as a fancy stats basher (much like I'm sure the big believers will miscast me after this article) but the truth is he is very much open to new ways.

“The first analytics system I see that will help us win, I’ll buy it and I’ll pay cash so that no one else can use it,” Burke said.

Why has he been miscast? All because of one of my favorite quotes of all time.

“What I think has happened is one quote from the MIT conference has been broadcast all over where I said, ‘statistics are like a lamppost to a drunk -- useful for support but not for illumination’,” he said. “I also think people confuse statistics and arithmetic and mathematics with analytics. Analytics to me are ‘can you take data and do some predictive work that will help me draft or trade better’?

“I haven’t seen a system that comes close to doing that. Statistical analysis about faceoffs and where guys play, we use that all the time. We’ve been using it for 20 years. To me, that’s not analytics. Anyone here that has a system worth buying, we pay cash.”

I recently saw a great example on Twitter. In a conversation about Sidney Crosby's recent goalless drought one notable blogger was quick to point out Crosby's low PDO numbers. That's a possession stat which simply adds the player's shooting percentage with the team's save percentage when he's on the ice. Not that complicated.

But I completely fail to understand what this stat had anything to do with Crosby's goalless drought. Sure, his PDO is low because he's not scoring. And he's been on the ice for a few goals against. But doesn't that boil down to a fair amount of luck? In a season's worth of data maybe there's a little more value but over a six game sample size?

The fancy stats guys couldn't tell me why Crosby has had trouble scoring goals lately. Or at the Olympics. They were quick to note their numbers to support their theory. Support. Not illumination. Contradictory support at that, because his Corsi (an improved +/- stat based on shots attempted for and against) suggests Crosby has been one of the top possession players much of this post season.

I am not about to suggest that I know why Sidney Crosby has had trouble scoring in the playoffs. If I had the answer Brian Burke would be paying me a lot of money. Maybe it's an injury or fatigue. Maybe it is the quality of opponents he faces every shift. Or maybe it is just really hard to score goals in the playoffs. Every NHL player would agree with that.

Trying to understand hockey at a higher level is a noble undertaking. But there is no substitute for watching the game rather than studying spreadsheets. The problem is most of us, despite what all hockey fans think at times, will never be able to watch hockey like Scotty Bowman or Ken Hitchcock or Mike Babcock. We all wish we could understand just one game like they do.

And that is why fancy stats are not going anywhere.

1 comment:

reoddai said...

If you're interested in lengthy reads, I suggest this blog:

Its a post about a guy trying to explain New Jersey's strategy using stats to pin point where in the game he should look.

In response to your post (to the best of my understanding). Crosby's high corsi does indicate good puck possession. I;ve also read that PDO tends to regress to the average. That mean that Crosby's low PDO and high corsi suggest that he's going through some bad luck. I agree though, that those states are at the point where they can tell him what to do. Hopefully those who are analyzing the game are also looking at statistics for the individual player (maybe, where they shoot as well as average shooting distance from the net. Things that are solely the effect of the individual player)