#4. Roster Size

At first glance, our fourth and final adjustment may stand out as unique.  Contextualizing for the goal and assist environments is logical, while remedying inconsistent season lengths is clearly necessary.  But the concept of why a team’s roster size should matter in normalizing counting stats is a legitimate question.

Seven-time Lady Byng winner and Hall of Famer Frank Boucher debuted in 1921 on a maximum 9-man roster and retired on a team that could dress 14 skaters. "TOPPS NHL Playing Cards 1960-61 — All-Time Greats 17" by dugspr — Home for Good is licensed under CC BY-SA 2.0

Recognizing the NHL has only tracked the ice time of its players since the 1998-99 season, this would leave 81 seasons where no formal information exists on how much time a skater spent on the ice. Absent a real-life time machine courtesy of Doc Brown from Back to the Future, we will never know precisely how many minutes the superstars, run-of-the-mill players, or benchwarmers of bygone eras received. But it is not critical to our adjustment process that we account for every second a player is on the ice.  In modern times, if a player wins the Rocket Richard Trophy as the league leader in goals scored, he is not docked for averaging 30 more seconds per game than the runner-up. Players earn additional ice time for various reasons – their talent level, ability to play effectively in certain situations, quality of teammates, or simply the intuition or tactics of the coaching staff. 

The process of adjusted statistics, however, requires consideration for how an NHL roster has been shaped over the years. As we are trying to adjust for the fundamental differences in past seasons, this extends to the shifting number of skaters a team has been permitted to put in uniform. These differences acutely affect a player’s output in counting stats as the NHL’s mandated roster capacity inherently dictates the proportionate opportunity of the players.

Consider a required five skater unit of hockey players – three forwards, two defencemen. If restrictions limit a maximum of five players to dress for the game, each player will play a full 60 minutes. If the roster size is expanded to ten, regardless of the ability of the individual athletes, the available ice time will be split – not evenly, but undoubtedly split. If the same team has fifteen skaters at its disposal, each player will reasonably see the ice even less in order to ration resources and maximize the output of the group. This concept drives the fourth adjustment required for roster size.

The table below summarizes the the league’s regulations governing the maximum number of skaters (i.e., forwards and defencemen) an NHL team can dress in a game. Most of the significant movement and tweaks happened in the NHL’s formative years, eventually settling on 18 skaters in 1982-83, which serves as Hockey Reference’s baseline for roster size adjustment.

NHL MAXIMUM NUMBER OF SKATERS IN UNIFORM, 1917-18 TO 2020-21

Season(s)# of SeasonsMax Roster SizeRoster Factor
1918-19258None0.50
1926-19283110.61
19291120.67
1930-19323150.83
1933-19386140.77
1939-19424150.83
1943-19497140.77
1950-19512170.94
19521150.83
1953115.50.86
19541160.89
1955-19606170.94
1961-197111160.89
1972-198211170.94
1983-202138181.00
Roster Size = Skaters (i.e., forwards and defencemen)

A background on when and why the NHL has changed its maximum roster size so frequently may offer a deeper understanding of the league’s evolution. Though information on the seasonal tweaks to rosters is difficult to come by, the league’s history of gradual roster growth is logical. As we’ve learned, the earliest days of the NHL were fraught with uncertainty.  In the late 1910s, professional sports in North America continued to face uphill battles to establish viability as legitimate businesses, and as legitimate careers for its most talented players. Competition from professional circuits in Western Canada, World War I and its aftermath, social perceptions of the nobility of retaining one’s amateur status, cultural aversion of traditional families dedicating time to sport, and the hesitation of young athletes playing a violent game long-term conspired to limit the available talent and interest in hockey. 

In its first eight seasons (1917-18 to 1924-25), there is no evidence the NHL imposed a maximum number of uniformed skaters on its franchises; however, a minimum eight players were needed. Given the primitive structure of the league at this time, financial restraint remained a reality; consequently, teams had a natural incentive to ice (and pay) as few players as was required to do the job. It was understood that with such a small roster, the top players would play the bulk of the game, as accounts from those early days noted star players seldom left the ice in critical games. Hockey-Reference uses nine skaters as a reasonable proxy for those seasons, though dressing only seven or eight skaters was not uncommon. Any instance of teams using less than the maximum at any point benefit its players in terms of production; however, the roster adjustment uses the rules in place to right size the situation as judiciously as possible. Perfect? Certainly not. But undoubtedly a fitting and necessary adjustment for the NHL’s formative years.

Ahead of the 1925-26 campaign, the NHL established its first known salary cap, fearing escalating player salaries would threaten profitability. The enacted $35,000(!) team payroll ceiling came with the first limit to personnel at 14 players, only 12 of which could take the ice in each match (11 skaters, one goaltender). For context, that same $35,000 is approximately $562,000 in today’s dollars, comfortably below the NHL minimum per player of $750,000. As the pace of play continued to speed up by way of alterations to forward passing rules in 1929-30, teams could dress a maximum of 15 skaters; further reinforcements were added to manage an increasingly faster and complex game.

