Ratings at the Attribute Level

[jcard]

As someone who is trying to become familiar with the skill profiles of NHL players, I am curious at what level of granularity the developers try to rate players. For example, there is an extremely high correlation between the Getting Open and Shooting Accuracy attributes. I certainly understand that the complementary nature of these attributes means that you would generally expect to find a positive relationship in the proficiency of a given player for these attributes. The relationship seems so strong, however (a teamÂ’s top players at Getting Open and at Shooting Accuracy are often just about identical), that I wonder if basically the archetype system and overall talent is what is used to manually rate the players in the various Skill levels with the attributes relating to that skill then more or less mechanically assigned within a narrow range of that Skill rating (here, for example, manually deciding on Goalscorer Archetype and Overall Talent to get the desired Scoring Skill rating, with the Getting Open and Shooting Accuracy attributes each automatically set within a tight window around that Skill rating). This is not asked with a negative undertone; I just do not want to spend time studying Attribute ratings for a player if the player-specific assessment is made at the Skill level.

[skunt3m]

I am just a player, so don't take my word as gospel here:

-Yes, there are a lot of players rated based on templates. This is partly due to the scope of the project (tens of thousands of players) and partly based on how FHM works.

-As you've probably already realized, attributes in FHM are (mostly) based on rate stats. Getting open correlates with a player's shots per minute. Shooting accuracy correlates with goals per shot. Someone like Zach Hyman isn't an elite shooter in the real NHL, but in FHM he has an 18 in shooting accuracy because, as a very selective shooter, he consistently has a high shooting percentage. In other words, attributes in the game are based on results, not theoretical/observable skills.

-I say "mostly" because researchers have to a) project a player's production b) try and separate the player's production from the positive/negative effects of his linemates/coaches/icetime c) try and account for sample size, etc, etc... tough to get granular unless it's a player with a long history of consistent production.

-So in the case of, let's say, a young sniper just getting to the NHL, a) the researcher might know nothing about the player other than that he's a goalscorer, b) can only guess at how the player will translate/who he will play with... a rating of 15 is approximately NHL average in FHM. What he can safely guess is that the player will shoot more than average (16) and score more than average (16). In FHM that might equate to 20-25 goals with first line ice time and average linemates. Beyond that, might as well use a template, right?

[jcard]

I appreciate your considered remarks- and also your well-documented presentation of the issue with top-heavy scoring distribution.

To the point at hand: If you compare the respective Getting Open and Shooting Accuracy attributes of players with at least an NHL-average (15) Getting Open rating, you find that a difference of even two points between the two is very uncommon. As a layman, so to speak, I find it difficult to believe that there are basically no players who get their goals through a combination of, say, average ability to get open (shot attempt creation) and very good accuracy (conversion rate)- or vice versa.

[skunt3m]

Glad you enjoyed!

Zeroing in on the lack of extremes in GO (getting open) vs SA (shooting accuracy)... some food for thought:

---First, I would argue that this type of player is increasingly rare in the NHL - rare enough that he probably doesn't exist as a template for researchers to pick in the game.

Let's try to find some players who consistently fit your criteria over the past 7 seasons:
-have under 2.13 shots per game
-shoot at 16% or higher
-play over 16 mins a game
-play in 85% of their team's games

Here are the names that appear more than once:
-Ryan Dzingel (2 times)
-JT Miller (2 times)
-Mark Stone (3 times)

TJ Oshie should be there too. He has been injured too often to appear on my list but he has averaged 167 shots per 82 games over the past 7 seasons with a 17% shooting percentage. He is the poster boy for the type of player you are seeking IMO.

----Second, researchers attempt to weather-proof players against changes in role, linemates, coaches, luck, etc. This typically results in their attributes being fudged toward the mean (we are also ignoring attributes which logically would also have an effect on shot attempts/conversion rate - ie, speed, shooting range). They don't bet on players replicating the extreme results dictated by a big gap in GO vs SA. As evidenced by my list, very few guys do it season in and season out.

Take Mark Stone - he seems to be a reluctant shooter, especially in Vegas, but he has also had seasons where he has taken shots at an above average rate and knocked himself off my list. His results seem to be related to role... if he is called on to shoot, he can.

Andrew Mangiapane is another example. Whoever rated him made the safe bet that he wouldn't shoot at 20% again because who does that 2 years in a row?! Researchers may have to adjust their expectations next season.

----Third, the game seems to struggle with low GO, high SA guys. TJ Oshie has had some extreme ratings over the years (think 14 GO, 18 SA) and he never comes close to his real life results in my sims. If he scores, it's because he shoots a lot. If he doesn't shoot a lot, he doesn't score. Maybe it's better in the 2D engine. I wonder if the attributes are also fudged toward the mean for this reason.

Just for fun, let's try and find the opposite (high volume, low % shooters) over the past 7 seasons:
-have over 3.33 shots per game
-shoot at 10% or lower
-play over 16 mins a game
-play in 85% of their team's games

Here are the names that appear more than once:
-Taylor Hall (2 times)
-Nazem Kadri (2 times)
-Patrick Kane (2 times)
-Phil Kessel (2 times)
-Tyler Seguin (2 times)
-Brady Tkachuk (2 times)

[Colorado]

What site are you using to source these?

[skunt3m]

Originally Posted by Colorado

What site are you using to source these?

https://www.nhl.com/stats/skaters

Lots of filters here!