“What analytics taught me is your eyes and your mind are lying sons of bitches in the worst absolute way,” Dubas said.

Assistant General Manager, Toronto Maple Leafs

http://www.theglobeandmail.com/sports/h ... e23238104/" onclick="window.open(this.href);return false;

And remember, he only started there this past summer.

So, it probably isn't fair to blame him for the Leafs past woes.

Oh and he didn't reach this conclusion on his own.

There is a large body of empirical research behind this statement.

He has recommended this book Thinking: Fast and Slow, by Nobel Prize Winner Daniel Kahneman.

Out.

http://hockeysimplified.blogspot.com/" onclick="window.open(this.href);return false;

I pretty much disagree, with all due respect. Most things are really obvious if you watch a team all the time like most of us do.

Dubas will be out of a job long before the Leafs become relevant again. The Leafs foundation hasn't been good in quite some time....that's their issue..not analytics. Winning the lottery and getting Connor McDavid is the best possible analytics they could hope for.

NHLTIM wrote:Dubas will be out of a job long before the Leafs become relevant again. The Leafs foundation hasn't been good in quite some time....that's their issue..not analytics. Winning the lottery and getting Connor McDavid is the best possible analytics they could hope for.

I love someone who has sufficient strength in their convictions to make a prediction.

How are we defining "relevant?"

Point percentage?

Points?

Aren't the Leafs somewhat impervious to irrelevance no matter how bad they suck? They are always a topic of interest...

ledzeppelinfan1 wrote:I pretty much disagree, with all due respect. Most things are really obvious if you watch a team all the time like most of us do.

Totally agree. You can watch a player, knowing absolutely nothing about his stats, and form a pretty solid opinion on his worth as a productive hockey player

I think there is happy medium where the eye test and advanced statistics can coexist. However, if the Leafs want to just start drafting players based solely of advanced statistics, I fear the Leafs will somehow be an even larger train wreck than they are currently.

they should measure the net weight of all the jerseys tossed on the ice.

Or maybe the Leafs will end up like the Kings and Blackhawks who make extensive use of statistics. In fact, they may be the most sophisticated users of statistics in the NHL.

How many minutes of eye-ball observation are sufficient to draw a conclusion about a player?

Against how many different opponents?

At what level of play?

How recently?

The flip side of relying on the eye test is that, if that is your position, you cannot possibly have an opinion about a player that you have never seen play a sufficient amount, however that is defined (see above).

And, you can respectfully disagree that the Earth is round or that the Earth orbits the Sun (rather than vice versa).

Disagreeing about how our brains work is not dissimilar.

That would be kind of like disagreeing about how our kidneys work.

You are correct the Kings and Blackhawks do use advanced statistics extensively. These two organizations also have extensive and well developed amateur scouting and player development programs. As I said before I think there is a happy medium but relaying to heavily on one could be problematic.
Is there an advanced statistic that tells you how well a player with transition from what ever league they are in to the NHL?

Yeah I suspect most clubs refer to stats to back up their opinions on players, but in my lifetime anyway, it's never going to be used as much as visual judgment.

I don't have a problem with advanced stats at all...some of it I find useful. For the most part it's black and white with me. If the Leafs and Blues switched players and coaches, I think the Leafs would be pretty good. Talent and depth always wins out. I don't know for sure and I could be totally wrong but just for example, what do advanced stats say when comparing Bortuzzo and Cole? From an eyeball perspective Bortuzzo seems to be a much better player to me but I would think advanced stats say Cole is better but I honestly have no idea.
I do know that the Leafs are going to trade Phil Kessel in the offseason and will probably base some of their return on advanced stats but I know for a fact they aren't going to get anybody in return as good offensively as Phil is.
The Cole vs Bortuzzo is an honest question....I would just like to know if you have something that gives a clear edge to one of those guys being a better player and what you use to measure it?

barnburner wrote:Yeah I suspect most clubs refer to stats to back up their opinions on players, but in my lifetime anyway, it's never going to be used as much as visual judgment.

