This is the first part of many posts that is part of a larger project. The ultimate goal of this project is to develop a statistical model that will allow me to determine the probability of wins and losses for NHL teams. I am hesitant to say that I will be able to predict who will win or lose, but rather construct a model for success (i.e. if certain conditions are met, a team is more likely to win). As hockey fans, we recognize certain events during games that alter the momentum of one team. The model that I am developing is an attempt to capture important aspects of the game, such as shooting the puck or winning faceoffs. The data is sourced from war-on-ice.com, and has been organized per team by game spanning from 2005-2014 (totaling over 25,000 rows). As of right now, I expect to examine several game situations including but not limited to: even strength, power play, shorthanded and 4v4.
The first order of business is examining even strength goals. I think that everyone can agree that the absence of 5v5 scoring diminishes the ability of a team to be competitive. Along with goals against, scoring goals is one of the most important events that occur during a hockey game. Recently, the most common explanation for goal scoring is puck possession. It has been theorized that teams with high puck possession rankings tend to be more successful. This seems intuitively true, since the team who possesses the puck has more chances at scoring than the team without the puck. Since there is only one puck on the ice, it is a zero-sum game: whoever doesn’t have the puck cannot score. The two most common measurements of puck possession are Fenwick and Corsi. In simple terms, Corsi measures all shots taken, whether they miss the net, blocked, or on net. Fenwick also measures shots taken, with the exception of those blocked.
In other words:
Corsi = All shots directed toward the net
Fenwick = All shots directed toward the net – Blocked shots
Continue reading “NHL Scoring Model Part 1: Even Strength Goals”
Even though the NHL season is in the midst of the Stanley Cup Playoffs, it’s not too early to think about the off-season festivities of the NHL draft and free agency. One of the unique features of the NHL is that it is the most cap constrictive sport. This is in part because of three features: (1) the presence of a hard salary cap, (2) the nature of guaranteed contracts and (3) limited and no-trade clauses. The combination of these three factors separate the NHL from other leagues, namely the MLB, NFL and NBA. Other leagues have mechanisms for restructuring contracts, the lack of guaranteed money and/or soft salary cap restrictions.
How this translates to differentiate the NHL from other leagues is that the salary cap is the great equalizer, which can empower or impoverish a franchise with a few decisions. Aside from the compliance buyouts offered in the wake of the previous CBA, there are limited options for a team to jettison a player signed to a bad contract. Teams that sign players to long-term, big money contracts are held most accountable in the NHL, with limited options of movement that can hamstring a franchise for years.
Thanks to war-on-ice.com, I was able to analyze the relationship between Corsi values (shots taken for/against while on the ice/off the ice) and annual contract values for players. First, I map the Corsi for (herein CF) and Corsi against (herein CA) values on a (x,y) plot. The size of the bubbles (for each player) are relative to the value of their contract.
Continue reading “Contracts and Corsi”
Another NHL preseason has been completed and the regular season is set to start tonight. Upon looking over the point standings in the preseason, it is shocking to see teams like Toronto and Columbus leading their respective divisions (albeit they played more games than other teams within their divisions). This led me to search for a correlation between preseason and regular season success. After mulling through the myriad amount of information available for other sports, namely the NFL, it led me to create my own. Once collecting the data, I was able to run some common statistical tests using STATA. Unfortunately, the relationship was extremely weak. I produced a graph to illustrate the results below. In order to account for the difference in the number of games played in the preseason, I measure success using a points per game variable. This can be compared to the points per game each team earns in the regular season.
Continue reading “Does the NHL Preseason Matter?”