Advanced Stats

Forum for Pittsburgh Penguins-related messages.

Moderators: Three Stars, dagny, pfim, netwolf

Advanced Stats

Postby longtimefan on Wed Jul 18, 2018 2:34 pm

These are legitimate questions. From what I read, the advanced stat community has some differences of opinion with traditional scouting. Or the eyeball test. I can remember when teams didn't release +/- stats to the public, using them only in contract negotiations. The stat line was G A Pts PIM. That was the entire story. +/- was somewhat revolutionary once released publicly. Suddenly, it was the way to judge a player's defensive value. Really though, decisions had to be made based on what the eyes told you.

Advanced stats have been around for a long time now. They become more and more sophisticated. They tell a story. They are very useful, and much better than in the past. So I consider them to be a valuable tool, albeit not the last word on any discussion. The Jack Johnson signing really blew the boards up on his advanced stats. Which fit the narrative. I understand that, but I still have seen little mention about his 1st pass completion %, which was 6th in the league. The bottom line, he was an advanced stats nightmare. The ironic thing is the othe LGP whipping boy currently is Letang. Who was off by any standard of the eyeball test, but an absolute darling of the advanced stats community.

I think the only proper way to evaluate is a hybrid approach. The two players I mentioned sort of point that out. Advanced stats tell a story, but only in the context of the team. You have to project how his talents will fit with the players he will be on the ice with in your lineup. In addition, they are human beings, not ratings for a simulated computer program. Meaning past performance is not necessarily indicative of future success. Look at HBK. The advanced stats bear out that the line just went poof!

I've been using naturalstattrick.com. Are there any other sites to recommend? The advanced stats certainly allow you to find trends, and see who is playing well with whom. At this time of the year, they can be used, along with the eyes, to plot a strategy for the season. The one that really struck me was Kessel having a CF% of 58.10 with Rust and 49.24 without. In a smaller sample size, Rust had a similar effect on Brassard 69.81 to 45.02. Rust? He was over 50CF% with I believe every forward on the team. He truly seems to be a swiss army knife. Obviously, he's playing LW with Kessel.

The other thing I found interesting is this: http://www.naturalstattrick.com/linesta ... &p4=0&p5=0

I don't know much about Markus Nutivarra, but he looks to have had the second most ice time with Jack Johnson last season after Savard. Johnson was a horrific 46.69CF% without Nutivarra, but did produce a 53.17 with him. We obviously don't have him, but it does seem the Jackets managed to find a partner he was effective with, at least according to the stats.

Stats are great, and they also lie. It's foolish to believe the are fool proof, and it's foolish to dismiss them as useless. I'd like to know how others view advanced stats, and would like to know where I can learn more about them.
longtimefan
ECHL'er
ECHL'er
 
Posts: 1,469
Joined: Mon Jan 29, 2007 7:33 pm

Re: Advanced Stats

Postby FLPensFan on Wed Jul 18, 2018 3:09 pm

In terms of good sites to look at, naturalstattrick is definitely one. A few others:

- http://www.corsica.hockey/
- https://www.hockey-reference.com/
- https://public.tableau.com/profile/bill.comeau#!/vizhome/SkaterComparisonToolv2/Dashboard1
- http://ownthepuck.blogspot.com/

If you have a subscription to The Athletic, there were two really good articles recently on advanced stats usage:

- https://theathletic.com/435131/2018/07/17/bourne-what-does-an-nhl-statistical-analyst-do-how-do-the-numbers-shape-strategy-and-where-is-this-all-going/
- I can't find the better one. I'll keep looking.
FLPensFan
AHL Hall of Famer
AHL Hall of Famer
 
Posts: 9,427
Joined: Sun Jan 29, 2006 1:30 pm
Location: South Florida

Re: Advanced Stats

Postby Daniel on Wed Jul 18, 2018 3:48 pm

longtimefan wrote:These are legitimate questions. From what I read, the advanced stat community has some differences of opinion with traditional scouting. Or the eyeball test. I can remember when teams didn't release +/- stats to the public, using them only in contract negotiations. The stat line was G A Pts PIM. That was the entire story. +/- was somewhat revolutionary once released publicly. Suddenly, it was the way to judge a player's defensive value. Really though, decisions had to be made based on what the eyes told you.

