Paramjit Gill applied statistical analysis to the NHL’s hockey numbers before and after rule changes were introduced - photo ©iStockphoto/walik
UBC Reports | Vol. 53 | No. 6 | Jun 7, 2007
Pucks, Prose and Probabilities
From the post-lockout NHL to the texts of Alfred the Great, stats professor finds the stories lurking deep in the data
By Bud Mortenson
“He shoots, he scores,” rang out more often in the National Hockey League over the past two seasons. So, too, did the referee’s whistle.
More goals, more excitement. That was the vision when the NHL overhauled its rulebook two years ago, in part to rekindle fan interest following a 2004-2005 season lost to a player lockout. The game changed as expected, with teams averaging 1.92 goals in a home game before the new rules, and 2.12 goals after. But the game also changed in some unexpected ways, says Paramjit Gill, an associate professor of statistics at UBC Okanagan.
“The league promised zero tolerance for hooking, holding, tripping, slashing, cross-checking and interference,” says Gill. “This resulted in more penalties being awarded -- more than 14,000 in the 2005-2006 season, in comparison to about 10,000 in previous seasons.”
Gill and student Stephen Welsh have applied statistical modeling tools to NHL regular seasons 2003-2004 and 2005-2006 (before and after the lockout year), examining even-strength goals, power play goals and power play opportunities.
“The home team’s ability to score during an even-strength play increased during 2005-2006 as compared to the game under old rules,” Gill says, “but there wasn’t any change in the home-ice advantage during the power play.”
Power play scoring ability became more important under the new rules. On the other hand, penalty killing ability appeared to have less impact on a team’s standings.
This summer, they’re analyzing the NHL’s 2006-2007 regular season. Already, they’ve found that while the number of penalties assessed this season fell from 14,000 to about 12,000, the chances of scoring during a power play are the same as the previous year.
“The rule changes introduced by the NHL were designed to open up the game, increase scoring and present a more entertaining product,” Welsh observes. “No lead is supposed to be safe anymore. However, preliminary analysis shows that the winning percentage of teams trailing entering the third period has actually decreased in the post-lockout years -- teams are actually finding it more difficult to come from behind and win.”
Looking at sports is a great teaching tool, says Gill. “There’s no doubt the subject -- statistics -- is hard,” he says. “The advantage of using sports is that it’s much easier to get across to students.”
Gill’s primary research is a long way from the sports field, though. Projects include studying air quality and asthma in populations, and the analysis of rare-event phenomena. Over the years he has examined accident rates, teenage pregnancies, drug prescriptions by region, and the codling moth in Okanagan orchards -- always looking for patterns emerging from the data.
“To me, that makes the work very exciting,” says Gill. “I see the similarities in these things. It’s very fulfilling to see through the lens of statistical modeling.”
That lens has even allowed him to look a thousand years into the past. Collaborating with Michael Treschow, Assoc. Professor of English at UBC Okanagan, last year Gill applied stylometry -- statistically measuring word usage -- to 9th-century religious texts believed translated from Latin to Old English by King Alfred.
“Each writer has his or her own wordprint -- just like a fingerprint,” Gill explains. “Non-contextual or ‘function’ words -- and, or, it, whether -- have nothing to do with what you’re writing about, but by counting those, you can distinguish between authors.”
Three translations that had long been attributed to Alfred did, indeed, cluster together on the frequency of function words. However, says Gill, one other translation attributed to Alfred, The First Fifty Prose Psalms, was found not to be an Alfredian text – a conclusion that challenges some authoritative scholars of Medieval languages.
Gill’s application of statistical tools to ancient texts -- and to the NHL -- represents a new depth of analysis that takes advantage of today’s readily accessible data, he says. A decade ago, he painstakingly collected hockey stats from the daily newspapers. Today, an entire season’s data can be obtained almost instantly from the Internet.