In a continuation of my recent Hall of Fame Probability model for NHL forwards, I have now built a model to measure which NHL defensemen would make the Hall of Fame if they were to retire today. Using statistics and a logistic regression model, I have attempted to calculate each current NHL defensemen’s probability of receiving one of hockey’s greatest achievements; an induction into the Hockey Hall of Fame.
Similar to my probability model for forwards, I based this model on the Basketball Reference Hall of Fame Probability Model, using logistic regression to estimate each player’s chances of being inducted into the Hall of Fame. Contrary to the first rendition, this model only considers defensemen, as players are judged differently by the Hall of Fame committee based on their position.
I decided to make a sequel to my original article as I am intrigued by which factors are most important to a defenseman’s career, and by which current defensemen have the highest probability of making the NHL Hall of Fame. Using these two models, it is now easy to see which players are likely to make the Hall of Fame and which players are likely to fall short.
Methods
Similar to my most recent model, I used logistic regression to determine which NHL defensemen have the highest likelihood of making it into the Hockey Hall of Fame. A player is eligible for the HHOF if they have been retired from professional hockey for the previous three seasons. Players are judged based on their playing ability, sportsmanship, character, and contributions to their team and the sport. As in my last article, quantifying a player’s sportsmanship and character is extremely difficult, so this model focuses on a player’s contributions and abilities to their team. In this model, I used four different statistics to measure a player’s dominance, success, and production. These metrics include:
Stanley Cup Wins: Stanley Cup Wins measures the number of Stanley Cups a player has won throughout their career. Cups are often used to measure a player’s success in the NHL, and can greatly influence the legacy of a player after they have retired.
Era-Adjusted Points per Game (EAPPG): Era-Adjusted Points per Game measures allows the model to compare players’ point production across different eras. EAPPG is necessary to this study, as the amount of goals scored in the league fluctuates over time. For example, in the 1981-1982 season, there were 4.01 goals per game, compared to the 2019-2020 season, where there were just 3.02 goals per game. Thus, a player with one point per game in the 1980s would not be as valuable as a player with one point per game in 2020.
Defensive Point Shares (DPS): Defensive Point Shares quantifies the amount of standing points a player contributes defensively to their team. DPS allows us to measure how a defenseman contributes to their team on the backend and how effective they are at keeping the puck out of their net. Zdeno Chara has the best DPS amongst active players (99.0). A breakdown of how DPS is calculated can be found here.
NHL All-Star Team Designations: There are two NHL All-Star teams decided at the end of each season which honour the best players at every position, meaning that there are four All-Star defensemen each season, which helps us measure the dominance of a player throughout their career.
To gather my training data, I used Hockey Reference to pool together all defensemen with more than 250 games who are eligible or are in the Hall of Fame. Next, I fit a logistic regression model based off of these defensemen and applied it to my training data; all ineligible defensemen for the Hall of Fame with more than 200 games played. In my previous model for NHL forwards, I used only players from the modern era (post-1967), however, because there are only 22 defensemen from the modern era in the Hall of Fame, all defensemen were used to generate a large enough sample size.
Next, I built a logistic regression model that assigned each of my variables with the following weights:
These weighting can then be used to calculate the Raw Probability Score (RPS) of a player, using the following equation:
RPS = Intercept + (Stanley Cups * 0.309561) + (Era-Adjusted Points per Game * 5.654432) + …
This means that for each Stanley Cup a player wins, their RPS will be increased by 0.309561. The intercept functions as a way to normalize the RPS when it is then used in the logistic distribution function.
As a reference, Zdeno Chara has the highest Hall of Fame Probability of all non-eligible players. His RPS is calculated below:
Next, to find out a player’s Hall of Fame Probability, we use the following equation:
Using Zdeno Chara, we get a Hall of Fame Probability of:
This means that based on the model, Zdeno Chara has 99.9999% chance of being inducted into the Hockey Hall of Fame if he were to retire today.
Results & Verdict
Below is the list of defensemen with the top 18 probabilities of being inducted into the Hall of Fame if they retired today (minimum 200 games played).
Unsurprisingly, the top of the list contains Zdeno Chara, Drew Doughty, Erik Karlsson, Shea Weber, and Duncan Keith. These players are essentially locks for the Hall of Fame when considering their overall careers. Each player has suited up for over 700 games and has all been named to the All-Star team at least four times in their careers. Additionally, these five have been extremely productive over their careers, with Chara possessing the lowest era-adjusted points per game (0.46). Further, Chara, Keith, Doughty, and Weber all rank 1st, 3rd, 4th, and 8th respectively in active DPS, while Karlsson has a godly 0.91 adjusted points per game, all contributing to their high Hall of Fame Probability.
In the next tier, we have Brent Burns, Alex Pietrangelo, Victor Hedman, and P.K. Subban, who have all had strong careers with high point production. All four players have an era-adjusted points per game over 0.60 and are within the top 32 for active DPS. Additionally, all four defensemen have been All-Stars at least three times over their career.
Kris Letang, John Carlson, and Mike Green make up the next tier in our list. These three are mostly known for their offensive ability in the NHL. Mike Green is probably the most surprising player in the top three tiers; however, it is hard to forget just how productive he was early in his career.
Photo credited to Getty Images
In a 3 season stretch from 2007-2010, Mike Green put up 205 points in 225 games and was a First Team All-Star in two of those years, contributing to his high Hall of Fame probability. Letang and Carlson have also been named to an All-Star team twice, while also providing tremendous offensive production in their careers. Additionally, both players have won a Stanley Cup, and Carlson recently put up back-to-back 70-point seasons.
One surprise in this list is that reigning Roman Josi Norris Trophy winner Roman Josi has only a 9.67% Hall of Fame Probability. While Josi just completed his ninth NHL season, he has only been productive for seven of them and was only an All-Star for the first time this year. If he can build off of his recent success, his Hall of Fame Probability can likely climb in the next few years.
Conclusion
The Hockey Hall of Fame considers a variety of different attributes regarding a player, such as international accolades, NHL careers, and legacies. Unfortunately, this model only looks at NHL success, which limits its viability for players that have had success outside of the NHL. Overall, the model seems relatively accurate in predicting which defensemen will make the Hall of Fame, however, only time will tell. In the future, I plan to continually update both my forward model and my defensemen model as seasons progress. Using this model, we can see what contributes to a player’s legacy and what factors of a player’s career will influence their chances of being inducted into the Hockey Hall of Fame.
Statistics retrieved from Basketball Reference, Hockey Reference, The Hockey Hall of Fame
Cover photo credited to David Berding — USA Today Sports