News Article: "Fun With Numbers" - Advanced Stats Talk Here

[MatchesMalone]

A goal-based analysis is very problematic when used for special teams. You will never have a stable, representative sample in a predictive time frame.

I saw you ask a question with your last post, so I will answer. AZ3 player comparison is your best bet for micro stats.

Oh sick. Thank you very much sir.

Yeah I've been noticing the problem with the goal-based stats. Been starting to look more at scoring chances and high-danger scoring chances.

I'm also working on learning about "expected goals for." Not just gonna start using it until I fully understand it.

 

[Peptic Balcers]

Oh sick. Thank you very much sir.

Yeah I've been noticing the problem with the goal-based stats. Been starting to look more at scoring chances and high-danger scoring chances.

I'm also working on learning about "expected goals for." Not just gonna start using it until I fully understand it.

Expected Goals are a better predictor of future scoring than Corsi, Goals

 

[Sens of Anarchy]

A goal-based analysis is very problematic when used for special teams. You will never have a stable, representative sample in a predictive time frame.

I saw you ask a question with your last post, so I will answer. AZ3 player comparison is your best bet for micro stats.

Any way to get the Az3 data on all the players?

 

[ijif]

Any way to get the Az3 data on all the players?

I don't think so. Whatever is on the tableau page is what is available, at least to my knowledge. To be honest, I thought they had every player. They have had every player I have looked up, but I have only looked at "star" players.

 

[NyQuil]

I like these zone exit graphs. This is one measurable area Ottawa has to improve in. Insisting on playing D who are far more comfortable going off the glass and out vs making a play to get it out is part of the problem.

Karlsson was generational when it came to zone exits. We were a little spoiled I think - it might have been his greatest talent as a defenceman.

 

[Sens of Anarchy]

I don't think so. Whatever is on the tableau page is what is available, at least to my knowledge. To be honest, I thought they had every player. They have had every player I have looked up, but I have only looked at "star" players.

They have quite a number of players . Did not verify for all. What I was able to do was a 1 vs 1 comparison over a variety of time periods. I found you can download that data from the 1 to 1 comparison but , what I was looking for was a way to download the data for all the players they have listed in one shot, not necessarily all the players.

 

[DrEasy]

Toronto's retrieval of those dump outs was much more successful. Almost like they were part of a plan vs desperation.

They also probably have better personnel to do the retrieval. Still, they were the outlier in terms of successful retrieval; nothing to base a team plan on IMO.

 

[ijif]

They have quite a number of players . Did not verify for all. What I was able to do was a 1 vs 1 comparison over a variety of time periods. I found you can download that data from the 1 to 1 comparison but , what I was looking for was a way to download the data for all the players they have listed in one shot, not necessarily all the players.

There is a download button on the top right. It says it will download the workbook, but you need the Tableau app, and I do not have it. I'm not sure if by workbook they mean the whole data set, but I would guess that is what they mean.
That is pretty much the full extent of my knowledge.

 

[Micklebot]

I hadn't seen that A3Z data before, seems interesting, but at first glance I'm a little skeptical of it's value, at least from what I've seen.

Using breakup rates per 60 and entries per 60 sounds good, but it's going to be skewed by whether or not the team is a high pace team or a lower pace team. Strong defensive teams will have fewer chances to create breakups per 60, so is it good to be higher or lower?

Idk, I think I prefer his work on pokemon similarity scores
Tableau Public

 

[Sens of Anarchy]

I hadn't seen that A3Z data before, seems interesting, but at first glance I'm a little skeptical of it's value, at least from what I've seen.

Using breakup rates per 60 and entries per 60 sounds good, but it's going to be skewed by whether or not the team is a high pace team or a lower pace team. Strong defensive teams will have fewer chances to create breakups per 60, so is it good to be higher or lower?

Idk, I think I prefer his work on pokemon similarity scores
Tableau Public

I like the defensive stats , especially the controlled exit stats. Possession is a big part of the game ...and measuring contribution .. Controlled exits are an increasingly important stat imo.

 

[Micklebot]

I like the defensive stats , especially the controlled exit stats. Possession is a big part of the game ...and measuring contribution .. Controlled exits are an increasingly important stat imo.

sure, but per 60 mins... Idk. If you're constantly hemming the opposition in their end, you'll never get a chance to break out of your own end and your per 60 numbers will suffer. I'd rather see something like success % that isn't dependent on how high pace the team is. It's not that the action being measured isn't important, it's that the way they are measuring it might not be providing the ability to derive useful insight.

 

[Sens of Anarchy]

sure, but per 60 mins... Idk. If you're constantly hemming the opposition in their end, you'll never get a chance to break out of your own end and your per 60 numbers will suffer. I'd rather see something like success % that isn't dependent on how high pace the team is. It's not that the action being measured isn't important, it's that the way they are measuring it might not be providing the ability to derive useful insight.

always room for improvement. I expect we will see some in these dzone measurements

 

[Do Make Say Think]

Karlsson was generational when it came to zone exits. We were a little spoiled I think - it might have been his greatest talent as a defenceman.

That and his ability to block a shot and start skating the other way, with the puck, seamlessly.

What a stud

 

[MatchesMalone]

sure, but per 60 mins... Idk. If you're constantly hemming the opposition in their end, you'll never get a chance to break out of your own end and your per 60 numbers will suffer. I'd rather see something like success % that isn't dependent on how high pace the team is. It's not that the action being measured isn't important, it's that the way they are measuring it might not be providing the ability to derive useful insight.

Good point that I hadn't considered.

 

[DrEasy]

Bump! The annual Ottawa Hockey Analytics Conference at Carleton is on Nov 15-16. But the program only lists the names of the people speaking, but not what they will be speaking of.

 

[Sens of Anarchy]

Lund is good people, not out to make it the be all, end all. Look forward to this.

Concur

 

[Sens of Anarchy]

Bump! The annual Ottawa Hockey Analytics Conference at Carleton is on Nov 15-16. But the program only lists the names of the people speaking, but not what they will be speaking of.

Do you know if Total Hockey Ratings (THoR) is still a thing that is being modeled?
Its interesting stuff. I have an extended background in data but more business data, metadata and content management so I am drawn to it. It does take an investment in time to really get a handle on whats available to collect and the tools to collect it. Just keeps ending up seeming like work.

 

[DrEasy]

Do you know if Total Hockey Ratings (THoR) is still a thing that is being modeled?
Its interesting stuff. I have an extended background in data but more business data, metadata and content management so I am drawn to it. It does take an investment in time to really get a handle on whats available to collect and the tools to collect it. Just keeps ending up seeming like work.

Doesn't look like it's been calculated the past few years, at least based on this link. Since it's Schuckers who apparently came up with the metric, you can go to the conference and ask him in person!

The workshops on Friday look interesting, since they talk about how to scrape NHL data, and how to use Tableau.