Post-Game Talk: Oilers win the battle but lose the war against their arch nemesis…

[McAsuno]

5 large is crazy

That will clearly teach him not to do it again!
NHL's DOPS is such a crock of shit lmao.
Kempe makes 5.5M annually, but that 5k will surely hurt him.
George Parros, f*** off.

 

[PositiveCashFlow]

LA has it down to a science. They will just rotate the next guy to do the slewfooting. Nothing to see here from the leagues standpoint.

What is that, like $50 for most people?

 

[BlackDogg]

McDavid got lucky twice on the same night

Did his wife hit the post as well?

 

[McBeastMode]

That will clearly teach him not to do it again!
NHL's DOPS is such a crock of shit lmao.
Kempe makes 5.5M annually, but that 5k will surely hurt him.
George Parros, f*** off.

That's about $90.00 to a person making $100,000 a year..

 

[McSpecial DraiBlend]

That will clearly teach him not to do it again!
NHL's DOPS is such a crock of shit lmao.
Kempe makes 5.5M annually, but that 5k will surely hurt him.
George Parros, f*** off.

They need to fix those fines

 

[BlackDogg]

Parros is likely jealous that he's basically a broomball player as compared to McDavid as far as skating goes.

 

[Sra1974]

That will clearly teach him not to do it again!
NHL's DOPS is such a crock of shit lmao.
Kempe makes 5.5M annually, but that 5k will surely hurt him.
George Parros, f*** off.

Hate parros but this is a negotiated amount with the union, he can’t do more.

 

[TheNumber4]

5K fine to slew foot Makar/MacKinnon/Stone/Eichel in the playoffs, noted.

 

[bellagiobob]

5K fine to slew foot Makar/MacKinnon/Stone/Eichel in the playoffs, noted.

If that were only true. Gary's wet dream would be to suspend McDavid if he even got near any of his darlings.

 

[brentashton]

Hate parros but this is a negotiated amount with the union, he can’t do more.

Unless he chooses to suspend him. No where does it say that a slew foot is a 5k fine. It’s all judgemental.

 

[TB12]

Parros is a f***ing idiot. I have said this many, many times but it needs repeating once again. f***ing clown.

 

[McTedi]

That will clearly teach him not to do it again!
NHL's DOPS is such a crock of shit lmao.
Kempe makes 5.5M annually, but that 5k will surely hurt him.
George Parros, f*** off.

The 5K fine is a joke, how do 4 refs miss that call entirely.

 

[Tobias Kahun]

The 5K fine is a joke, how do 4 refs miss that call entirely.

They didnt "miss" the call.

They ignored it.

It was 1-0 game in which they were trailing.

 

[Senor Catface]

Something like that happens almost every game

Wow, so weird. He got a fine for something that happens every game.

 

[CrazyJoeDavola]

If that were only true. Gary's wet dream would be to suspend McDavid if he even got near any of his darlings.

Yup. He wants Oilers failure as it increases the chances of McD leaving for greener pastures down south.

Think of all of the league sabotage against us:

Chiarelli draft picks
Keith retirement cap hit sabotage

I can't wait until they somehow sabotage our ltir cap plans. They've already announced that there will be an investigation lol

 

[McTedi]

Something like that happens almost every game

Yah sometimes when guys are challenging for a puck and engaged in a battle. This wasn't that, Kempe came in after the puck had left just to take a dirty run at McD. It's a dirt bag play and can easily turn into an injury, they really need to start suspending guys for it.

 

[bucks_oil]

Just some additional research to add to what I think is implicitly true analyzing the 2 on 1 on Stu last night:

This article also denotes that Pass Location and East-West One timers are not considered in these XGF models:

"So when new data, new shot information, becomes available from NHL they run it through the code and show you xG value, but it will be only about 80% accurate prediction, in addition to all inaccuracies of shot registration by NHL, to tell how probable was that shot to become a goal.

For example, when the model says shot had 2% or 0.02 xG value in reality chances were somewhere between 1.6% and 2.4% and it was still possible that shot will go in on any given shot like that.

