No Distinction Between Different Games States In HockeyReference's Adjusted Formula - How Much Does This Matter?

[GlitchMarner]

I like HR's adjusted formula for valuing goals and points. Points and goals are worth more (ie more valuable) when scoring across the League is lower. This is simply a basic mathematical fact. There is really no subjectivity involved.

A 50 goal season in a league where 8.02 goals are scored on average per game is not equal in value to a 50 goal season in a League where 5.67 goals are scored per game. The latter season is unquestionably more valuable to a team.

However, when we get into making comparisons and asking which player played better, there will always be a degree of subjectivity involved. You cannot simply look at or consider one thing and say, "This says Player A was better; therefore, he was better; case closed."

When it comes to the 80-game era and beyond, the main flaw in Hockey Reference's adjusted formula seems to be the fact that it does not differentiate between scoring at different states of a game (ES, PP, SH). This is only a problem for the following reason: When there is a rise in PP opportunities, teams have a tendency to give increased PP time to their best scorers at a rate that is disproportional to the rate at which second-rate scorers receive an increase in PP time. This skew in additional PP time toward the best scorers has the effect of allowing them to score a greater percentage of the League's goals than the top scorers score in seasons where PP opportunities are lower, thus inflating their overall adjusted points as per HR in relation to the adjusted point totals of players playing in seasons with lower Power Play rates. If teams tended to distribute increased PP ice time uniformly between first and second-rate scorers, then it wouldn't matter for comparison purposes how much higher PPOs are in certain seasons than in others. If scoring was to rise considerably or significantly in seasons with more PPOs, top scorers would be in a free market type of competition against each other and against top scorers from other seasons to score as much as possible relative to the League-wide Goals per Game rate. If increased PP rates were to make little to no different to overall League scoring, then they wouldn't be relevant when comparing scoring across seasons.

This isn't to say other comparison methods aren't without their flaws. The strength of the top ten, top 15, top 20 and especially top five scorers will vary from season to season. Some eras simply have more high-end scoring talent than others.

Click to read more...

 

[pabst blue ribbon]

It's probably very hard to make era adjustments for scoring because Time On Ice stats have been non existent for most of NHL History. It's hard to determine the value of the average 5v4 PP in a certain era without knowing how efficient the 'average' team was scoring on it. I think the lack of actual TOI stats is quite determintal to the cause of using stats in a historic context in many ways

 

[BraveCanadian]

I'll throw in my 2 cents: when comparing players in different years, it isn't just trends in the proportion of game states like ES/PP/SH - although PPO and PP time in general is a big one- it is also:

• TOI and shift length changes - shift lengths changed dramatically in the 80s and there have been big fluctuations in TOI standards for top players
• usage/deployment of non-1st liners because these change in trends as well, for example in the 70s/80s it wasn't unusual to have lower lines still have some decent production whereas before and after they are often specialists (or plugs). In some eras there are "star" players and specialist defensive players and in some eras there is a premium on two-way players.
• "external" factors. For example, the introduction of TV timeouts in '93 that helped it become such an outlier year, or obstruction crackdowns in 95-96 and 05-06 (although accounting for PP time will catch a lot of this).

In short, yes, a goal in a league with average gpg of 8 is worth less than one in a league with an average of 6.. but IMO, assigning portions of value to individuals is a LOT more complicated when comparing seasons across time.

Which, unfortunately, is mostly what people try to do with adjustments based on average scoring.

 

[The Panther]

I like HR's adjusted formula for valuing goals and points. Points and goals are worth more (ie more valuable) when scoring across the League is lower. This is simply a basic mathematical fact.

But is it really this simple?

For example, what about in the following situation. Let's imagine that in one given NHL season...

-- Team A scores 400 goals
-- Team B scores 240 goals
-- The statistically average team (offensively) scores 320 goals

-- Forward A on Team A scores 55 goals (13.8% of team's total)
-- Forward B on Team B scores 46 goals (19.2% of team's total)

Clearly Forward B made a bigger contribution (goals-wise... let's not overcomplicate this) relative to his team's offense than Forward A did.

However, if Hockey Reference (or another source) were "adjusting" these goals totals according to individual teams' goals that season instead of against historically-average scoring rates, then Forward B (who scored fewer goals) would come out with "more" goals scored than Forward A... even though both played in the same league in the same season and shot pucks against the same goalies.

My point is that it's false to assume that league-wide scoring rates affect all players and all teams equally. But Hockey Reference's system does this.