I agree with this conclusion. Stats are good backup information, to help analyse and make a "best guess" at a player's potential value to a team, in the absence of having seen him enough (virtually every player other than one that plays in a local league in the metro area of the scouting team).

But, they should NEVER be the ONLY source of decision-making fodder, and with any luck, shouldn't need to be the "major" influence on such decisions.

We need to be careful, I think, about false dichotomies and straw man arguments.

The numbers don't come from the ether.

For now, at least, they come from people actually watching the games and recording what occurs.

Who shot?

Who scored?

Who passed the puck to the guy that scored?

These are all based on visual observations.

So, eye-balls versus data is a false dichotomy.

But, what is being observed is being recorded in a fairly objective and reproducible fashion.

So, it's not some guy saying, "Wow, this guy is great!"

Instead, it's "This guy did such and such."

With the increased usage of stats in hockey expect the number of scouts to increase. In fact, it already is increasing.

But, instead of "send this guy out to look at this kid and give us his opinion" it's "send this guy out and observe these things, measure them, and record them."

Maybe there is someone who posts here who works for the Kings or Blackhawks who can share what they are doing, but this is what the publicly-available info indicates.

Re: NHLTim's question

What a great question.

How should we value a player and, as a result, compare one player against another?

Ian Cole played 812.0 minutes (please notice these are decimal minutes) for the Blues this season.

Robert Bortuzzo has played 62.5 minutes.

So, there is a sample size issue.

More generally, maybe we can come up with a general model.

For example,

Salary - low is good, high is less good

Age - low is bad, middle is good, high is bad; using established performance curves by age for NHL players.

Etc.

Interesting question.

Or we could just wait for BarnBurner to tell us.

Very stimulating...

Web, I definitely agree with you that advanced stats can be a great tool in cases where X player is somewhat unknown to a scout or GM. However, intangibles which hockey is chock full of, cant be measured even with even with basic hockey stats. Regardless, both will be used. To be honest, baseball is probably the best advanced stat sport due to the structure and definable and trackable situations. Hockey is so random that it's hard to fit everything into a framework that doesn't leave room for interpretation, error, and misunderstanding.

WebSant wrote:Re: NHLTim's question

What a great question.

How should we value a player and, as a result, compare one player against another?

Ian Cole played 812.0 minutes (please notice these are decimal minutes) for the Blues this season.

Robert Bortuzzo has played 62.5 minutes.

So, there is a sample size issue.

More generally, maybe we can come up with a general model.

For example,

Salary - low is good, high is less good

Age - low is bad, middle is good, high is bad; using established performance curves by age for NHL players.

Etc.

Interesting question.

Or we could just wait for BarnBurner to tell us.

Can't we use Bortuzzo's stats with the Pens?

NHLTIM wrote:

WebSant wrote:Re: NHLTim's question

What a great question.

How should we value a player and, as a result, compare one player against another?

Ian Cole played 812.0 minutes (please notice these are decimal minutes) for the Blues this season.

Robert Bortuzzo has played 62.5 minutes.

So, there is a sample size issue.

More generally, maybe we can come up with a general model.

For example,

Salary - low is good, high is less good

Age - low is bad, middle is good, high is bad; using established performance curves by age for NHL players.

Etc.

Interesting question.

Or we could just wait for BarnBurner to tell us.

Can't we use Bortuzzo's stats with the Pens?

That's where the whole thing gets fangled, how do you compare hockey advanced stats when a player goes from a team who plays one system to a team that plays another?

One team plays a stern dump and chase trap then another plays a run and gun system where the d is encouraged to join the rush...messes things up. It has to.

ledzeppelinfan1 wrote:Very stimulating...

Web, I definitely agree with you that advanced stats can be a great tool in cases where X player is somewhat unknown to a scout or GM. However, intangibles which hockey is chock full of, cant be measured even with even with basic hockey stats. Regardless, both will be used. To be honest, baseball is probably the best advanced stat sport due to the structure and definable and trackable situations. Hockey is so random that it's hard to fit everything into a framework that doesn't leave room for interpretation, error, and misunderstanding.

