Advanced Analytics might be used to judge Selke Trophy candidates this year

[kickazz 5,459]

http://sports.yahoo.com/news/if-bergeron-doesnt-win-who-fills-nhls-selke-trophy-void-201919246.html

Selke winners have to have a:

1 – Faceoff winning percentage among the League’s best.

2 – Positive plus-minus rating, and especially a double-digit one.

3 – Offensive stats that are above average, even though this shouldn’t be a primary factor in defensive prowess.

On the positive side, PHWA voters have made an effort to dabble in analytics to make a Selke nominee’s case. Corsi and possession metrics are seen as mandatory in 2017; one hopes that further context will follow.

---

So for a couple years now LGW posters have been adamant about Corsi/Fenwick being an important way of judging defense. Selke trophy is given to the best defensive forward and it looks like the voters for the trophy will be taking a look at the advanced stats when deciding. 

 

Edited March 8, 2017 by kickazz

[toby91_ca 620]

I think there are a ton of flaws in using some of these stats to try and hand out a best defensive forward award.  I have no issue with considering the stats, but if you start placing too much reliance on it, that's when I have a problem.  Way too much in some of those stats that are way beyond the control of the player in question.  

[Echolalia 2,961]

17 minutes ago, toby91_ca said:

I think there are a ton of flaws in using some of these stats to try and hand out a best defensive forward award.  I have no issue with considering the stats, but if you start placing too much reliance on it, that's when I have a problem.  Way too much in some of those stats that are way beyond the control of the player in question.  

corsi is just a more sensitive +/-  It has its merits as long as people are aware of the limitations, the problem is most people aren't aware and use it as a holy grail of statistical evidence that player A is superior to player B.  Having said that, I prefer judging players based on quantitative evidence over the eyeball test whenever possible.

[Echolalia 2,961]

24 minutes ago, Jonas Mahonas said:

I don't think you understand Corsi based on what you just said.  Advanced stats ARE based off of quantitative evidence.  They are not advanced opinions based on someone's perspective.  People recognized that +/- was only a good indicator of a player's 5 on 5 value over longer periods of time and also saw that having a better cast of supporting players lead to a better plus minus.  So Advanced Stats were created to take other factors into account.  Those factors are not judgemental or open to interpretation, however.  They are statistics.

 

I don't think you understand my post, based on what you just said. I agree that advanced stats are quantitative.  I never suggested otherwise.  

[ChristopherReevesLegs 7,022]

Sorry Toews. No more media Selke's for you.

[GMRwings1983 8,794]

This is the hardest award to judge.  Can't just look at the numbers.  One has to attend games in person and see how these players play when they don't have the puck and when the camera isn't necessarily focused on them.  

I wouldn't mind seeing the players vote for this award.  

[Number19 159]

It really shouldn't come down to points even if it's the least important criteria. Was Gainey ever a point producer?

[toby91_ca 620]

5 hours ago, Echolalia said:

Having said that, I prefer judging players based on quantitative evidence over the eyeball test whenever possible.

This is where I'd differ...extremely and this is what I want to avoid with these advanced stats.  I'd prefer the eyeball test 100 times out of 100.  I don't want someone voting for an award having never watched a guy play.  Stats are needed as well, but especially for this award, way to dangerous to try to simply use advanced stats to judge.

2 hours ago, Number19 said:

It really shouldn't come down to points even if it's the least important criteria. Was Gainey ever a point producer?

The issue is that it's the best defensive forward award.  A forward's main purpose is to score, so if they don't score, I don't think they deserve this award.

Gainey put up respectable enough numbers, certainly not a big scorer.

[kliq 3,755]

44 minutes ago, toby91_ca said:

This is where I'd differ...extremely and this is what I want to avoid with these advanced stats.  I'd prefer the eyeball test 100 times out of 100.  I don't want someone voting for an award having never watched a guy play.  Stats are needed as well, but especially for this award, way to dangerous to try to simply use advanced stats to judge.

 

The issue is how does any voter watch every top defensive forward play 82 games a year, its impossible. Not that I have a better way to judge, though I like GRM's idea of having the players vote. Everyone who plays knows who is the hardest person to play against, anyone who has played sports can identify with this.

[Echolalia 2,961]

1 hour ago, toby91_ca said:

This is where I'd differ...extremely and this is what I want to avoid with these advanced stats.  I'd prefer the eyeball test 100 times out of 100.  I don't want someone voting for an award having never watched a guy play.  Stats are needed as well, but especially for this award, way to dangerous to try to simply use advanced stats to judge.

The issue is that it's the best defensive forward award.  A forward's main purpose is to score, so if they don't score, I don't think they deserve this award.

Gainey put up respectable enough numbers, certainly not a big scorer.

The only reason why watching someone play is still valued for evaluating a player's performance is because we haven't come up with an entirely fool-proof of quantifying someone's on-ice worth just yet.  Unfortunately human observation is riddled with bias, and I suspect a lot of the people who actually vote on things like this have seen only a handful of games of each candidate, and rely more heavily on statistical evidence anyway.  

[ChristopherReevesLegs 7,022]

16 minutes ago, Echolalia said:

The only reason why watching someone play is still valued for evaluating a player's performance is because we haven't come up with an entirely fool-proof of quantifying someone's on-ice worth just yet.  Unfortunately human observation is riddled with bias, and I suspect a lot of the people who actually vote on things like this have seen only a handful of games of each candidate, and rely more heavily on statistical evidence anyway.  

There's a heavy media narrative as well that voters fall into line with. Kopitar, Bergeron, and Toews have all basically been crowned the only viable choices for the Selke for the foreseeable future.

One of my favorite narratives from the past few years is that Shea Weber is secretly a bad D-man. It's been going on for a while. First it was "he's only good cause of Suter" then "Only good cause of Josi" then "His fancy stats prove he's bad". Now that he's with Montreal that narrative has died off amongst the media folk. And look who he's paired with.. Markov. Guess who's having a career year at 38... Markov. He's a great defensemen. 

[Echolalia 2,961]

11 minutes ago, Jonas Mahonas said:

Ok, as long as you understand that the corsi and fenwick stats aren't eyeball tests.  There really aren't any limitations.  They are just stats.

Again, I never suggested that it was.  But there absolutely are limitations to how they can be used.  You wouldn't say someone with a high corsi for is a good goal scorer if he only has 2 goals on the season.  And you wouldn't say someone that is terrible with giveaways/takeaways, but takes a lot of low percentage shots from the perimeter whenever they get the puck (thus inflating their corsi) is a solid possession driver for their team (in fact I would argue that taking s***ty shots with virtually no chance of going in is essentially a turnover, thus worsening the possession of the team, yet corsi would say the opposite).  There is a context that needs to be realized with all statistics, and therein lies the limitations of their use.

[kickazz 5,459]

4 hours ago, toby91_ca said:

This is where I'd differ...extremely and this is what I want to avoid with these advanced stats.  I'd prefer the eyeball test 100 times out of 100.  I don't want someone voting for an award having never watched a guy play.  Stats are needed as well, but especially for this award, way to dangerous to try to simply use advanced stats to judge.

Eye ball tests are near impossible without the element of bias and the amount of games played by each player.