How Do Adjusted Goals Work?

[ted2019]

I've been trying to understand all these advanced metrics for stats, but I really don't understand the adjusted goals stat. How does a player lose goals scored from real life? Example, Pat Lafontaine had 468 goals in His career, but it went down to only 414 goals according to His " adjusted stats" . How does He lose 54 goals that He actually scored?

Click to read more...

 

[Bear of Bad News]

All "adjusted" statistics work by normalizing between different eras (by choosing a benchmark era, and translating the goals (and other stats) from each era to that benchmark era).

So to directly answer your question - Lafontaine lost goals that "he actually scored" because, when normalizing to a consistent era, it was determined that goals had less value in Lafontaine's era than in the benchmark era.

These calculations do offend some people (generally those who have a vested interest in a player who was "robbed"). It should be noted that no players ever scored *any* adjusted goals, and therefore nothing has been "taken away" from them.

 

[kdfsjljklgjfg]

In 2000-01, Pavel Bure scored 59 goals.

The top 5 goal scorers that year were:
Bure, 59 goals
Sakic, 54 goals
Jagr, 52 goals
Bondra, 45 goals
Kovalev, 44 goals

In 81-82, Gretzky had his NHL-record 92 goal season.

That year, the NHL goal leaders were:
Gretzky, 92
Bossy, 64
Maruk, 60
Ciccarelli, 55
Vaive, 54

As you can probably see, it was easier to score goals in 81-82 than 2000-01. Therefore, one can argue that each of Bure's goals was more difficult to score, and thus "worth" more.

Ultimately, when adjusting those goals, Gretzky gets 68 adjusted goals, because it was relatively very easy to score goals in the 81-82 season. Bure gets 65 for 00-01 because it was a relatively difficult time to score goals.

So the way it comes out, Gretzky's season was "worth" 68 goals, and Bure's was "worth" 65. It's an estimation of the goal value relative to the era.

 

[ted2019]

Ok then. What is the formula used to figure all this stuff out?

 

[hatterson]

Ok then. What is the formula used to figure all this stuff out?

The simplest adjustment is to adjust by league wide average goals per game.

So in the 81-82 example used above, Gretzky scored 92 goals in a season that had an average of 8.025 goals per game scored league wide.

In 00-01 Bure scored 59 goals in a season that had an average of 5.513 goals per game scored league wide.

Next you pick either a comparison year (95-96 for example, 6.285 average) or you choose a comparison era (say average scoring from 90-91 to 96-97) or you choose an average scoring level to adjust to (say 6 goals/game).

Then it's just a little bit of math

Bure: 59 / 5.513 * 6 = 64.2218
Gretzky: 92 / 8.025 * 6 = 68.785

There's more advanced methods that are applied depending on which source computes them, but generally the difference isn't that large. The most significant is when dealing with shortened seasons or with seasons in the 50s and earlier where players only had 60 or 70 games instead of 82 (the 80 games played in the 70s/80s and the odd 84 game season is adjusted for, but the difference is very small either way)

Hockey-Reference, for example, adjusts for roster size (smaller roster means a player has more chance to score theoretically, not sure how I really feel about it), season length (70 games vs. 82) and for era differences excluding the player in question (Average goals in the year *excluding* the ones that player scored). They describe them in a little more detail on their page, and include an example of Gordie Howe's 52-53 season.

http://www.hockey-reference.com/about/adjusted_stats.html

 

[TheDevilMadeMe]

This thread on methods of comparing offense across eras may be of interest: http://hfboards.mandatory.com/showthread.php?t=1484885

H-R's adjusted goals formula is a rather crude method of doing so.

 

[Uncle Rotter]

As you can probably see, it was easier to score goals in 81-82 than 2000-01.

It's also easier to score goals on the power play, but Adjusted Goals doesn't take that into consideration.

 

[Bear of Bad News]

It's also easier to score goals on the power play, but Adjusted Goals doesn't take that into consideration.

I've seen versions that attempt to adjust for that.

Why don't you propose something?

 

[TheDevilMadeMe]

I've seen versions that attempt to adjust for that.

Why don't you propose something?

Are you talking about adjusted even strength goals/points, which completely removes PP production? I haven't seen a version that then separately adjusts PP production and adds it back in.

 

[Bear of Bad News]

Are you talking about adjusted even strength goals/points, which completely removes PP production? I haven't seen a version that then separately adjusts PP production and adds it back in.

I've seen versions that attempt to not only control for the league-wide scoring environment, but for the number of power plays (and times shorthanded) that an individual player's team receives.

Having said that - I can't find one online at the moment.

 

[Black Gold Extractor]

Adjusting between eras is complicated stuff. One of the most comprehensive ones would be "Goals Versus Threshold" (GVT), which attempts to weigh various factors while giving a player how much they affect goal differential versus an arbitrary "marginal player".

(In all honesty, I would rather look at a modified version, the yet to be defined "Goals Versus Average" (GVA), assuming that an average player contributes 0 to goal differential. That way, all GVA ratings on a team could be added to see if it comes up with the team's actual goal differential as a test.)

In terms of looking at star players, I've grown fond of just seeing how many top 10 or 20 finishes they have in their careers as well as awards and honors. Even during the 21-year reign of terror of Gretzky, Lemieux, and Jagr, star players will still end up on the leaderboard at least occasionally...

For Pat LaFontaine,

Goals: 5, 5
Assists: 2
Points: 2, 8
Honors/Awards: 1 Bill Masterton Memorial Trophy win, once selected to the Second All-Star Team

 

[Bear of Bad News]

(In all honesty, I would rather look at a modified version, the yet to be defined "Goals Versus Average" (GVA), assuming that an average player contributes 0 to goal differential. That way, all GVA ratings on a team could be added to see if it comes up with the team's actual goal differential as a test.)

This is true (and would have value). Ideally, both would be published.

The reason that a replacement-level benchmark is preferred (at least by me) is that being "average" does have defined value (just look at the salaries of average-level NHL scorers). When you compare against average, the injury-plagued superstar fares worse than the consistent (but healthy) above average player.