On Nate Silver and the future of hockey analysis

What, if anything, does New York Times blogger Nate Silver's resounding success in using fancy stats to project the 2012 U.S. presidential election teach us about how we should view statistical analysis in hockey?

If you avoided the barrage of pandemonium and partisan hackery permeating yesterday's national festival of waiting with bated breath for the faceless and outmoded Electoral College to graciously take into consideration citizens' suggestions of which corporate-backed candidate should spend the next four years in the White House...well, I can't say I blame you.
But if one person comes out of last night smelling like roses, it's statistician and New York Times political blogger Nate Silver. For those of you who are unaware, Silver gained notoriety during the 2008 presidential election cycle when he used state poll aggregation models to correctly project the winner in 49 out of 50 states. He did himself one better this time around, accurately picking every state in the nation.
What does any of this have to do with hockey? Well, Silver started out crunching numbers in baseball, developing the PECOTA projection system, before wading into the murky waters of politics. And, perhaps unsurprisingly, he faced the same type of "watch the games, nerd!" detractors in the latter as he did in the former. If anything, the derisive screams echoed even louder in the political arena with members of the media in the past week alone calling Silver everything from a "one-term celebrity" to a "joke [and] ideologue" to "thin and effeminate," because apparently that's relevant.

David Roher noted this clear parallel between the anti-stats hivemind in sports and politics in an excellent piece for Deadspin. What it essentially boils down to, in both cases, is a class of media pundits and inside operatives desperate not to lose their monopoly on disseminating analysis to the public and/or their team or political party.

The issue is a prominent one in hockey, where much of the mainstream media covering the sport has yet to adopt the statistical tools commonly wielded by the blogosphere. But the debate really isn't the stats versus eyes clash it's often made out to be; in both sports and politics, the conflict is far more often stats versus inferior stats.

In electoral coverage, that might mean focusing on a small handful of polls rather than weighted averages of all available data or, as was often the case with particularly misinformed pundits during this election cycle, making the fatal mistake of believing national polls are anywhere near as relevant as what state polls suggest about how the Electoral College will shake out.

In hockey, as the prolific Cam Charron noted in his piece on Silver's success, the error is usually in ascribing way too much value to things like a team's record in one-goal games, a hot month by a goaltender or even more frequently cited numbers like a player's goal total or the length of a club's losing streak, rather than paying attention to the stats that really matter when it comes to judging and predicting performance.

We know random fluctuations in a team's shooting and save percentage make score-adjusted shot differential a better predictor of their future record than their current record or goal differential is, at any point in the season. We know the tendency of a player's PDO to converge to 1000 usually means his Corsi number paints a better picture of his two-way value than his +/- does. We're starting to develop working theories on how well a team's ability to create scoring chances lines up with their time of possession and how important it is for players to enter the offensive zone with the puck.

While posts like this one have a tendency to attack straw man caricatures of those who insist watching games are all that matters, I'd argue that we need to become more intent observers of the sport to broaden our knowledge base and devise new areas of study and new events to quantify. What we need to do away with is reliance on bare-bones, traditional statistics that represent relics of a bygone era. To ape the title of Silver's own bestselling book, we need more of the signal and less of the noise.

The point of this article isn't to establish a false equivalency between the very-much-nascent field of hockey sabermetrics and Silver's election predictions. Even though the bounces probably went his way last night, Silver's methodology is certainly more advanced and predictive than anything we hockey fans have at our disposal via Behind the Net, Time On Ice, David Johnson's site or any number of tracking projects. But hopefully what last night hammers home is that spreadsheets and calculators aren't just for basement-dwelling, near-sighted nerds. We're all faced with the same questions (how many games will the Sharks win next season? How many years do Joe Thornton and Patrick Marleau have left in them as first-liners? Should Doug Wilson trade Michal Handzus for pork belly futures?) and fine-tuning ways to quantify what we observe is the best way to come up with the answers.