Notes on the Alberta Analytics Conference

You may have heard about the September Hockey Analytics Conference hosted by Rob Vollman of HockeyAbstract.com. It took place in the Calgary Saddledome, and was recapped by notables like Scott Cullen of TSN. November's conference was nothing of this sort: it was in a smaller venue, had no media presence, and was more discussion-based than presentation-based. Nevertheless, it was a great meeting, and I'll try to give a quick rundown below of all that went on.

Rob Vollman opened the event by talking a bit about where he gets his data from. He showed us a bit on his own website, hockeyabstract.com, including player usage charts, and links he gives to lots of other stats websites. It also features articles on newer advanced stats, and attempts at stat translations between the NHL and other leagues. Below is a list of other sites he showed either that morning or in the afternoon, with notable information:

• war-on-ice.com has the best player usage charts in the business and allows for home/away splits
  
• Ninja Greg Sinclair (somekindofninja.com) has a super shot search tool
  
• behindthenet.ca is not the most user-friendly, but is pretty much the original advanced stats site. If you use the address bar on your browser, you can see there are up to 67 categories available, and you can type in your own URL to view some of the other stats that aren't up by default
• puckalytics.com, the new site for the author of stats.hockeyanalysis.com, has a large glossary, as well as lots of categories of WOWY
• progressivehockey.com shows player usage trendlines, team-adjusted zone starts, confidence intervals for save percentage, and data from the AHL
• hockey-reference.com has player index tools, that make it easy to sort players by certain stats (ex. players under the age of 22 with at least as many goals as Jarome Iginla by his age 22 season)
• fenwick-stats.com shows data over a specified date range

There were also shout-outs to S7S writers, in that B_T's naturalstattrick.com was mentioned as among the best sites for game-by-game data, and Sareon's CHLStats.com shows a bunch of the hard-to-get stats from junior.

Lastly for the long list of websites, Bob Sullivan from SportingCharts.com showed some of the features of his site. Important features included shot and goal location comparison charts for players or teams, game previews showing how teams match up in a tonne of categories, and lots of available graphs.

The first speaker of the morning was Andrew Thomas of war-on-ice.com, calling in remotely from Pittsburgh. He ran through a bunch of the features of the site (there are many), and explained a little bit about the site. For example, the stats run back to 2002. However, before 2005, nobody recorded missed or blocked shots, so only data relating to shots on goal is available. He also said that the pinnacle in hockey nerdiness is finding humour in the question "Who would win in a fight between Kirk Maltby and Alexandre Picard?"

Justin Azevedo of Flames Nation talked a little bit about his quest to chronicle microstats for all the Calgary Flames games. He charted over 60 last year, and was five games into this year as of Saturday. He had some interesting patterns he'd noticed, such as how the stretch pass failed to connect 76% of the time last year. He commented that the average successful zone entry took an average of 1.9 passes from their own defensive zone, which seems to run against the common theory of the one-pass breakout. He also commented on how certain players like Brian McGrattan were just terrible at getting off the ice in a reasonable amount of time. The second-period long change definitely made line changes more difficult, and he was interested to see if adding the long change to overtime was going to have any effect.

One of the limitations of this type of study is that it's focused on one team. People studying other teams in as much detail would need to validate some of his findings. The obvious limitation is time, as it takes at least three hours to do this kind of detailed study. He also had an interesting sidenote in that in doing this study, he's found himself feeling detached from his team. In breaking down every play they make, it's become harder to simply enjoy hockey.

Then some Senators fan who goes by a dumb pseudonym like Absolute Insanity led a discussion on the analytics of faceoffs. People have found statistical significance in using faceoff wins contributing to Corsi, in some cases even more than zone starts. A faceoff differential of 76.5 is about equivalent to a goal differential of 1, with offensive/defensive zone draws being more important than neutral zone draws, and special teams wins being more important the even strength. This compares to only 50-70 faceoffs in an average hockey game. A team increase of 10% in faceoff win percentage is equivalent to about two more wins, but last year the Nashville Predators led the league with a faceoff win percentage of 53.3%. Though there may be statistical significance, it's hard to see a practical significance for something that has such a minimal effect.

Jamie Cochrane, a Blackhawks fan working in accounting for the Flames, led a discussion on goaltending stats. He argued that our best current stat is save percentage, which isn't really great, since it doesn't account for quality and frequency of shots faced very well. Arguments arose such as the fact that after 2500-3000 shots faced, shot quality isn't a huge factor, and the idea of home plate save percentage, which appears to be more telling than ordinary save percentage. Other points were brought up, such as how shootout save percentage tends to be repeatable but not correlatable to regular-play save percentage, and how the confidence intervals in save percentage found on Progressive Hockey are very large. There also appears to be very little correlation between save percentage in the AHL and NHL, Pekka Rinne being the most famous example. One person also pointed out that all of this talk is only useful if we consider goalies to be consistent.

The day was wrapped up by William Loewen, of the site checkrepublics.wordpress.com, who talked a bit about parity. He first pointed out that parity is a very vague word, giving three example of what it could mean for the NHL: each team has an equal budget with which to work; each team has equal chance to win a championship; and each team has a 50% chance of winning each game. He also pointed out the flaws of a complete-parity system, in that a Stanley Cup win is less significant if it happens for every team in a 30-year rotation. In comparison to a binomial distribution, he found that there are far more teams just below the mean section of the distribution than would be expected statistically, suggesting that loser point is currently creating what he coined "artificial parity".

(If anyone knows of anyone who is also studying parity, William would love to hear from you. You can contact him on his website above, or through Twitter @checkrepublics. No one at the conference knew of anyone studying it.)

All in all, it was a good day. I heard lots of interesting topics, and got to talk hockey with a bunch of people I'd never met before. I learned lots, and also learned of many online sources from which I'll learn lots in the coming weeks. The next conference will likely be in Edmonton, and will be advertised on HockeyAbstract.com. I'm sure carpool opportunities from Calgary will exist too. I'd recommend for anyone who gets the chance to check one of these out soon. Maybe you'll even meet me!