Two-time Norris winner Erik Karlsson will be paid a league-high salary of $14.5 million by San Jose for the 2021-22 season - a far cry from the $35,000 team payroll of the mid-1920s. "Karlsson Shoots From Point" by clydeorama is licensed under CC PDM 1.0

Over the next two decades, 14 or 15 skaters in uniform became the standard, depending on the season. In fact, the first seven seasons of the Original Six era featured only 14 skaters dressed – less than three full lines of reinforcements. As the NHL expanded from 60 to 70 games played in 1949-50, it permitted teams to dress 17 skaters for the first time. In the early 1950s, the roster rules were refined on multiple occasions: 16 skaters at home and 15 on the road in 1952-53 (hence the 15.5 listed in the table above); 16 for every game in 1953-54; 18 through December 1st and 16 thereafter in 1954-55 (averaged to 17 in the table above). Game accounts from this era note that standard shifts on the ice run several minutes long, an exercise in pacing oneself for short bursts of explosiveness – an outrageous concept in the present-day NHL.

The NHL would make only three further amendments to maximum roster sizes. In 1960-61, uniformed skaters dropped back down to 16 for eleven seasons, returning to 17 for the 1971-72 season. It remained common for teams to lean heavily on just three forward units and two pairs of defencemen; an elite defenceman could occupy the ice for more than the half the contest.  Finally, in 1982-83, the 18-skater formation in effect today arrives, which in nearly all cases is four three-man forward units, anchored by three defence pairings. With the pace of play perpetually rising, additional talent arriving from Europe, and a bevy of players arriving (or returning) from the WHA, further depth was needed to compete. For the next two to three decades, bottom lines and pairings continued to be sheltered and are often used primarily for physical play and intimidation.

An interesting development begins at the turn of the millennium, as pugilism declines swiftly. In 2018-19, you could expect roughly an average of only one fight per five games. One of the main causes of the downward trend is that a perpetual increase in game pace and the development of sports science leads teams to issue fewer minutes to its top players. Despite unprecedent athlete fitness, coaches must deploy depth players more frequently, eliminating the utility of a one-dimensional, low-minute bruiser. The phasing out of a Rob Ray-type, playing only 4:13 per night in 1999-00, is a side effect. From an adjustment perspective, while superstar minutes have come down slightly, NHL coaches have had the same number of skaters (18) at their disposal since the 1980s.

The chart below illustrates the maximum roster sizes over the course of the entirety of the NHL’s existence, with 18 skaters as the baseline for our fourth adjustment.

NHL MAXIMUM NUMBER OF SKATERS IN UNIFORM, 1917-18 TO 2020-21

Skaters = Forwards and defencemen

The above image provides a snapshot of the game’s roster construction. Picture the founding 1917-18 season, where a tiny roster allows Joe Malone to notch over two goals per game, rarely leaving the ice surface. By the 1940s, Rocket Richard counts himself amongst the Canadiens’ 14 nightly skaters, affording him significant ownership of the team’s supply of ice time. By the 1960s, Bobby Hull’s Black Hawks have more reinforcements in uniform, however the 16-skater limit keeps his minutes plentiful.  Next, we arrive in the 1980s and 18-skater lineup cards as a permanent fixture. Each of 1980s Mario Lemieux, turn-of-the-millennium Jaromir Jagr, and peak Sidney Crosby share the ice with 17 teammates. Our fourth adjustment accounts for this evolution in roster construction for each superstar’s era.


For the fourth and final time, we check in with Gordie Howe’s 1952-53 season. Seeking to neutralize Howe’s counting stats to a roster with 18 skaters, we use the Roster Factor in the table NHL Maximum Number of Skaters in Uniform, 1917-18 to 2020-21 created earlier on the page. The required multiplier from that 15.5-skater season is captured as follows:

  • 15.5 ÷ 18 = 0.86

Through our first three steps in the process, we amended Howe’s basic counting stats to now read 76 goals and 77 assists.

Taking Howe’s 76 goals and right sizing for an 18-skater era, we calculate:

  • 76 x 0.86 = 65 goals

Our final adjustment chisels Howe’s goal total down to 65.  Doing the same for his assist output, we arrive at a final quantity of 66 assists (i.e., 77 x 0.86).

Adjustment #4: Roster Size                  

Adjusts a player’s goals and assist totals to an 18-skater roster

In the eyes of Adjusted Hockey, Gordie Howe’s 1952-53 season measures: 65 adjusted goals (“AdG”), 66 adjusted assists (“AdA”), totaling 131 adjusted points (“AdPTS”). These results are a far cry from the 49 goals, 46 assists, and 95 points you’ll find in the NHL’s archives. The first three adjustments benefit Howe’s season and career (unfavourable goal environment, unfavourable assist environment, shorter season length) while the fourth reduces his output for the advantage of playing in a smaller lineup than the baseline 18 skaters.

In completing the walkthrough of this final adjustment, we’ve established the foundation supporting the concept of Hockey Reference’s adjusted statistics.  With this framework in place, we can now put goals, assists and points for every season and career on a level playing field. The results are incredible, providing new information demanding reconsideration of hockey’s historical record, and challenging long-held beliefs about the accomplishments of its 8,000-plus players.

Data from Hockey-Reference.com.