News and Blog Report: Analytics and their evolution

[CantHaveTkachev]

The Oilers need a shaman or cleric or exorcist to help with their drafting.

Like honestly it's so bad, it's not even explainable. It's not even a Chiarelli/Holland thing exclusively, it's been bad since like 1983.

The Kevin Lowe era was the best drafting era we've had post-1983 when they got Tikkanen in the 3rd or 4th round.

It's like Pocklington made some kind of deal where the 79, 80, and 81 drafts would bring him 5 Cups, but the catch was the team can't draft for another 50 years.

Like I had to literally look this up and really realize "holy crap ... this is even worse than what I remember".

The Oilers were actually pretty good at drafting under Chiarelli
The 2015 draft was a home run
Then Desharnais in 2016
Then Yamamoto and Skinner in 2017
Then Bouchard McLeod and Kesselring in 2018
Holland’s drafts so far have been poor

 

[guymez]

This was my biggest deterrent to the analytics at the beginning. "If you don't agree with me, you either don't understand or are just stupid." Igor and Tyler did that early but Tyler certainly softened as time went on.

Personally I had to evolve because I just wanted to watch the game an enjoy it and not drill down into Corsi, PDO, zone starts, pts/60, WOWY. But dinosaurs like me have to change and I have come to use the tools to see how I feel about a player. If I am going to say someone is shit, I should at least figure out why I think they are shit. LOL

You have been around longer than 95% of the posters on HF...even longer than I have but I can confirm what you are saying here.
Back in the day (almost 20 years ago) on here that arrogant attitude was very prevalent and it soured me on analytics but over time I came to realize their value. The moderation wasnt wound up quite as tight in those days and Igor and I had a couple of somewhat 'animated' disagreements. Not so much about the stats themselves but more about the delivery.
Tyler (Mudcrutch79) did soften as time went on...Igor didnt...LOL
Arrogant or not I did ultimately respect the work they put into the data they were coming up with.

Now I find that I appreciate analytics (I eventually evolved as well) but I would sure like to see the data the teams actually get. I know that isnt going to happen unless I become an employee but I am willing to bet that data accounts for more nuance/context than the data the fans can access.

 

[Behind Enemy Lines]

I've alway enjoyed looking at stats in hockey and baseball going back to childhood and curious about the numbers tracking as it evolved over decades. The 'Moneyball' book was like the apes in 2001 weaponizing themselves haha in that it broke through and crossed into mainstream the chase of underdog teams to find value and efficiencies against the sport's Goliath spenders.

I've kind of categorized this as Wild West phase in which many, untrained hobbyists flooded into the field with awesome curiosity but at times an arrogant attitude drawing definitive conclusions from faulty or limited data which was then often parroted into the vortex by acolytes. Now, entering a Maturing phase with trained data scientists now working in the industry and as hobbyists.

We still see some of the challenges with integrity of source data in advance stats like Goals Saved Above Expected where there can be a high variance in conclusions like Binnington 1st at 25.36 on one list and 5th at 16.2 on another (a variance of 9.16 GSAE). I also think some of the advance stats which tilt to offence creates some blindspots in the valuation of goal suppression d-men and forwards.





The source data through NHL Edge is becoming more reliable and standardized with puck and player tracking informed by in rink camera system. Reading how coaches like Cassidy incorporate data into their game strategy is interesting; how it can help pro scouting find latent value and such is interesting. The NHL is big business so utilizing information fully as a support function to improve all facets of operation is a no brainer. Fuels good fan discussion too.

 

[bobbythebrain]

Analytics should be left to the professionals. Average fans shouldn't try to use them. This is why you had countless people arguing that Jesse Puljujarvi and Nail Yakupov were dominant and valuable players.

"JP is elite defensively and makes his lines better"
"ConYak is one of the best combos in the NHL"

Those comments were pure lunacy if you had 2 working eyes

 

[McDNicks17]

Analytics should be left to the professionals. Average fans shouldn't try to use them. This is why you had countless people arguing that Jesse Puljujarvi and Nail Yakupov were dominant and valuable players.

"JP is elite defensively and makes his lines better"
"ConYak is one of the best combos in the NHL"

Those comments were pure lunacy if you had 2 working eyes

It would have been nice to have real analytics writers during the Pulju+McDavid era.

From what little glimpses of advanced metrics we saw, McDavid did have his best results with Pulju that year. He created chances and goals at the highest rate with him. That likely came down to Pulju's basically only NHL level skill was transition/neutral zone defense that he used to create a bunch of extra possessions for McDavid who was in absolute god mode and was having probably the best stretch of one-man rush hockey ever played in the league.

It would have been interesting for someone to break down stuff like neutral zone break-ups and solo rush numbers instead of the just seeing the Oilers "blogosphere" reference GF% and xGF%.


The media have access to those types of numbers, so I don't know if it just doesn't generate enough money/clicks or if it's just too much actual journalism for them. Like Gregor put out an interesting article a few years ago that broke down how the Oilers defensemen exit the zone and how they ranked amongst the other defenseman around the league using SportsLogIQ data. I don't really understand why Sportsnet or someone that already pays for SportsLogIQ isn't putting out articles using their data on the regular.

