I once saw a stray comment on an APBA Hockey post that defensemen get too many assists. I had never really noticed that until I started looking closer at some of the statistics in my own projects.

One project I play every year is a 10-game, 12-team draft league. Based on the season I last played (2019-20), and adjusting for the assist numbers would be a little higher because a smaller league produces more scoring (25% more goals), pro-rating the assist leaders would produce the following:

Player	Pos	Exp.
Draisatl	F	12
McDavid	F	12
Panarin	F	11
Carlson	D	11
Marchand	F	11
MacKinnon	F	11
Huberdeau	F	10
Kucherov	F	10

However, when the season was done, it was littered with defensemen:

Player	Pos	Exp.
Carlson	D	17
Makar	D	10
Josi	D	10
S. Jones	D	10
Pietrangelo	D	9
4 tied (2 D, 2 F)	F	8

Yikes. Doesn’t look like a typical scoring leader board. With seven seasons in this project, a defensemen has led in assists five times. So there is probably some tweaking to be done.

At first, I wanted to compare real life stats to the actual season (2019-20). The idea was to grab one team’s scoring, figure out who was on the ice for each goal, and simulate awarding assists. However, I was thwarted by this as the NHL does not make the individual +/- for a goal available in their public API anymore (an API is a method of retrieving data without going web page by web page.)

So I still had some data leftover from some of my other innovation investigations from 2012-13. I picked the most average team available (Columbus), and ran each goal through the current assist rules 10,000 times in order to get an average of each goal.

In that season, Columbus had 205 assists — 64 from the D and 141 from the forwards — for a percentage of 31% for the D. In my first run of the normal rules, I only had on average 185.4 assists per season, as Columbus was a bit above average in receiving assists. That being said, the average simulation saw more D assists (68.1) and a lot less from the forwards (117.3), for a D percentage of 37%. So maybe some tweaking needed to take place.

As a lark, for the second trial, I went back to the original 1.0 game rules, which was simply check from High-to-Low for all assists, rather than rolling for an initial position first. Although the D percentage was better (35%), players with higher assist ratings were wildly higher and players with lower assist ratings were significantly lower.

So for the third test, I tweaked the rules that instead of a one die roll to determine the first position, I went with two (LW: 11-22, C: 23-34, RW: 35-46, LD: 51-53, RD: 54-56, Rank: 61-66.) With this setup, the D percentage was 32% and everyone was pretty proportional to their actual number.

The fourth test was to see what would happen if we went a little closer to the current rules (LW: 11-21, C: 22-32, RW: 33-43, LD: 44-52, RD: 53-61, Rank: 62-66.) This ended up being a little worse with the D percentage was 34%, although the proportions were similar. With that being equal, I would still lean towards the results in the third test.

However, when typing up the new chart, I realized the positions within forwards and within defensemen weren’t equal. So after six more trials, and sparing you the details of those tests, I finally came up with a chart I’m going to use in my next season, one that had a 31% D assist percentage and better proportionality than the original winning test:

Dice Roll	Assist Order
11-22	LW-C-RW-LD-RD-G
23-26	C-RW-LW-LD-RD-G
31-34	C-RW-LW-RD-LD-G
35-46	RW-LW-C-RD-LD-G
51-54	LD-LW-RD-C-RW-G
55-56	LD-C-RD-RW-LW-G
61-62	RD-C-LD-RW-LW-G
63-66	RD-RW-LD-LW-C-G

So this begs the question, if all things are equal, how come the defensemen get more assists. The reason is that forwards score goals more often, and therefore are removed from the check, giving defensemen more of a chance to get the assist. By tweaking the chances for each position, and giving forwards more of a chance, the numbers work out much better.