Nice post.

Several quick thoughts:

I hate the term "advanced stats." Even worse is "advanced analytics." ("Analytics" not even being a word according to the spell-checker here.)

There is only data.

And, if you have read my blog you know that most of the "advanced analytics" aren't worth the electricity that it took to calculate them.

I agree that baseball is an easier case.

The really hard part in hockey is trying to tease out the contribution of the individual player, with a highly dynamic game and only five skaters.

This has led some people down a rat-hole of QoC and QoT with ever decreasing situations (e.g., close) and ever decreasing sample sizes and ever increasing complexity of calculation.

If it was easy it wouldn't be fun.

NHLTIM wrote:

WebSant wrote:Re: NHLTim's question

What a great question.

How should we value a player and, as a result, compare one player against another?

Ian Cole played 812.0 minutes (please notice these are decimal minutes) for the Blues this season.

Robert Bortuzzo has played 62.5 minutes.

So, there is a sample size issue.

More generally, maybe we can come up with a general model.

For example,

Salary - low is good, high is less good

Age - low is bad, middle is good, high is bad; using established performance curves by age for NHL players.

Etc.

Interesting question.

Or we could just wait for BarnBurner to tell us.

Can't we use Bortuzzo's stats with the Pens?

I would.

But, I didn't see any of those games.

That wouldn't stop me.

It should stop some others here, based on what they have posted.

For defensemen I like, even strength goals allowed divided by even strength TOI, as a measure of defensive ability.

It's not perfect. Nothing is.

I'm certainly open to other thoughts.

Still noodling.

Great topic and great discussion! Really enjoying reading what everyone has to say and find myself agreeing with points made by both sides of the discussion.
I think Led Zeppelin brings up a good discussion point worth exploring (if I am interpreting correctly). How do we apply data when a player goes from one team to another or from one league to another? Does a statistic quantify this (legit question)?

WebSant wrote:

NHLTIM wrote:

WebSant wrote:Re: NHLTim's question

What a great question.

How should we value a player and, as a result, compare one player against another?

Ian Cole played 812.0 minutes (please notice these are decimal minutes) for the Blues this season.

Robert Bortuzzo has played 62.5 minutes.

So, there is a sample size issue.

More generally, maybe we can come up with a general model.

For example,

Salary - low is good, high is less good

Age - low is bad, middle is good, high is bad; using established performance curves by age for NHL players.

Etc.

Interesting question.

Or we could just wait for BarnBurner to tell us.

Can't we use Bortuzzo's stats with the Pens?

I would.

But, I didn't see any of those games.

That wouldn't stop me.

It should stop some others here, based on what they have posted.

For defensemen I like, even strength goals allowed divided by even strength TOI, as a measure of defensive ability.

It's not perfect. Nothing is.

I'm certainly open to other thoughts.

Still noodling.

If I remember correctly, the Blues have a system (advanced stats) in place that they put to use last off-season. It's why they switched around the d-man pairings to start the season. I remember Hitch or Army talking about that. One x-factor to consider, and it's too small of a sample size right now, but given a few more games, you can compare Jax's numbers with Bort or Cole and see if Jax got better or worse depending on who he's playing with. One of the other x-factors is the opponents level of talent and execution.

A quick and dirty answer to NHLTim's question:

Robert Bortuzzo Ian Cole
age 25 26
height 6'4" 6"1"
weight 215 219

shoots Right Left

salary $650,000 $900,000
status RFA after this season RFA after this season

last season: games played 54 46
last season: minutes played 763.58 653.5
last season: goals scored 0 3
last season: GF / min. ES (z-score) -0.5 0
last season: GA / min. ES (z-score) 0 -2.9

this season: games played 38 54
this season: minutes played 538.13 783.55
this season: goals scored 2 4
this season: GF / min. ES (z-score) -1.2 0.3
this season: GA /min. ES (z-score) -0.2 -2.8

But, of course it won't paste nice and pretty.

Darn.

Working on a blog post, where it will be nicer to look at.