Another example, let’s say shot was in close and the model shows 40% chance of scoring, 0.40 xG value. In reality it was somewhere between 32% and 48% chance but model can’t say exact because it’s missing information/attributes/features like: was the goalie screened, was it cross crease pass for “back door one timer” or where goalie or opposition player were located. At the same time commercial (non-public) models will have some of those additional features and will narrow the range saying that this shot had between 34% and 46% chance to become a goal."

Expected Goals (xG) Models Explained

<em>(Editors Note: <strong><a href="https://twitter.com/YuriGelfat">Yu...

jetsnation.ca

Here's another article:

"Perhaps the biggest gap is pre-shot movement. We know that passes before a shot affect the quality of the scoring chance, but the pbp data does not include them."

Expected Goals Model with Pre-Shot Movement, Part 1: The Model

There are few questions in hockey analytics more fundamental than who played well. Consequently, a large portion of hockey analysis has been focused on how to best measure results. This work is som…

hockey-graphs.com

Wow... I CANNOT believe that something as fundamental as East-West pass (distance, speed) to a one-timer is not factored in to xG models.

That's absolute hogwash.

I'm all for advanced stats, but this just shows how much they are in their infancy when it comes to goaltending.

Eye-test and error assessment for me. Goalies (just like players) make plenty of errors per night, if you can't see an error on a given goal, you gotta give the benefit of the doubt to the guy making millions. The best goalies stand up to this assessment and eye-test... the worst ones show you enough errors to let you pass judgement.

 

[oobga]

Yup. He wants Oilers failure as it increases the chances of McD leaving for greener pastures down south.

Think of all of the league sabotage against us:

Chiarelli draft picks
Keith retirement cap hit sabotage

I can't wait until they somehow sabotage our ltir cap plans. They've already announced that there will be an investigation lol

Got us with McLellan too. 2 picks penalty for reworking our GM/coach with people that were fired on expiring deals using a rule that I think only existed for a year.

I will say though, many GM's would have fought it, like Lucky Lou does every time. As usual we just smiled and said "OK!". Clearly the league thought the rule was stupid.

 

[TheNumber4]

Wow... I CANNOT believe that something as fundamental as East-West pass (distance, speed) to a one-timer is not factored in to xG models.

That's absolute hogwash.

I'm all for advanced stats, but this just shows how much they are in their infancy when it comes to goaltending.

Eye-test and error assessment for me. Goalies (just like players) make plenty of errors per night, if you can't see an error on a given goal, you gotta give the benefit of the doubt to the guy making millions. The best goalies stand up to this assessment and eye-test... the worst ones show you enough errors to let you pass judgement.

I agree. It’s definitely a huge omission. And it’s something that needs to be considered when assessing goalie performances on a single game basis. XGF will tell you Kuemper outplayed Skinner in that game. The same Kuemper who was not challenged with as sure fire a goal as that 2 on 1, and that same Kuemper who let in a goal to Stu’s 0 goals. Now if Kuemper had made a miraculous save on that tough McDavid shot off the Nurse rebound, I’d give it to him, but he didn’t. The GSAX in this game ended up 2.15 for Kuemper and 2.05 for Stu, that wrong assessment of only 6% chance on Stu’s 2 on 1 explains that difference, proper assessment of the danger on that 2 on 1 would have put Stu ahead, where he belongs. For a single game I would trust an informed observer like yourself assessment over xGF.

 

[Behind Enemy Lines]

That wasn't a particularly "short pass between circles", it was a lot closer to "dot-to-dot".

And it doesn't look here like he was particularly late with the read, he needs to stay with the passer to respect a potential shot until the puck is actually off his stick.

Tremendous save (and read)... especially since we know Skinner is (big and therefore) not the fastest post to post.

View attachment 960805


As for Draisaitl's 2:1 in the third, are you referring to McDavid's goal... this play below?...

To me the first stop (which was made) is easier than Skinner's, since Keumper doesn't have to move that far... the pass is a longish pass, but the lateral distance (post-to-post) required of Keumper is much less than Skinner's save, since his came off the rush and he was well on top of the crease...

