Being inducted into the Hall of Fame is one of the greatest honours a professional athlete can receive.
As an avid hockey fan, debating who does, and who doesn’t deserve to be in the Hall of Fame is one of my favourite pastimes. With this in mind, I wanted to create a model that takes various player statistics and uses them to calculate the probability of a player making the Hall of Fame if they were to retire today.
After various attempts, the model shown below is based on the Basketball Reference Hall of Fame Probability Model. So far, this model only considers players in the forward position, as the criteria for evaluating the careers of forwards for the Hall of Fame is different compared to defencemen or goalies. Specifically, there is a greater emphasis on offensive production for Hall of Fame careers than in other positions. In the end, the model appears to be able to accurately predict who gets the ‘Call to the Hall,’ and who will fall short.
While this model follows the method of Basketball Reference’s Hall of Fame Probability tool, there are other similar models that have calculated likely Hall of Fame inductions for hockey. One similar article is from Matt Pfeffer of the Hockey News, who bases a player’s Hall of Fame probability by assessing their overall careers, as well as their peaks specifically. RDJ of Oddacious posted a similar analysis that uses a logistic regression model. While both valid analyses, the criteria used to assess a player’s career is slightly different than the criteria I put forward. However, both models have similar results to my own, indicating that the model I have created may, in fact, be a good predictor for an NHL forward’s probability of making it to the Hall of Fame.
As stated previously, this model is based upon Basketball References Hall of Fame Probability Model. The model uses logistic regression to predict a player’s likelihood of making the Hall of Fame. The criteria for a player to be inducted into the Hall of Fame requires players to have been retired from professional hockey for at least three seasons. Additionally, players are judged based on their ability, sportsmanship, character and contributions to their team and the sport. Due to the difficulty in quantitatively measuring a player’s sportsmanship and character, I decided to focus on fitting my model around players’ abilities and contributions instead. Specifically, I used eight different statistics that can be used to measure a player’s production, dominance, success, and longevity. These include:
Goals: The classic statistic for measuring a player’s production. It’s simple: Goals win games. In fact, only five eligible players with more than 500 goals are not in the Hall of Fame.
Era-Adjusted Points: This counting stat measures a player’s points but adjusted for the time period in which they played. This allows a player’s production to be easily compared over multiple eras. For example, in the 1980s the average goals per game was much higher than it is today, meaning that it would be unfair to compare a forward who debuted in 80s with a forward who debuted in the 2010s. For example, Connor McDavid has a total of 469 career points, but 503 era-adjusted points.
Points per Game (PPG): PPG is a rate statistic that measures the average amount of points a player will score a game. This statistic was included in the model to account for a player’s consistency For reference, only 13 eligible players with a points per game of at least 1.00 are not in the Hall of Fame.
Point Shares: This is a Hockey Reference statistic that estimates the number of standing points that are contributed by a player. This metric allows us to measure how important a player was to their team. To demonstrate, Sidney Crosby has 154.0 total point shares over his 15 year NHL career. The formula for calculating point shares can be found here.
Stanley Cups: Measures the number of Stanley Cups a player has won in their career. Rings are often used as a defining aspect of a player’s legacy, and sometimes, a Stanley Cup can push a player out of the ‘Hall of Very Good’ and into the Hall of Fame.
Games Played: The amount of games played in one’s career helps to measure the longevity of a player’s career. Dave Andreychuk and Mark Recchi are two examples of this, as both players were never superstars, however both were great for a very long time, which eventually led to their inductions into the Hall.
Selke Trophy Wins: The Selke Trophy is awarded to the forward who demonstrates the most skill on the defensive side of the game. This accounts for players that contribute to their team’s success through their defensive play rather than offensive production.
Hart Trophy Wins: The Hart Trophy awarded to the most valuable player in the league. This helps us measure a player’s success at the top of the league, as only the most dominant forwards will ever win a Hart Trophy. In the modern era, every eligible forward that has won the Hart trophy has been inducted into the Hall of Fame.
Using these statistics, I pulled data from Hockey Reference for all forwards who have played more than 600 NHL games who are also eligible for the Hall of Fame in the modern era (post-1967). I limited the model strictly to the modern era due to the increase in competition and talent that has occurred post-expansion. Further, the model only uses players who played at least 600 games, as all Hall of Famers who played the majority of their career in the NHL achieved this milestone.