Advanced stats have been around for a long time now. They become more and more sophisticated. They tell a story. They are very useful, and much better than in the past. So I consider them to be a valuable tool, albeit not the last word on any discussion. The Jack Johnson signing really blew the boards up on his advanced stats. Which fit the narrative. I understand that, but I still have seen little mention about his 1st pass completion %, which was 6th in the league. The bottom line, he was an advanced stats nightmare. The ironic thing is the othe LGP whipping boy currently is Letang. Who was off by any standard of the eyeball test, but an absolute darling of the advanced stats community.

I think the only proper way to evaluate is a hybrid approach. The two players I mentioned sort of point that out. Advanced stats tell a story, but only in the context of the team. You have to project how his talents will fit with the players he will be on the ice with in your lineup. In addition, they are human beings, not ratings for a simulated computer program. Meaning past performance is not necessarily indicative of future success. Look at HBK. The advanced stats bear out that the line just went poof!

I've been using naturalstattrick.com. Are there any other sites to recommend? The advanced stats certainly allow you to find trends, and see who is playing well with whom. At this time of the year, they can be used, along with the eyes, to plot a strategy for the season. The one that really struck me was Kessel having a CF% of 58.10 with Rust and 49.24 without. In a smaller sample size, Rust had a similar effect on Brassard 69.81 to 45.02. Rust? He was over 50CF% with I believe every forward on the team. He truly seems to be a swiss army knife. Obviously, he's playing LW with Kessel.

The other thing I found interesting is this: http://www.naturalstattrick.com/linesta ... &p4=0&p5=0

I don't know much about Markus Nutivarra, but he looks to have had the second most ice time with Jack Johnson last season after Savard. Johnson was a horrific 46.69CF% without Nutivarra, but did produce a 53.17 with him. We obviously don't have him, but it does seem the Jackets managed to find a partner he was effective with, at least according to the stats.

Stats are great, and they also lie. It's foolish to believe the are fool proof, and it's foolish to dismiss them as useless. I'd like to know how others view advanced stats, and would like to know where I can learn more about them.


I think no matter what the stats say, it is still history and doesn't take emotion into it. I won't use this as an excuse for Jack Johnson, merely as an example, but what if his emotional state was the reason for his bad stats (because of his personal nonsense) rather than skill and talent?

Sometimes history can predict the future, but sometimes people grow and change and learn and improve.....making the historical stats moot.
Daniel
AHL'er
AHL'er
 
Posts: 4,032
Joined: Sat Jan 28, 2006 4:10 pm
Location: Fort Worth

Re: Advanced Stats

Postby longtimefan on Wed Jul 18, 2018 4:42 pm

Daniel wrote:
longtimefan wrote:These are legitimate questions. From what I read, the advanced stat community has some differences of opinion with traditional scouting. Or the eyeball test. I can remember when teams didn't release +/- stats to the public, using them only in contract negotiations. The stat line was G A Pts PIM. That was the entire story. +/- was somewhat revolutionary once released publicly. Suddenly, it was the way to judge a player's defensive value. Really though, decisions had to be made based on what the eyes told you.

Advanced stats have been around for a long time now. They become more and more sophisticated. They tell a story. They are very useful, and much better than in the past. So I consider them to be a valuable tool, albeit not the last word on any discussion. The Jack Johnson signing really blew the boards up on his advanced stats. Which fit the narrative. I understand that, but I still have seen little mention about his 1st pass completion %, which was 6th in the league. The bottom line, he was an advanced stats nightmare. The ironic thing is the othe LGP whipping boy currently is Letang. Who was off by any standard of the eyeball test, but an absolute darling of the advanced stats community.

