Hockey fans, analytic gurus make compelling cases in Oilers Hackathon
|Oilers Hackathon 2.0 was launched in December 2012.
Five months ago, Edmonton Oilers fans were given the opportunity to become insiders with the team.
The club opened up its vault of in-depth and historical analytical information, allowing contestants to launch their own methodology of advanced data to help the Oilers win games. Entries were to be measured real game results, and the best submission by year's end would win a position on the Oilers' analytical team -- a staff backed by Edmonton-based Darkhorse Analytics, who's worked with the team for several seasons now.
With the chance for participants to delve deep into information dating as far back as 1979, the Hackathon 2.0 was born.
One of the most compelling submissions has come from New York resident Michael Schuckers, an Associate Professor of Analytics at St. Lawrence University, and a hardcore hockey fan that heard of the Hackathon through social media.
"It seemed like an interesting challenge," said Schuckers. "I'm actually teaching a course on data competitions this semester and was hoping that my students could actually participate, but the timing didn't quite work with the spring semester."
Broken down to the bare essentials, Schuckers' entry is a comprehensive rating of forwards and defencemen based upon all on-ice events recorded by the National Hockey League's Real-Time Scoring System (RTSS).
The end result is what he's dubbed a Total Hockey Rating (THoR), which is ultimately a real-life version of EA Sports' ‘overall rating,' seen in their popular NHL video games, or a more complex and all-inclusive version of a pitcher's WHIP in baseball.
According to the study, the goal "is to create a rating system for NHL forwards and defencemen that values their role in creating goals, as well as preventing them."
"The idea is to use that information to extract the individual value of a player and their individual impact on what's happening on the ice during a game," said Schuckers. "How we do that is we give a value to each event recorded by the NHL, account for who the player is on the ice with for and against, where their shift started (zone starts) and who's playing at home and on the road.
"We're pretty happy with the results so far. The way that we've really been addressing how accurate it is, is when players switch teams and determining whether or not that value is consistent. We didn't have a whole lot of player movement this year because of the short season, but we've been pretty happy with how it's done in terms of addressing that."
Schuckers applied the methodology to the results of the 2010-11 and 2011-12 seasons, cases in which over 300,000 individual on-ice events were recorded, collected and extrapolated to determine the League's "top players" based on the Total Hockey Rating.
"We've found that the NHL's most impactful forwards and defencemen are worth as much as or more than five wins per season," said Schuckers, adding that the study primarily focuses on even-strength data but extensions on special teams (both the power-play and penalty kill) are available.
In order to calculate a win, Schuckers' study uses a formula that models how many goals a player creates over the course of an 82-game season.
Some of the results are surprising. Understanding that there's more to determining a player's worth than the obvious could mean that the NHL is now entering a previously untapped network of information and pro scouting.
It's something Schuckers believes could revolutionize the sport. With baseball out in front, football not far behind and basketball now entering the world of advanced statistics, it's time hockey gets involved because "the potential is very, very big."
"What we've seen in other sports is that those early adopters of methodologies like this or another analytical approaches have gotten some big benefits from it," said Schuckers. "As a result, certain teams in other sports such as the MLB or NFL are way ahead of the curve, forcing other teams to play catch-up.
"We'll get there."
Other, but equally as compelling entries in the Hackathon have included measuring the effect of coaching strategy on standing points, which looks at three main points of study concerning line matching, zone starts and line composition. In terms of special teams play, a number of submissions dug deeper into a team's success (or lack thereof) on the power-play and penalty kill, and whether that data accurately predict regular season results.
Certainly, a wide variety of information and methods have been submitted.
"The Edmonton Oilers are impressed with the level of enthusiasm and knowledge provided by Oilers Hackathon 2.0 participants," said Oilers President of Hockey Operations, Kevin Lowe. "It has already proven to be a great way to identify new ways of looking at hockey and opened new discussions.
"We are fortunate to have such a strong following, and it is obvious that our fans and supporters include mathematicians, analysts, computer programmers and other industry users of analytic techniques who are focused on trying to help us make more informed choices."
"I think what the Oilers are doing is tremendous," added Schuckers. "The idea of making this data available is a wonderful thing and I think from the Oilers' perspective, it's a smart thing for them to do. They get to look at ideas from a lot of people who have spent a lot of time analyzing their data, and to be able to harvest those ideas and be able to pick through them, I think is a really smart initiative."
With eight games remaining in the 2012-13 season, participants' entries will soon be graded and a winner will be chosen.
-- Ryan Dittrick, edmontonoilers.com | Follow me on Twitter @ryandittrick