AI seminar: Qualification and elimination in the NHL using constraint programming

Speaker: Tyrel Russell

Sports fans in many sports anxiously watch their team's performances and their chances of winning a championship or securing a playoff spot. Typically, they obtain their information from major newspapers and websites, which publish standings along with remarks on the qualification and elimination of the individual teams. It has long been conjectured that these papers use a heuristic solution that fails to capture the interplay between the standings and the remaining schedule. In this talk, I present the progress that I have made on three qualification and elimination problems that arise in the National Hockey League (NHL). First, I address the problem of determining exactly when a team has qualified for the playoffs. I used a constraint programming solution to show qualification earlier than the results posted by the Globe and Mail. I also show that the different scoring models can have an effect on the date of qualification depending on the schedule. Second, the simple decision problem can be extended to examine the associated optimization problem. The optimization problem is the problem of determining the number of games needed to secure or guarantee a playoff spot. Third, with the results of the optimization problem, it is possible to calculate the likelihood of winning the number of games needed to guarantee a playoff spot. This requires counting or approximating the number of scenarios where a team wins the specified number of games. The results from this third problem could have interesting applications in sports management and sports analysis.