Over this past weekend the second annual RIT Hockey Analytics Conference took place. Ryan Stimson and Matt Hoffman once again brought together the analytics community to Rochester, NY for the latest advances in the field.
The event ran flawlessly and there was plenty of information to sift through. I won't be breaking down every single speaker in full detail, but I will provide some general thoughts about some of the topics. The presentations were filmed and are available to view for those that want to see everything in detail.
Killing Time: Analyzing Zone Exits on Penalty Kills - John Fischer (All About the Jersey)
John did a great job of breaking down the past two seasons of the New Jersey Devils penalty kill. He manaully tracked the different ways New Jersey was exiting the zone on the PK and what impact it had on their ability to kill penalties.
Fischer made the point that the Devils are the lowest scoring team since 2012 and that something like the PK matters more to them. He separated NJ zone exits into eight different categories on the PK. At the conference all speakers are limited in how long they have to speak. John recorded this YouTube video that goes into more depth than what he talked about at the conference.
This type of tracking is in its infancy and it will be interesting to see if other people take the time to track other teams to see if there is a correlation between success and the different kind of exits
Advancing Statistical Analysis of Goaltending - Nick Mercadante (Hockey Graphs)
Nick Mercadante does some of the best goalie work out there now in my opinion. His Mercad/60 stat isn't the end all be all, but nothing is. Being able to evaluate goaltenders based on their current peers is a great analysis tool and one that can be used in any time period.
He also raised a great point about the current state of goalie analysis
Mercad: Majority of public offerings on goalie are descriptive. Predictivity is elusive. #RITHAC
At the present moment we do a better job at saying what happened, not what is going to happen.
For Nick's presentation this past Saturday he visited the concept of Win Threshold Percentage. Nick claims that goaltenders play about 48 minutes of even-strength action each game and face on average 23 shots. Using this data he is trying to build a win threshold stat that should indicate how well a goalie has to play to give their team the best chance of winning. Goalie wins are a brutal metric to analyze the position. By creating a win threshold percentage he is hoping to eliminate goalie wins as something we talk about anymore (I agree with him).
Here is the information from his initial Win Threshold Percentage findings
Defending the Pass - Matt Cane (Hockey Graphs) and Ryan Stimson (Hockey Graphs)
This dynamic duo teamed up to discuss how shot assists and the ability to defend against them are paramount in playing defense. Currently we do not have a lot of statistical options to evaluate the defensive side of the puck. By tracking Shot Assists Against per 60 Ryan and Matt are taking the firsts steps to gather this kind of intel. A shot assist is any pass that leads directly to a shot attempt. Colorado was the worst team at hockey at preventing shot assists and that should come as no surprise because of how unorganized Patrick Roy's team was. Ryan had video evidence of why Colorado was bad at this and why the Florida Panthers were great on this front. Florida was very aggressive with their puck pursuit all around the rink while Colorado was passive and chased a lot. The aggressiveness led to less shot assists for Florida.
Matt Cane "randomly" chose two players to analyze defensively. Those players were PK Subban and Shea Weber. The perception is that Shea Weber is a much better defensive player while PK Subban's value comes on the offensive side of things.
There is plenty more work to be done on this front but the early findings are encouraging and with more tracking and information we can definitely start to analyze players' defensive abilities better.
Maximizing NHL Player Lineups Using a Linear Optimization Model -Asmae (Hockey Graphs) & Dawson Sprigings (Hockey Graphs)
In another very good presentation Dawson Sprigings aka DTMAboutHeart discussed the merits of stacking lines versus spreading the talent out. Basically he is trying to figure out lineup optimization with tangible evidence.
Here is how he looked at things
Here is what he found and how he did it
Basically he found that "loading" up lines over the long haul is not getting the most out of those players. At a certain point you get diminishing returns on the value of those higher end players. Placing three average players together helps get bang for the buck. Loading up with a super line over the course of a full season will see minimal optimiation. In speaking with Dawson after the conference he had mentioned Pittsburgh and how they split Crosby, Malkin, and Kessel as one of the reasons he looked into this.
