Rumor: Leafs significantly expanding their analytics department

[Bluelines]

Worth noting the Caps had 105 points last year and won the cup. The year prior they had 120 and 118 respectively.

Dominating the regular season is nice but it's really how they come together the playoffs that matter.

Right and this team feels a bit like Frankenstein's Monster right now, lots of parts but IMO they don't work perfect together (yet). I think that is in part due to frequent movement of people in and out of the roster. I'm thinking 4 more good players and time will turn this team into a power house.

 

[Notsince67]

Yup. But is it the one you're implying?

Not implying anything. Teams like the Canes and Oilers built their teams around advanced analytics. I'm only saying that there is a lot of evidence that move too far from tried and tested methods developed over a 100 years or so is a recipe for failure. I said this last summer and I will repeat it. I have quite of bit of experience with quantitative models. A perfect comparison would be credit risk models for auto adjudication in the lending industry. They are handy for quickly wading through the low hanging fruit. Everyone in that business knows that nothing beats the manual adjudication of a skilled risk guy. The measurement of portfolio performance contrasting people vs models clearly support people. If I needed to burn through 50MM, I'd spend a smaller % on stats and a larger amount on boots on the ground scouting. The stats can direct the deployment more efficiently

 

[4thline]

Not implying anything. Teams like the Canes and Oilers built their teams around advanced analytics. I'm only saying that there is a lot of evidence that move too far from tried and tested methods developed over a 100 years or so is a recipe for failure.

There is a lot of evidence that mistaking work akin to a 12th grade Data Management project for "advanced analytics" and making decisions based on it isn't a smart move.

In your analogy, does the skilled risk guy look at numbers or judge based on solely applicant appearance ,charisma, and the curb appeal of their house?
Where do the skill guys numbers come from? How did they develop the benchmarks and guidelines that they use to make their skilled

You're underrating the infancy of sporting analytics relative to financial. Implementing analytics for hockey isn't at the stage of letting models make decisions for management, it's still about finding and gathering the information to help the skill guys make decisions rather than loaning money to person based on gut feel and the fact their father is "good people" and they wore their sunday best to the bank

 

[Notsince67]

There is a lot of evidence that mistaking work akin to a 12th grade Data Management project for "advanced analytics" and making decisions based on it isn't a smart move.

In your analogy, does the skilled risk guy look at numbers or judge based on solely applicant appearance ,charisma, and the curb appeal of their house?
Where do the skill guys numbers come from? How did they develop the benchmarks and guidelines that they use to make their skilled

You're underrating the infancy of sporting analytics relative to financial. Implementing analytics for hockey isn't at the stage of letting models make decisions for management, it's still about finding and gathering the information to help the skill guys make decisions rather than loaning money to person based on gut feel and the fact their father is "good people" and they wore their sunday best to the bank

In the world of risk, application comments from originator matter as does verification (ie income) etc. Real world forces at the moment such as interest rate speculation from economists, competitor strategy changes also have a factor. Face to face lending would certainly have a huge impact but that kind of relationship lending really only happens with credit unions and smaller regional US banks. Big Banks dont really do this to a large extent unless you are into private banking.

 

[4thline]

In the world of risk, application comments from originator matter as does verification (ie income) etc. Real world forces at the moment such as interest rate speculation from economists, competitor strategy changes also have a factor. Face to face lending would certainly have a huge impact but that kind of relationship lending really only happens with credit unions and smaller regional US banks. Big Banks dont really do this to a large extent unless you are into private banking.

So the skilled credit risk guys do have numbers that they use, and those numbers come from somewhere? And someone at some point analyzed other numbers to develop the guidelines and understand the relationships that the skilled guys now use? Of course.

Your post #452 is the hockey equivalent of the old-timey bankers saying "bah humbug, I know everything I need to know, I can tell the measure of a man/company by a handshake and a plant walk through so this stuff is useless" to modern (20th century) analysis and techniques

 

[Menzinger]

One of the first teams to jump on the analytics bandwagon was the Oilers. A cautionary tale for sure.

