Tuesday, 27 December 2011

Point Allocation on Offence

Previously we've talked about Marginal Goals, which form the basis of the historical Point Allocation system. We use Marginal Goals not only to establish a team's quality, but also to allocate this quality (in the form of points) between offence and defence. From there you merely have to allocate these offensive and defensive points to each player on the team, in a rational way, to estimate the value that each of said players contributed to the team achieving what it did.

I should say, at the outset, that Point Allocation will never by anything more than a rough estimate of a player's value, especially the further back you go in history. The surprises it uncovers will be few; those considered the best players of the day will largely be shown by the system to be great players. It's really more about trying to establish an equal footing to allow players from different eras to be compared to each other, and also to identify the value of the non-great players from any era. There are hundreds of players from history that were really good, solid players, with the level of talent that would allow a nice, long NHL career in today's world, but who are relatively unknown today because they were not among the very best players of their day. Part of the purpose of the system is to figure out approximately how good these players were.

Going back to the system, we can look at allocating offensive points, which is easier and more straight-forward than dealing with defence. Point Allocation gives offensive points to players based on their scoring point totals, as a proportion of the team's total scoring points. We don't worry about adjusting for the relative rate of assists per goal; rather we assume that the officials of the day basically knew the relative contributions of playmakers and goals-scorers at the time. There were many fewer assists awarded per goal in the pre-forward pass era, for example, and individual rushing was a more important factor in scoring goals at that time, so we figure that's alright.

We can have a look at the 1907 Montreal Wanderers again, one of the greatest teams of all time, to examine some of the details.

Russell, Ernie C 9 43 4 47 26 514
Johnson, Moose LW 10 15 5 20 42 558
Blachford, Cecil RW 7 14 3 17 5 415
Glass, Pud R 10 13 4 17 21 579
Patrick, Lester P/R 9 11 3 14 21 532
Marshall, Jack RW 3 6 0 6 0 180
Stuart, Hod CP 8 3 2 5 20 460
Kennedy, Rod P 3 0 1 1 2 178
Totals 10 105 22 127 124 610

Wow - this team had five Hall-of-Famers among its eight skaters (Russell, Johnson, Patrick, Marshall and Stuart), and another tending goal in Riley Hern. It was truly an all-star team that played together throughout the season.

(As an aside, you might realize that the assist totals above are completely unofficial. I compiled them by reading the game reports for all of the ECAHA games that season. I'll have a brief discussion of compiling assists like this in the next post.)

Based on the above, Ernie Russell, one of history's best pure scorers, would receive 47 out of 127 offensive points the Wanderers earned, or 37%. But it's not quite that simple. Some players produce points at such a low rate, compared to their positional norms, that they effectively costs their team goals. Remember that Point Allocation is built upon the concept of Marginal Goals, which has at its core the idea that there is a certain minimum level of performance that anyone good enough to get a sniff in a particular league will be able to contribute. This applies to individual offensive performance as well as at the team level.

For the Wanderers, this applies to only one player. Pud Glass was a hard-working, professional player and a very good defender. But in 1907 his offensive production left much to be desired, given that he played rover, which as a position scored just as much as centres league-wide. Among all centres and rovers, he places ahead of only the men from the sad-sack Montreal Shamrocks on a per-minute basis:

PlayerPosTeamGPPtsMINPts/60 MIN
Russell, Ernie C Mtl W 9 47 514 5.49
Bowie, Russell R Mtl V 10 47 587 4.80
Smith, Harry C Ott 9 25 484 3.10
Hale, Chandler C Mtl V 7 21 412 3.06
Jordan, Herb C Que 5 15 300 3.00
Sargent, Grover C Mtl A 9 22 538 2.45
Constantin, Charles R Que 7 16 409 2.35
Westwick, Rat R Ott 9 18 528 2.05
Smaill, Walter R Mtl A 10 18 578 1.87
Glass, Pud R Mtl W 10 17 579 1.76
Brennan, Johnny R Mtl S 5 7 285 1.47
McCarthy, Frank C Mtl S 5 4 225 1.07

The system figures if Glass can't outscore Charles Constantin, who played precisely one season of senior hockey, his offence should be knocked down a peg when deciding who contributed how much to his team's goal-scoring. So he's given 10.7 scoring points instead of 17 when allocating the team's offence. Rather than just adjust everyone's point totals to remove this "sub-marginal" level of offense, we just apply an adjustment to those below that level. Why this way and not the other? No particular reason. The effect is the same.

And don't worry. We know that Glass has a reputation for good defence, and he will earn his share of points that way, when we look at allocating defence. In fact, the system considers him the team's most valuable defensive forward. But more on that later.


  1. Where did you get minutes played? I didn't know that was available - is it an estimate?

  2. The beauty of this era is that you can determine minutes played with a high degree of accuracy. Substitutions were fairly, and generally the game reports will mention when the sub was made. Mostly players played the full game, except when they were serving a penalty.


Hostgator promo codes