The Advanced Stats Thread Episode VI: RIP To Our Databases

[SA16]

Karlsson is a fraud and I recommend placing a large wager on the under of his point total next year.

 

[Bob Richards]

Fun fact: Jimmy Vesey has more ixGF and a higher ixGF/60 than Wild Bill.

Its settled. Wait for Bill to get his deal and match Jimmy cent for cent.

But 4 srs.....lol.

Gotta be the HDSC, no? I think Jimmy is actually pretty adept at registering those, at least relative to the rest of our team (not sure about the NHL as a whole).

 

[Mac n Gs]

This is kind of shocking. 5v5 or all situations? How does their ixFsh% compare?

Karlsson is an arbitration eligible RFA. Is he gonna get paid based off this? What do you do here? I think he's UFA eligible after the 19-20 season.

Two year bridge at $4.5m AAV?

That’s 5v5. Trying to analyze Vesey’s powerplay numbers would give me a stroke because he’s an awful powerplay player.

I forget what ixFSh% is, but Vesey is at 9.49 and Karlsson is at 7.78. They’re both at basically 12 iCF/60 (11.8 vs 12.3)

Vesey really isn’t as bad as people make it seem. He’s overused, but he’s fine as a 3rd or 4th liner. The whole media hype around him skewed people’s expecations and still does. I’m curious to see him under a new coach.

 

[silverfish]

That’s 5v5. Trying to analyze Vesey’s powerplay numbers would give me a stroke because he’s an awful powerplay player.

I forget what ixFSh% is, but Vesey is at 9.49 and Karlsson is at 7.78. They’re both at basically 12 iCF/60 (11.8 vs 12.3)

Vesey really isn’t as bad as people make it seem. He’s overused, but he’s fine as a 3rd or 4th liner. The whole media hype around him skewed people’s expecations and still does. I’m curious to see him under a new coach.

ixFSh% is based off Manny's xG model and is essentially the % chance a shot has at going in the net. Used as a complement to raw xG, basically "average danger".

 

[Mac n Gs]

ixFSh% is based off Manny's xG model and is essentially the % chance a shot has at going in the net. Used as a complement to raw xG, basically "average danger".

Oh sweet, thanks for that. Vesey is better at that than Karlsson too hes actually 7th in the NHL when filtering for 400min

Scanning through the list quickly basically shows me guys that are good around the net and guys who can beat dmen to getting shots in tight. It doesn’t make up for all of his flaws, but it does show that he at least does have value.

Side note: I’m really hoping we can catch Chicago being stupid with Saad since there were some rumors floated out earlier this year. Or, I wonder if you can get them to give up a boatload of assets to get rid of Seabrook

 

[Machinehead]

I'm learning how to use excel for work and I just ran some correlations.

Of the top 30 most used defensemen the last 4 years:

Rel CF% to TOI QoT - -0.07 (Almost nothing)
Rel CF% to CF% QoT - -0.21 (Crap)
Rel CF% to TOI QoC - -0.41 (Below Average)
Rel CF% to CF% QoC +0.50 (Moderate Correlation)

1)I was surprised by how well QoC performed. I think we've underestimated it.

2)CF% QoC > TOI QoC. CF% beat TOI for both. This confirms what I thought. Actual results > what coaches think they are seeing.

3)None of these even sniffed the +/-0.7 barrier of being a strong correlation. At the end of the day, the driving force behind a player's analytics is THE PLAYER. Everything else is context.

 

[silverfish]

Excel? Peasant.... We actually still use at Excel at work and it is a royal pain in my ass, on the other hand, the things we've figured out how to do in it, in terms of automated reporting since we refuse to move to R and the agency won't get Tableau for us... I mean... some of these formulas are stupid at this point.

 

[Machinehead]

Work gives us excel for free, and I ain't spending any money to get yelled at on HF

Very few people appreciate the work anyway.

 

[silverfish]

Work gives us excel for free, and I ain't spending any money to get yelled at on HF

Very few people appreciate the work anyway.

R and Python are free

 

[Machinehead]

R and Python are free

The only pythons I know live in Australia. One step at a time.

 

[silverfish]

I want to make a Zipay Twitter bot.

 

[SnowblindNYR]

I'm learning how to use excel for work and I just ran some correlations.

Of the top 30 most used defensemen the last 4 years:

Rel CF% to TOI QoT - -0.07 (Almost nothing)
Rel CF% to CF% QoT - -0.21 (Crap)
Rel CF% to TOI QoC - -0.41 (Below Average)
Rel CF% to CF% QoC +0.50 (Moderate Correlation)

1)I was surprised by how well QoC performed. I think we've underestimated it.

2)CF% QoC > TOI QoC. CF% beat TOI for both. This confirms what I thought. Actual results > what coaches think they are seeing.

3)None of these even sniffed the +/-0.7 barrier of being a strong correlation. At the end of the day, the driving force behind a player's analytics is THE PLAYER. Everything else is context.

You've never used excel before?

 

[SnowblindNYR]

Excel? Peasant.... We actually still use at Excel at work and it is a royal pain in my ass, on the other hand, the things we've figured out how to do in it, in terms of automated reporting since we refuse to move to R and the agency won't get Tableau for us... I mean... some of these formulas are stupid at this point.

Most companies use excel.

 

[Machinehead]

You've never used excel before?

Never really had a reason to

 

[silverfish]

Most companies use excel.

Doesn't make it right.

 

[SnowblindNYR]

Never really had a reason to

So what have you been using to do your advanced stats on?

 

[SnowblindNYR]

@silverfish I prefer to use R Squared over correlations because it's more intuitive what the number actually means. Thoughts? You think I'm missing out by not using correlation?

 

[Machinehead]

So what have you been using to do your advanced stats on?

I don't really do advanced stats. I just read the ones already available.

 

[silverfish]

@silverfish I prefer to use R Squared over correlations because it's more intuitive what the number actually means. Thoughts? You think I'm missing out by not using correlation?

I lean towards R^2 as well, but everything needs to be caveated. Correlation != causation, etc...

The main question I try to focus on is how well does one variable predict another.

 

[Machinehead]

I lean towards R^2 as well, but everything needs to be caveated. Correlation != causation, etc...

The main question I try to focus on is how well does one variable predict another.

Chris Kreider existing predicts me being happy.

 

[SnowblindNYR]

I don't really do advanced stats. I just read the ones already available.

If you need excel help, let me know.

 

[Bob Richards]

I want to make a Zipay Twitter bot.

Love bots like that. They can really come up with some brink-of-sanity type profound shit lol

 

[silverfish]

Love bots like that. They can really come up with some brink-of-sanity type profound **** lol

I think I'm squared up with the NWHL for the most part so this seriously may be my next project

 

[Irishguy42]

I want to make a Zipay Twitter bot.

Inspired by this?

 

[silverfish]

Inspired by this?

Oh my God, I didn't know this existed!