Below, Keumper only has to slide central off the post and get as close to the top of the blue as he can.

Both good saves, I think Skinner's took more work, harder push than Keumpers.

I would hope the model would not value them the same.

View attachment 960809

Good discussion here with appreciation to all. Will add that your post and visuals reinforce another subjective, bias element involved with event interpretation that comes into play notably close to 'home' with Skinner and Oiler goaltending broadly.

I find the Jets fanbase article @TheNumber4 posted really interesting which shared some interesting insight into the tracking limitations.

***"They call it the accuracy of the model (often measured by AUC value or/and Log Loss).
For these public models the accuracy of predicting shot to become a goal is between 76.7% to 79.9% from these tests. Based on Maurice’s comments and other sources, commercial models show accuracy of about 85% which is not perfect either."

These are all great resources with big data ability to parse out the patterns in the game. However the limitations should be acknowledged in the evaluation of what it tells us. Important to also distinguish that real life application work by teams includes a deep dive into each and every goal to assess a number of considerations including: did the system fail? were there player mistake(s) that led to the goal?

This is the fastest game on earth with a high degree of random, non-repeating events. I love the new approaches to evaluate this game's complexity, however, models with 20% variance in reliability need to be couched in terms of conclusions being drawn. And adding personal bias above the data has a further distorting effect.

On the Skinner early game stop. He made it and at a critical, potentially defining moment of that game. LA as road team pushed hard in the first period and with style of game they play and the close checking playoff style by both teams, the game result could have been different if the Oilers were chasing.

EDIT: ***Adding my own assumption that the math used within this analysis is accurate!

 

[AddyTheWrath]

Wow... I CANNOT believe that something as fundamental as East-West pass (distance, speed) to a one-timer is not factored in to xG models.

That's absolute hogwash.

I'm all for advanced stats, but this just shows how much they are in their infancy when it comes to goaltending.

Eye-test and error assessment for me. Goalies (just like players) make plenty of errors per night, if you can't see an error on a given goal, you gotta give the benefit of the doubt to the guy making millions. The best goalies stand up to this assessment and eye-test... the worst ones show you enough errors to let you pass judgement.

They're included in private models (Clear Sight Analytics, Sportlogiq) because they have information about what happens in the play before a shot is taken. But this information isn't publicly available.

It's the reason why NHL teams pay Sportlogiq and not Dom lol

 

[OilerTitanFan]

The back up was the reason they went to the finals. Skinner at his best Paris like this, he needs someone pushing him to keep him honest.

How's the backup vs the wild last night?

The Oilers need to ice games when the opportunity arises. They will get scored on at some point. Empty nets may not seem like much in the grand scheme of things, but putting a game out of reach is always key regardless if there's a goalie in net or not.

True, extra point could give mcD additional bonuses

 

[bucks_oil]

Good discussion here with appreciation to all. Will add that your post and visuals reinforce another subjective, bias element involved with event interpretation that comes into play notably close to 'home' with Skinner and Oiler goaltending broadly.

I find the Jets fanbase article @TheNumber4 posted really interesting which shared some interesting insight into the tracking limitations.

***"They call it the accuracy of the model (often measured by AUC value or/and Log Loss).
For these public models the accuracy of predicting shot to become a goal is between 76.7% to 79.9% from these tests. Based on Maurice’s comments and other sources, commercial models show accuracy of about 85% which is not perfect either."

These are all great resources with big data ability to parse out the patterns in the game. However the limitations should be acknowledged in the evaluation of what it tells us. Important to also distinguish that real life application work by teams includes a deep dive into each and every goal to assess a number of considerations including: did the system fail? were there player mistake(s) that led to the goal?

This is the fastest game on earth with a high degree of random, non-repeating events. I love the new approaches to evaluate this game's complexity, however, models with 20% variance in reliability need to be couched in terms of conclusions being drawn. And adding personal bias above the data has a further distorting effect.