 

[Fourier]

It would have been nice to have real analytics writers during the Pulju+McDavid era.

From what little glimpses of advanced metrics we saw, McDavid did have his best results with Pulju that year. He created chances and goals at the highest rate with him. That likely came down to Pulju's basically only NHL level skill was transition/neutral zone defense that he used to create a bunch of extra possessions for McDavid who was in absolute god mode and was having probably the best stretch of one-man rush hockey ever played in the league.

It would have been interesting for someone to break down stuff like neutral zone break-ups and solo rush numbers instead of the just seeing the Oilers "blogosphere" reference GF% and xGF%.


The media have access to those types of numbers, so I don't know if it just doesn't generate enough money/clicks or if it's just too much actual journalism for them. Like Gregor put out an interesting article a few years ago that broke down how the Oilers defensemen exit the zone and how they ranked amongst the other defenseman around the league using SportsLogIQ data. I don't really understand why Sportsnet or someone that already pays for SportsLogIQ isn't putting out articles using their data on the regular.

SportsLogIQ is a very serious operation. Here is one of their job ads.

AI Researcher / Chercheur.euse en intelligence artificielle

We are the world's sports analytics leader, using cutting-edge AI technology to produce unique sports data and performance insights. These insights provide an edge to teams, leagues, media organizations, betting operators, and rights holders of all sizes. We use a variety of techniques to...

sportlogiq.bamboohr.com

 

[McDNicks17]

SportsLogIQ is a very serious operation. Here is one of their job ads.

AI Researcher / Chercheur.euse en intelligence artificielle

We are the world's sports analytics leader, using cutting-edge AI technology to produce unique sports data and performance insights. These insights provide an edge to teams, leagues, media organizations, betting operators, and rights holders of all sizes. We use a variety of techniques to...

sportlogiq.bamboohr.com

Yeah. They're doing some really cool stuff with AI.

I've seen a few demos where they're using it to track body and stick position. It can see the distance between players, if there's a stick/player in shooting/passing lanes, etc.

Last I heard they were contracted by 30 teams in the league. I wouldn't be surprised if it's every team now.

 

[bobbythebrain]

It would have been nice to have real analytics writers during the Pulju+McDavid era.

From what little glimpses of advanced metrics we saw, McDavid did have his best results with Pulju that year. He created chances and goals at the highest rate with him. That likely came down to Pulju's basically only NHL level skill was transition/neutral zone defense that he used to create a bunch of extra possessions for McDavid who was in absolute god mode and was having probably the best stretch of one-man rush hockey ever played in the league.

It would have been interesting for someone to break down stuff like neutral zone break-ups and solo rush numbers instead of the just seeing the Oilers "blogosphere" reference GF% and xGF%.


The media have access to those types of numbers, so I don't know if it just doesn't generate enough money/clicks or if it's just too much actual journalism for them. Like Gregor put out an interesting article a few years ago that broke down how the Oilers defensemen exit the zone and how they ranked amongst the other defenseman around the league using SportsLogIQ data. I don't really understand why Sportsnet or someone that already pays for SportsLogIQ isn't putting out articles using their data on the regular.

I'll give you the breakdown. It's easy

JP wasn't good in the neutral zone. He was just there WAAAAAAAAAAAAAAAAAAAAY to early, so the analytics reflect that he was in good positions defensively. He never carried a puck in EVER and rarely got caught deep cuz stopped skating at the blueline. He abandoned cycles like crazy also. He left that to McDavid and co.

The result for McD was that he didn't have to play defense, hence, why he was accused several times of being lackluster in that department, but cuz he's so gifted offensively, his #'s benefited.
I know he's not having a runaway season this year offensively, and I have no idea what the anyltics say, nor do I care, but his defense has been top notch this year. He is the first guy back most of the time, and he's broken up more threats than one can count. I'd say he's more gawd like this year than last tbh

Federov was one of the most dominant offensive players of his time. Then Bowman pulled him aside and asked if he could focus on more D as Stevie Y would provide the offense, and winning cups also takes defense. Federov went from top 5 point totals to mid totals, but was getting Selke nods like crazy, But I'm sure some "analytics" would show he regressed, or some fans would read it that way

McD is a better player this year imo

 

[Stoneman89]

The Oilers were actually pretty good at drafting under Chiarelli
The 2015 draft was a home run
Then Desharnais in 2016
Then Yamamoto and Skinner in 2017
Then Bouchard McLeod and Kesselring in 2018
Holland’s drafts so far have been poor

To be fair though, Chia was working with some higher picks and they were keeping their first rounders. And his drafts have the advantage of having an extra few years on Holland picks in which to develop. For all we know, some of the guys currently on the farm may turn out to be fine. For me, the jury is still out on Holland's drafts.

 

[Fourier]

Listening to the comments coming out of the NVIDIA event it is very possible that we will see a revolution in how analytics are used in the next decade. The new platform NVIDIA has developed has a focus on understanding video. This means that a new world opens up for the integration of AI in scouting. SportsLogIQ already advertised that their metrics are all linked to video. But we are probably in the embryonic phase of what might soon be possible.

 

[Slats432]

Interesting take.