New York Rangers Top 60 Producers of Offense

I am counting down my list of the best producers of offense for the New York Rangers over their first 97 seasons with individual posts for each player, linked below. (Unfortunately, new posts are going to be few and far between during the school year. I’ll post when I can and hopefully do quite a few during school breaks.)

1.	22.	43.
2.	23.	44.
3.	24.	45.
4.	25.	46.
5.	26.	47. Mike Gartner
6.	27.	48. Derick Brassard
7.	28.	49. Bill Fairbairn
8.	29.	50. Edgar Laprade
9.	30.	51. Butch Keeling
10.	31.	52. Ryan Callahan
11.	32.	53. Babe Pratt
12.	33.	54. Wally Hergesheimer
13.	34.	55. Ron Duguay
14.	35.	56. Harry Howell
15.	36.	57. Reijo Ruotsalainen
16.	37.	58. Grant Warwick
17.	38.	59. Mark Pavelich
18.	39.	60. Pete Stemkowski
19.	40.	61. Murray Murdoch
20.	41.	65. Mike Rogers
21.	42.

How I Ranked the Players

The Short Answer

My rankings are based on the sum of:
1. playoff expectancy added through adjusted goals created above replacement (GCAR) during regular seasons
2. adjusted playoffs GCAR multiplied by .01

The Loooooooong Answer

Here are explanations of the two factors mentioned in the short answer:

1. Playoff Expectancy Added in Regular Seasons
   
   This is by far the bigger factor.
   
   I have come up with a method to attempt to properly rate the value of shorter-term, higher-peak production compared to longer-term, lower-peak production. A system that rates Player A who puts up 160 points in two seasons the same as Player B that puts up 160 points in four seasons is flawed because Player A has done more to push the team towards better records and more playoff appearances. In short, highly productive seasons are worth exponentially more than decent/pretty good seasons.
   
   Quantifying that difference is tricky, but here is how I have attempted it: I compared the adjusted team goals scored of teams that would or would not qualify for a spot in the postseason if 61% of all teams made the playoffs each year (which is the historical average). Adjusting all seasons to 82 games and an average of 250 goals scored per team, 98% of NHL teams have scored 191 or more goals in a given season, and I set that 191 goal level as a baseline, minimum starting point for a team. Teams that score 191 goals in a season have an approximate 4% chance of making the playoffs, and as goals for go up, playoff chances increase in a non-linear curve something like this:

2. Adjusted Goals Created Above Replacement in Playoffs

My goals created formula and replacement level used are explained below.

Playoffs GCAR for the top 60 players ranges from 0 to 24 and averages 7. For my rankings, I multiplied those raw numbers by .01063 for no reason other than after experimenting with different weightings, that felt like the best credit to give playoff production compared to regular seasons. Multiplying playoffs GCAR by .01063, players in my top 60 range from 0-.25 and average .07.

I’ll post a spreadsheet here showing the playoff expectancy added and playoff GCAR numbers after I post my top 60 players.

---

Adjusted Stats Used

I have made my own adjustments to actual games played, goals, and assists throughout NHL history. The easy aspect of adjusting stats is setting a consistent number of team games played per season, and my adjustments force every year to an 82 game schedule. The extremely challenging aspect is determining how to adjust for league-wide levels of scoring. This might at first glance seem as simple as forcing yearly goals-per-game to a constant, but that approach breaks down at the individual player level due to expanding roster sizes and the evolution from the early NHL days of relying on the starting five to skate the vast majority of games to today’s game where 18 skaters usually play significant minutes every contest.

There has been at least one consistent strand in this regard throughout NHL history: there has always been the concept of teams having a top forward line and a top pair of defensemen. With that in mind, I have identified “first-line forwards” and “top pair d-men” from every season. These players are defined as a certain number of the top point-scoring players every year. The number of such forwards each year is equal to teams in the league multiplied by three, and the number of such defensemen each year is equal to teams in the league multiplied by two. In 1926-27, there were 10 NHL teams, so the top 30 point scoring forwards are the “first-liners” and the top 20 scoring d-men are “top-pair.” Today, with 32 teams, these groups include the top 96 point-scoring forwards and top 64 point-scoring defensemen. Taking all NHL “first-liner” seasons into account, the average first-line forward plays 96% of their teams’ games, scores .39 goals per game, and tallies .54 assists per game (which equals 79 games played, 31 goals, and 43 assists in an 82 game schedule). The average “top-pair” d-man plays 93% of their teams’ games, scores .13 goals per game, and adds .42 assists per game (76 GP, 10 goals, and 32 assists).

The approach I have taken to attempt to level out the scoring environment throughout league history is this: Adjustments are made to all individual forward player seasons by multiplying their actual goals and assists per game by whatever factors are needed to force each year’s “first-line forwards” as a group to score .39 goals and .54 assists per game. All d-mens’ yearly goals and assists per game are multiplied by whatever factors are needed to force that year’s “top-pair d-men” as a group to score .13 goals and .42 assists per game. There is no one correct or perfect way to approach adjusting stats, and there are plenty of issues with this method, but I do consider it to yield more interesting and meaningful results than any other method I am aware of.

Goals Created

To these home-cooked goal and assist totals I apply a home-cooked “goals created” formula. There is no consensus about how much a goal is worth relative to an assist, and reasonable arguments could be made about different weightings of the two. I feel sure that on average the goal scorer contributes more to a goal than the zero, one, or two players credited with an assist. (There are many exceptions when those setting up a goal contribute more than the goal scorer.) For reasons that would take too long to explain, the goals created formula I am currently using is .476*goals + .318*assists. (If you are curious about how I landed on those numbers, feel free to shoot me an email.)

Goals Created Above Replacement

To determine GCAR, I have identified “replacement level” in a manner similar to how I defined “first-line” and “top-pair”: For each season, I separately sort forwards and defensemen by games played, and all players below a certain ranking are considered “replacement” players. The number of players above replacement level each year is equal to the number of active roster spots multiplied by teams in the league. For example, in today’s NHL, there are 32 teams that typically dress 12 forwards per game, so the top 384 (32 x 12) forwards by games played are above replacement and all remaining forwards are considered replacement level, and the top 192 (32 x 6) d-men by games played are above replacement and remaining d-men are replacement level. Using my adjusted stats for this group of players, replacement level was set at .12 G/GP and .16 A/GP for forwards and .04 G/GP and .15 A/GP for d-men. (Replacement level shifts down very slightly in the playoffs to account for the slightly lower level of offense in the postseason.)

For every player in my top 60 I will post their adjusted Rangers stats which will look something like this:

Here are explanations for all of those stat categories: