The Bonk's Mullet Advanced Stats Series - Part 1

Friday, October 18, 2013

by Chet Sellers


How often do you find yourself talking to a stranger in a doctor’s waiting room, and the subject turns to your favourite Senators blogs, and then they say something like, “Yeah, I like some of the stuff at Bonk’s Mullet, but they don’t do any . . . *lowers glasses to bridge of nose* . . . serious analysis.” Well, it’s happened to me twice now, and I’m sick of it!

It's no secret that statistical analysis is becoming more prevalent in hockey, even if it requires the use of obscure names and complicated abbreviations (Fenwick; WOWY; EvOZS%) to represent simple, intuitive concepts (puck possession; linemate chemistry; no idea on that last one, actually). It's easy to get intimidated in the face of alphabet soup like that, but rather than throw up your hands, why not put them in our silky mitts and let us lead you through some basic concepts? That's right - we're starting a new series here at RBM, where we take the fanciest #fancystats we can find, fancy them up, and then apply them, fancily, to YOUR Ottawa Senators. Let's go!

One of the objectives of statistical analysis is testing conventional wisdom to determine if it's correct. Here's an example - have you ever heard the old adage "look good, feel great"? Of course you have . . . but is it true? Does a hockey player that looks good in fact feel, and by extension play, great? We ran data for the season-to-date through a few sophisticated regressions and then drew the results in a notebook:

CLICK HERE FOR FULL SIZE IMAGE. UNDERLYING DATA PROPRIETARY TO BONKSMULLET.COM

In the graph above, we've put Performance on the x-axis and Style & Grooming on the y-axis, and even over a sample of seven games, that's a pretty strong relationship! We can see from our graph that the fresher a player looks, the more likely he is to be a stud performer. Usefully, we can then use this analysis to forecast the performance of some of the outliers - in the case of the Senators, we expect well-coiffed gentlemen like Marc Methot, Zack Smith, and J-G Pageau to improve their play over the course of the season, while tat-man Joe Corvo is a good bet to regress from his early-season scoring rate. And look at Erik Karlsson, all the way up in the Euro Zone - there's no way he can manage a whole season at his current looks-to-performance rate. He's either got to muss up his flow - not likely! - or pick up his game. That's good news!

But wait - did you remember to control for nationality? It's important to examine whether noise in the data is the result of grouping together too many disparate sets of players with their own sub-trends. For example, what happens if we isolate just Russian players?

WE DID THIS ONE ON A BETTER COMPUTER.
It's a completely inverse relationship! And that's really the lesson here - numbers are useless without context, proving that visual analysis will always have a place in the game. And we'll see you in Sochi, you talented, funny-looking Russians!

So what have we learned so far? We've learned that statistical analysis isn't so complicated once you know how to do it! Join us next time for something else!