From there, I built a logistic regression model that calculated and assigned each variable the following weightings:
These weightings contribute to what we will call the raw probability score (RPS). This score is calculated by taking the equation:
RPS = Intercept + (Goals*0.008109) + (Points per Game*11.4046) + …
The interpretation of these weightings means that each goal a player scores will increase their RPS by 0.008109 and each Stanley Cup won will increase their RPS by 1.078. The intercept is used to normalize the score for when it is inputted into the logistic distribution function.
For reference, Jaromir Jagr, who has the highest Hall of Fame Probability of non-eligible players, has an RPS of 13.90522. An example of his calculation is shown here:
To find our Hall of Fame Probability we use the following equation:
In the case of Jagr, we will get:
Meaning that based on the model, Jaromir Jagr has a 99.9999% probability of being inducted into the Hall of Fame.
Below is a list of the forwards (minimum 200 games played) with the top 50 probabilities of being inducted into the Hall of Fame if they retired today as well as other notable forward probabilities.
Other Notable Hall of Fame Probabilities:
As expected, players such as Jaromir Jagr, Alex Ovechkin, Sidney Crosby, Evgeni Malkin, and Patrick Kane are basically guaranteed locks when factoring in their production and accolades. What I also found surprising at face value is that players such as Marian Hossa and Patrik Elias are assigned high Hall of Fame probabilities. However, both players have eclipsed 1000 games played and 1000 points while also having won multiple cups. While these two may not necessarily be first ballot hall of famers, given their stats, it is not hard to see the two Czechs making the Hall of Fame.
Even more surprising, though, is that Connor McDavid already has a Hall of Fame probability of 95.8% and is by far the youngest player on this top 50 list. McDavid currently has the highest points per game among active players (minimum 50 games) and has already won a Hart trophy in his 5-year career. For reference, after five seasons, Sidney Crosby and Alex Ovechkin posted a Hall of Fame probability of 99.1% and 99.8% respectively.
As a Maple Leafs fan, I was interested to see where Auston Matthews stood amongst current players. Through his first four seasons, Matthews is ranked 58th among active and non-eligible players with a 4.76% probability of making the Hall of Fame if he were to retire today. Patrick Kane, another American superstar had a Hall of Fame probability of 4.96% through his first four seasons, meaning that given a strong continuation to his career, Matthews may certainly enter the Hall of Fame conversation one day.
The Verdict: Ranking which players will make the Hall of Fame
These players have defined the last generation of the NHL. If these players retired today (if they haven’t already), they will undoubtedly be first ballot Hall of Famers. Every single player on this list has been dominant for the majority of their careers and have many accolades to go along with their dominant stats.
The Probable Hall of Famers:
If these players retired today, they are likely be Hall of Famers eventually, but not necessarily in their first year of eligibility. Their presence has been notable throughout the past decade and have contributed to the sport through their exceptional play. These players were always some of the best in the game, but never included in the same conversation as the likes of Crosby and Ovechkin. It should also be noted that Ilya Kovalchuk has a deflated probability due to the time he took off to play in Russia as well as the numerous international accolades that he has achieved.
A Few More Good Seasons:
With just a few more season at their current pace, these players are sure to be locks for the Hall of Fame. If these players were to retire today, they would join the Probable Hall of Fame class.
The Hall of Very Good:
If these players retired today, if they have not already, it is unlikely they would be inducted into the Hall of Fame. These players were good, and they were good for a long time, however, they were never dominant enough to be considered Hall of Famers.
The Could-be Hall of Famers:
These players still have a lot of hockey left to play in their careers and have shown flashes of greatness. If these players continue to play well for the rest of their careers, they will have a strong case to get the Call.
Conclusions and Limitations
This model offers insight into which players are on track to become Hall of Famers. One issue with the data is that it only considers NHL statistics, however, the voting committee also considers international accolades, such as the Olympics and IIHF honours. For players such as Ilya Kovalchuk and Henrik Zetterberg, who both had incredible international success along with their NHL careers, this model may underestimate their true probability of making the Hall of Fame. We must also mention that while intangibles are not exactly loved in the advanced statistic community, these factors do have an influential role in a player’s legacy. For example, the perception of a player’s leadership, popularity, and “grit” may not contribute much when it comes to a player’s on-ice effectiveness. However, these categories are important factors for the Hall of Fame induction committee.
Going forward, I plan to update this model as new classes are inducted, and players progress in their careers. It would also be an interesting experiment to create similar models for the careers of the many defensemen and goalies that belong in the Hall of Fame. With this probability model, we can see how different factors contribute to the legacy of a player’s career, and with a fine combination of statistics and awards, we can see how that translates to a Call to the Hall.