Hello again everyone,
Recently I’ve put up a few fanposts about the relationships between puck possession and save percentage on playoff success. Using some statistical tools, we could see that there was a weak but significant link between regular season Fenwick Close and playoff performance, and could generate a regression formula to predict an ‘expected’ number of playoff wins based on RS Fenwick Close.
We then saw that the relative over/underperformance of each team’s goaltender explained a pretty significant amount of the discrepancy between Fenwick-expected wins and actual playoff wins, reinforcing the notion that goaltending carries a whole lot of weight during the postseason. However, there was still a great deal of variability left to explain that neither puck possession or goalie stats could explain.
Well, the wait is over. The holy grail of hockey prediction - the One Stat that Matters - may have been found.
Some background: back in 2011, Megalodon from Battle of Cali put up a highly scientific and serious post about a new metric for use in advanced hockey analytics: the SONA index. This metric, the ‘Sex Offenders Near Arena’ statistic, initially showed a high degree of predictive power by revealing that the team with the single highest SONA statistic in 2011 was named the Predators. Impressively, later that spring the SONA index successfully identified 5 of the top 6 teams during the regular season, outperforming many more established metrics such as goal differential and Fenwick Close. Unfortunately, while the SONA index showed phenomenal predictive power at determining who made the playoffs, it was unfortunately not so good at predicting actual playoff success. (Sound familiar? Wink) A new set of SONA predictions were made for the 2013 playoffs, but unfortunately they appear to have born little fruit as well. So, I decided to tackle the SONA index to see if any adjustments could be made, or to see if it might have more subtle predictive powers that have not yet been revealed.
The first thing I looked at was whether SONA could historically explain the set of recent playoff data I’ve been working with (the last 5 full playoff seasons). I used a slightly modified methodology from Megalodon’s original formula: rather than looking at registered sex offenders in the area immediately surrounding the arena, I considered the number of identified sex offenders within the city limits to eliminate the problem of arenas by the water (and get rid of the mer-vert confounding factor). Sex offenders were identified using the website Family Watchdog, and all searches were conducted by entering the zip code of the home arena and considering each sex offender that appeared in the initial results map. Unfortunately, only American teams were considered, due to the difficulty of determining this metric for Canadian cities. In order to normalize to population, and also add in another element of hilarity, I divided each city’s sex offender number by the number of minors in that city according to the 2012 US Census, which should give us a good ‘sex offender ratio’. These numbers were then multiplied by 1000 to form the ‘Adjusted SONA Statistic’ (A.S.S.).
I checked out the Adjusted SONA Statistic of each American team to make the playoffs since the 2007-08 season and compared it to 1) their total number of playoff wins, and 2) their playoff wins relative to Fenwick expectations (see this previous article for methodology there). Unfortunately, not much came out of that:
Clearly, Adjusted SONA is not a good statistic for predicting playoff success, nor playoff over/underachieving. Undeterred, I tried a number of different approaches but all of them turned up empty.
However, as I was about to give up on SONA as a metric for playoff prediction, a thought struck me: maybe sexual predation loses value in the playoffs because of something fundamentally different about postseason hockey: frustration. It might be that the collective predatory influence of playoff cities’ sex offenders is not adequately leveraged if teams have been recently successful in the playoffs, for instance. With that in mind, I considered a new metric: the Cup-Adjusted SONA Statistic.
This metric is the following:
[number of sex offenders in the immediate area surrounding arena] x [years team has gone without winning a Stanley Cup] / [number of minors living in city according to US Census].
I generated Cup-Adjusted SONA Statistics for the same data set of teams, and then separated each individual playoff appearance into Low, Medium, or High levels of hockey-related sexual frustration. Low was defined as a CASS of 50 or lower, Medium from 50 to 250, and High for 250 or greater. I then looked at number of playoff games won and playoff success relative to Fenwick Expectations for these three groups, and here is what came out:
Success! It appears as though we have a Goldilocks formula for postseason victory: your city cannot have too much nor too little Cup-Adjusted Predation; the middle group is just right. Groups with ‘the right amount’ (scoring between 50 and 250 on the CASS Index) not only won more playoff games, but also outperformed their expectations based on regular season puck possession. Interestingly, both the teams with low and high CASS index scores had trouble performing during the playoffs relative to the expectations placed upon them. I wonder what that could mean.
Finally, in order to visualize the CASS effect in a different way, I considered relative likelihood of low, medium, or high-frustration teams to perform well or poorly under playoff pressure:
Speaks for itself, doesn’t it.
Finally, as positive proof that the CASS index is for real, I offer you the top three hockey-related sexually frustrated teams of the past half-decade:
3) The Buffalo Sabers. Buffalo recently made a jump to the top of the 2013 SONA rankings, so it should be no surprise that they make this list. Plus, let’s be honest- basically everything about the Sabers confirms a top-3 spot.
2) The Washington Capitals. A surprise entry this high on the list, Washington nonetheless displays several important characteristics of a high-ranking CASS team- an excruciating run of playoff disappointments, an abundance of sexual predators in local government, and Alex Semin.