On the Skinner early game stop. He made it and at a critical, potentially defining moment of that game. LA as road team pushed hard in the first period and with style of game they play and the close checking playoff style by both teams, the game result could have been different if the Oilers were chasing.

EDIT: ***Adding my own assumption that the math used within this analysis is accurate!

They're included in private models (Clear Sight Analytics, Sportlogiq) because they have information about what happens in the play before a shot is taken. But this information isn't publicly available.

It's the reason why NHL teams pay Sportlogiq and not Dom lol

Agree... I'm all for the discussion. It is an interesting one. And I agree BEL, so many times I see us score a goal and 100% of the attribution goes to our "great play" or "net front presence" that lead to the goal. Then the same goal gets scored against us and Skinner "was slow across" or "gave up a bad rebound". It's infuriating... I honestly think Oiler fans are unreasonably harsh on goaltending (I've noticed it on here for YEARS) and give nearly zero consideration to the idea that goals against are TEAM events... remind me never to come play beer league in Edmonton lol.

Having said that... I come at this from (yea goalie brothership code) but also a biotech guy. If we developed an assay in our labs that had even 15% false positive or false negative... well then we DON'T HAVE AN ASSAY. I'm not a statistician myself, but with that much variance what you are looking at is mostly noise... and more problematic, if fundamentally MEANINGFUL information is systematically omitted from the model... well then not only do you have noise, but you also can't even argue that that noise will "balance out in the end" and as the n-size increases, you can eventually see signal from the noise.

Take the extreme example from our discussion (east-west pass to one-timer). If goalie A is on a team with inexperienced D that cannot defend a cross crease pass, and gives up a diproportionate number of 2on1s... and goalie B is on a team that doesn't, then goalie A is going to be SYSTEMATICALLY slaughtered by the, frankly, completely incorrect model. Why? Well not only is he consistently seeing more extremely high danger shots, but he is also repeatedly being penalized by the model for not saving a shot that it has erroneously assigned too low a value to.

Addy, thanks for sharing Sportlogiq, I had heard of it, but didn't know it is considered a gold (probably not the right word given the above... let's say bronze) standard.

 

[AddyTheWrath]

Agree... I'm all for the discussion. It is an interesting one. And I agree BEL, so many times I see us score a goal and 100% of the attribution goes to our "great play" or "net front presence" that lead to the goal. Then the same goal gets scored against us and Skinner "was slow across" or "gave up a bad rebound". It's infuriating... I honestly think Oiler fans are unreasonably harsh on goaltending (I've noticed it on here for YEARS) and give nearly zero consideration to the idea that goals against are TEAM events... remind me never to come play beer league in Edmonton lol.

Having said that... I come at this from (yea goalie brothership code) but also a biotech guy. If we developed an assay in our labs that had even 15% false positive or false negative... well then we DON'T HAVE AN ASSAY. I'm not a statistician myself, but with that much variance what you are looking at is mostly noise... and more problematic, if fundamentally MEANINGFUL information is systematically omitted from the model... well then not only do you have noise, but you also can't even argue that that noise will "balance out in the end" and as the n-size increases, you can eventually see signal from the noise.

Take the extreme example from our discussion (east-west pass to one-timer). If goalie A is on a team with inexperienced D that cannot defend a cross crease pass, and gives up a diproportionate number of 2on1s... and goalie B is on a team that doesn't, then goalie A is going to be SYSTEMATICALLY slaughtered by the, frankly, completely incorrect model. Why? Well not only is he consistently seeing more extremely high danger shots, but he is also repeatedly being penalized by the model for not saving a shot that it has erroneously assigned too low a value to.

Addy, thanks for sharing Sportlogiq, I had heard of it, but didn't know it is considered a gold (probably not the right word given the above... let's say bronze) standard.

Sportlogiq actually employs an army of part timers to validate the information they receive from their video processing AI. So I would tend to put a bit more stock into their information compared to others.

That being said, even they’re aware of the limitations of analytical models and that you have to marry what you see on the ice with the eye test. All numbers are meaningless without proper context.