Real-World Example:
-- In 1970-71, Boston scored 399 goals (pre-overtime) in 78 games, or 5.12 goals per game.
-- Boston's Ken Hodge scored 43 goals. The Bruins thus scored 356 goals not by Hodge.
-- In 1970-71, Minnesota scored 191 goals (48% of Boston's total goals).
-- Minnesota's Bill Goldsworthy scored 34 goals to lead the club. The North Stars thus scored 157 goals not by Goldsworthy.
-- According to Hockey Ref, for that season, Hodge scored 39 adjusted goals and Goldsworthy scored 31.

In other words, Hockey Ref. tells us (devoid of further context) that Hodge goals were "more" in the subjective sense than Goldsworthy's goals were. Is this correct? No. If you take away every goal that Hodge scored for Boston, the Bruins were still the #1 offensive team in the NHL by a country kilometer. Clearly, Goldsworthy's 34 goals for Minny were more valuable than Hodge's 43 for Boston.

It's a very random example, but I think you get my point: Putting a blanket value on what one goal is, and then applying it to every player in the League, across generations, is extrememly sketchy. So, I can't conclude that every one goal in 2004 or whenever is automatically worth more than every one goal in 1984 or whenever.

I think these formulae for adjusting stats vs. historical averages are nice to give us very rough, ballpark-figures for long-past seasons in eras we might know nothing about or have little context for. Like, someone might look at Beliveau's eye-popping RS and playoffs in 1956 and wonder how that would translate into later, higher-scoring, 80+ game eras. A quick look at one formula's "adjusted stats" can give us a very rough idea of this. But it's only a very rough idea, and there is no accuracy at all to the calculation, needless to say.

 

[BraveCanadian]

Yeah, as soon as you start looking at some examples with stats adjusted by average scoring, it quickly falls apart.

 

[GlitchMarner]

Great post, Panther. Very well-thought out and articulated.

I'm not ashamed to admit that I had never really thought of it from that angle.

I think maybe we can make the following distinction: That scoring more relative to the overall scoring rate of the League does not necessarily make one a better or more apt scorer.

In the simplest, perhaps most technical sense, a player who scores more goals relative to the League-wide GPG total likely adds more value in the sense that his goals probably directly translate into more wins and points for his team.

If, for example, Rick Vaive scores 53 goals in a league that averages over 7.50 goals per game and Mats Sundin scores 40 goals in a League that averages just over 5.50 goals per game, I think Sundin's goals probably directly add more wins to his team's total than Vaive's.

But does that mean that Sundin was a more talented goal scorer or a more natural one? I think these are inherently different matters. You would have to require personal judgement (and perhaps the opinions of others) to answer the question of who the more gifted or natural scorer was.

I would say that even the question of whose scoring was more impressive or better when certain or all contexts are considered (who had more PP time, who had more time on ice in general, who had better linemates, who was given more offensive freedom, etc etc) is to an extent a subjective question (you can track TOI and such, but opinions will differ on which linemates were more talented relative to the League and how much a player's team helped his scoring versus how much he carried the load). When it comes to Vaive and Sundin, I'd say the latter's scoring is both more valuable to his team in the standings and more impressive all things considered; both often played with rather lackluster players, but one played in an era where blatant cherry-picking (or "goal-sucking") was less prevalent and where goals were generally much harder to come by, so a raw difference of about a dozen goals is more than negated.


This season Auston Matthews received a ton of attention for his 69 goals scored. He placed 12 goals ahead of Sam Reinhart (who was second in League scoring with 57 goals).

Matthews' Toronto Maple Leafs scored 298 goals. Reinhart's Florida Panthers scored 265 goals. AM34 accounted for 23% of his team's goal total. Sam Reinhart accounted for 22% of his team's goal total (rounded up). Does this mean that these players provided almost equal value to their respective teams with their respective goal totals even though raw stats and adjusted totals (which in this case are exactly equal to one another) suggest Matthews' goal scoring was clearly more valuable to his team? To answer that one would have to delve deeply into team dynamics. Maybe I'm biased here (and this is only a feeling after all), but without any further thought, I would say that in this case GPG-based adjusted totals are more indicative of relative value. At any rate, Auston's goal scoring was certainly more impressive considering he had more than double Sam's Even Strength goal total.


It seems we have gotten to a point where people are used to distinguishing between the concepts of greater and better player. Furthermore, the player who wins the Pearson (best player) doesn't always win the Hart (most valuable to his team). Maybe we need to accept that being the more valuable scorer according to one particular metric is just that and it doesn't mean everything when it comes to making player comparisons.

 

[daver]

Here is a simple solution, don't use HR adjusted stats.

Use performance vs. peers as a statistical starting point.