This isn’t particularly true, there were a fair number of teams who beat them to the punch before they finally hired Tyler Dellows.

Not to mention, it was a common theme for a few teams to hire a stats guy (usually directed by ownership) and then have them be completely ignored by the rest of the front office. Stats writers have been by far the loudest critics of the Oilers the past few years (even during that year they made the playoffs). Chia is in record saying he doesn’t pay attention to them (remember that infamous Seguin trade video?)

 

[Leafs87]

I would love to hear how the Leafs cap would have been better spent than on Tavares. By all means prove me wrong.

That’s not that point.... he raised our internal cap up when he became the second highest paid center without being the second best when we had three high priced RFAs to sign.

 

[Mess]

Yup. But is it the one you're implying?

Does cap management fall under the analytics umbrella ?

If so, then despite the production of McDavid and Draisaitl finishing 2 & 4 in NHL scoring combining for 221 points and their individual analytics through the roof, the impact of their combined cap hits of $21 mil and the impact on the rest of the players around them is most influential in the teams overall demise and 7th last in standings.

Analytics match individual performances but team metrics in regards to cap management seems to be the problem.

Hopefully that +$50 mil invested in more analytics is going to help Dubas solve this dilemma when he is faced with 3 players with great individual analytics consuming too much salary cap %, so the Leafs team doesn't follow the Oilers down the drain.

Dubas will need to use these individual analytics to find cheap contracts with strong supporting corsi metrics to surround a team with limited cap space.

 

[Albi34]

That’s not that point.... he raised our internal cap up when he became the second highest paid center without being the second best when we had three high priced RFAs to sign.

So what, would you would have been happier going into the season with matthews kadri ??? as our top 3 centers because it saves some cap space?

 

[Martin Skoula]

This isn’t particularly true, there were a fair number of teams who beat them to the punch before they finally hired Tyler Dellows.

Not to mention, it was a common theme for a few teams to hire a stats guy (usually directed by ownership) and then have them be completely ignored by the rest of the front office. Stats writers have been by far the loudest critics of the Oilers the past few years (even during that year they made the playoffs). Chia is in record saying he doesn’t pay attention to them (remember that infamous Seguin trade video?)

If the Oilers had a stats guy worth listening to he would have quit the day they signed Lucic.

 

[Menzinger]

Does cap management fall under the analytics umbrella ?

If so, then despite the production of McDavid and Draisaitl finishing 2 & 4 in NHL scoring combining for 221 points and their individual analytics through the roof, the impact of their combined cap hits of $21 mil and the impact on the rest of the players around them is most influential in the teams overall demise and 7th last in standings.

Analytics match individual performances but team metrics in regards to cap management seems to be the problem.

Hopefully that +$50 mil invested in more analytics is going to help Dubas solve this dilemma when he is faced with 3 players with great individual analytics consuming too much salary cap %, so the Leafs team doesn't follow the Oilers down the drain.

Dubas will need to use these individual analytics to find cheap contracts with strong supporting corsi metrics to surround a team with limited cap space.

The McDavid and Draisaitl contracts are no where even near the top 10 reasons why the Oilers are a disaster of an organization. Not even being close to the top 10

 

[Menzinger]

If the Oilers had a stats guy worth listening to he would have quit the day they signed Lucic.

Things the stats community has overwhelmingly hated:

Russell and Lucic signings, the Hall trade, the Eberle trade, the reinhart trade....

Which makes it a fairly dubious claim above that analytics supposedly “built” the Oilers lol

 

[Martin Skoula]

Things the stats community has overwhelmingly hated:

Russell and Lucic signings, the Hall trade, the Eberle trade, the reinhart trade....

Which makes it a fairly dubious claim above that analytics supposedly “built” the Oilers lol

Well you see, stats are numbers... and the draft lottery uses numbers...

I'm sure you can see where I'm going with this, it's pretty obvious that Edmonton is an excel spreadsheet team.

 

[Notsince67]

So the skilled credit risk guys do have numbers that they use, and those numbers come from somewhere? And someone at some point analyzed other numbers to develop the guidelines and understand the relationships that the skilled guys now use? Of course.