Optimal Roster Structure in Hockey - Eric Cantor (Shippensburg University)
Eric did some extensive research with the Tampa Bay Lightning. He broke down whether or not it was better for Tampa Bay to run with the classic 12F 6D approach or if the 11F 7D approach was more beneficial. Tampa Bay has gone with the 7D approach more often than other teams which made them ideal for a study like this. He found that Tampa did much better with the 7D approach. Most stats saw improvement including things like zone entries and shot assists percentage. Not all teams have the ability to do this and this research is only scratching the surface by covering one team but it is a start and is something to keep track of as we move forward.
Hockey Media Panel
Scott Cullen (TSN), Alison Lukan (Bluejackets.com & buckeyestatehockey.com), Bill West (Pittsburgh Tribune Sports), Carolyn Wilke (Site Manager and Editor, Today's SlapShot)
This was a great panel and talked about the media in hockey. Scott Cullen had a few noteworthy items he shared including:
"I'm not buying into the idea that you can win with better locker room culture. You need better skill to win"
He specifically mentioned that the Pittsburgh Penguins won because they had skill more than any other reason.
"We can all appreciate that no one wants to work with a jerk. Skill matters more."
He then mentioned how Toronto was better with Phil Kessel and worse without him. Also mentioned the correlation between Pittsburgh getting better and somebody like Phil Kessel being added to their team.
"Skill trumps culture when trying to win hockey games."
"Corsi isn't an answer. It's the start to more questions"
This quote is excellent. It addresses the strawman argument people try to use quite frequently. Nobody claims Corsi to be the god particle of hockey analysis. It takes the temperature and allows for a great starting point to analysis in order to answer many other questions.
Carolyne Wilkie discussed how online content makes things easier to visualize points and how traditional media cannot compete with that. Examples included embedding tweets, YouTube videos, gifs, etc... Blogs are also not on word counts. Endless "paper".
Alison Lukan discussed a lot about how she communicates with both players and coaches and how she tries to get informative answers from them. She explained on how using "statsy" terms like Corsi are a good way to turn off players/coaches from answering the kind of questions she wants to ask. She has to craft things in a language that is inviting for her interview subject in order to ultimately get the best and most descriptive answers. She talked about talking with Jody Shelley and him describing shot attempts (on goal, missed, etc) and how Shelley said he doesnt understand "all that Corsi stuff". She also spoke about being able to go into the room the next day and face the players and coaches that she covers (maintaining professionalism)
Bill West did a great job of explaining how he goes about his job in a traditional media outlet and attempting to use modern statistical analysis. He talked about how most people go to the paper to get the game score and they aren't necessarily looking for much more than that. He tries to focus on using phrases like shot-attempts as opposed to 'Corsi' in order to cater to his specific audience. He spoke about "easing" the audience into the stat conversation by doing little things like that. Bill had a great quote about what he is trying to do
"You don't write a stats story vs a traditional story. You write a story that merges both together"
Again, a great panel.
Drafting Overagers: Have the Leafs Uncovered the Newest Market Inefficiency?
Sean Tierney (Hockey Graphs)
On top of live tweeting RITHAC like a machine Sean had his own presentation to give which covered the topic of drafting over age players and if there is a benefit to it. He found that it makes more sense to draft second or third year eligible players in the later rounds. There is not a lot of value in doing it early.
Gap in value from pick 1 overall to pick 3 is same as pick 24 to pick 210. So if you want to pick overager after pick 24, have at it #RITHAC
Draft Order Incentives and League Competitiveness -Micah Blake McCurdy (HockeyViz.com)
Last and certainly not least Micah closed out the conference with his idea to change the draft lottery. He believes that the current system does not have the fans, players, coaches, management, and owners best interests aligned. He borrowed an idea he saw from the Sloan conference called Gold Drafting. In this concept teams with the most points after being eliminated from the playoffs will get the top picks in the draft.
McCurdy shows that statistically, Gold Plan: improves parity; doesn’t unduly punish very weak teams; “earlier tanking” has no benefit.