Your post #452 is the hockey equivalent of the old-timey bankers saying "bah humbug, I know everything I need to know, I can tell the measure of a man/company by a handshake and a plant walk through so this stuff is useless" to modern (20th century) analysis and techniques

My this post is a little cocky. 20th century? The information is contextual and not metrics based. You cannot regress information that has no metric. I am giving you valuable insight from hard nosed experience from a long practitioner of stats (myself), yet you arrogantly dismiss it as behind the times. So typical of a lot of people these days who read an article and immediately designate themselves as experts in the field. You keep trying to automate behaviour and your next lesson will be on how the population's behaviour shifts through your own actions. Negative or anti selection biases are well known. Predictive models require constant calibration to account for behavioral shifts. You dont need to know this stuff because it is so 20th century.

 

[Notsince67]

This isn’t particularly true, there were a fair number of teams who beat them to the punch before they finally hired Tyler Dellows.

Not to mention, it was a common theme for a few teams to hire a stats guy (usually directed by ownership) and then have them be completely ignored by the rest of the front office. Stats writers have been by far the loudest critics of the Oilers the past few years (even during that year they made the playoffs). Chia is in record saying he doesn’t pay attention to them (remember that infamous Seguin trade video?)

Well chia was a piece of work...I'll give you that. Dellows was hired in 2014 and fired 2 years later. I'm guessing over the Taylor Hall trade but I dont know for sure.

 

[FreeBird]

There is a lot of evidence that mistaking work akin to a 12th grade Data Management project for "advanced analytics" and making decisions based on it isn't a smart move.

In your analogy, does the skilled risk guy look at numbers or judge based on solely applicant appearance ,charisma, and the curb appeal of their house?
Where do the skill guys numbers come from? How did they develop the benchmarks and guidelines that they use to make their skilled

You're underrating the infancy of sporting analytics relative to financial. Implementing analytics for hockey isn't at the stage of letting models make decisions for management, it's still about finding and gathering the information to help the skill guys make decisions rather than loaning money to person based on gut feel and the fact their father is "good people" and they wore their sunday best to the bank

So does this mean we will be much better on Paper

 

[4thline]

My this post is a little cocky. 20th century? The information is contextual and not metrics based. You cannot regress information that has no metric. I am giving you valuable insight from hard nosed experience from a long practitioner of stats (myself), yet you arrogantly dismiss it as behind the times. So typical of a lot of people these days who read an article and immediately designate themselves as experts in the field. You keep trying to automate behaviour and your next lesson will be on how the population's behaviour shifts through your own actions. Negative or anti selection biases are well known. Predictive models require constant calibration to account for behavioral shifts. You dont need to know this stuff because it is so 20th century.

I'm not dismissing your skill or as experience as behind the times, I'm doing the opposite. I'm saying that your perspective is ahead of where things stand in hockey, and is taking for granted foundational work in your field, work that hasn't happened yet in hockey. I'm not advocating for the automation of behavior, I'm arguing that there is a lot of investment required in hockey analytics to provide the information required for evidenced based and well thought decisions.

The analytics argument in hockey isn't
go fully automated based on models vs. rely on skilled analysis
it's
develop information to inform and allow better skilled analysis vs. continue working on hunches and assumptions

 

[JT AM da real deal]

As Dubie has stated himself we still are figured out what are the best metrics and how to explain things properly in context using advanced stats. For goodness sake we have not even figured out save % yet and what is on the goalie, the team, the other team, the rink, strength and quality of shot etc etc etc. This whole analytics things is still in its infancy stages of development. So that is why no one is getting rid of scouting departments. Hockey is about people. And as long as it is there will be people who can project better than others. Where things are now is analytics can point you to certain types of players that you want to recruit. But that does not mean you are going to win anything. I still go with winners. Guys who have won their entire lives whatever the level. Those guys will typically keep winning. And when you look at a total team how many guys have won all the time throughout their lives. That will point you in direction of a Cup. Accumulate as many winners as you can.