I think the only proper way to evaluate is a hybrid approach. The two players I mentioned sort of point that out. Advanced stats tell a story, but only in the context of the team. You have to project how his talents will fit with the players he will be on the ice with in your lineup. In addition, they are human beings, not ratings for a simulated computer program. Meaning past performance is not necessarily indicative of future success. Look at HBK. The advanced stats bear out that the line just went poof!

I've been using naturalstattrick.com. Are there any other sites to recommend? The advanced stats certainly allow you to find trends, and see who is playing well with whom. At this time of the year, they can be used, along with the eyes, to plot a strategy for the season. The one that really struck me was Kessel having a CF% of 58.10 with Rust and 49.24 without. In a smaller sample size, Rust had a similar effect on Brassard 69.81 to 45.02. Rust? He was over 50CF% with I believe every forward on the team. He truly seems to be a swiss army knife. Obviously, he's playing LW with Kessel.

The other thing I found interesting is this: http://www.naturalstattrick.com/linesta ... &p4=0&p5=0

I don't know much about Markus Nutivarra, but he looks to have had the second most ice time with Jack Johnson last season after Savard. Johnson was a horrific 46.69CF% without Nutivarra, but did produce a 53.17 with him. We obviously don't have him, but it does seem the Jackets managed to find a partner he was effective with, at least according to the stats.

Stats are great, and they also lie. It's foolish to believe the are fool proof, and it's foolish to dismiss them as useless. I'd like to know how others view advanced stats, and would like to know where I can learn more about them.


I think no matter what the stats say, it is still history and doesn't take emotion into it. I won't use this as an excuse for Jack Johnson, merely as an example, but what if his emotional state was the reason for his bad stats (because of his personal nonsense) rather than skill and talent?

Sometimes history can predict the future, but sometimes people grow and change and learn and improve.....making the historical stats moot.


Hence why they tell a story, but not the whole story. They are tools to use in search of combinations that work. They don't take into account emotion, leadership, clutch etc. As a stat, look at hits. Oleksiak was 4th on the team in hits, and Jake Guentzel led the team in hits. Somehow, those hits aren't created equal. It doesn't explain how Rust has 16 playoff goals, and 7 in decisive games. They do help formulate strategy though. They are far from perfect, but that' true of the eye test too. I remember Harry Sinden was frustrated when he traded for Kevin Stevens, and found him not to be the 50 goal scorer he expected. He didn't consider that Mario wasn't coming too.
longtimefan
ECHL'er
ECHL'er
 
Posts: 1,469
Joined: Mon Jan 29, 2007 7:33 pm

Re: Advanced Stats

Postby longtimefan on Wed Jul 18, 2018 4:44 pm

FLPensFan wrote:In terms of good sites to look at, naturalstattrick is definitely one. A few others:

- http://www.corsica.hockey/
- https://www.hockey-reference.com/
- https://public.tableau.com/profile/bill.comeau#!/vizhome/SkaterComparisonToolv2/Dashboard1
- http://ownthepuck.blogspot.com/

If you have a subscription to The Athletic, there were two really good articles recently on advanced stats usage:

- https://theathletic.com/435131/2018/07/17/bourne-what-does-an-nhl-statistical-analyst-do-how-do-the-numbers-shape-strategy-and-where-is-this-all-going/
- I can't find the better one. I'll keep looking.

Thank you. I've found all of those, but was wondering where to find the advanced stats about first pass completion percentage etc. Obviously, that's going to any further depth.
longtimefan
ECHL'er
ECHL'er
 
Posts: 1,469
Joined: Mon Jan 29, 2007 7:33 pm

Re: Advanced Stats

Postby FLPensFan on Wed Jul 18, 2018 6:01 pm

longtimefan wrote:
FLPensFan wrote:In terms of good sites to look at, naturalstattrick is definitely one. A few others:

- http://www.corsica.hockey/
- https://www.hockey-reference.com/
- https://public.tableau.com/profile/bill.comeau#!/vizhome/SkaterComparisonToolv2/Dashboard1
- http://ownthepuck.blogspot.com/

If you have a subscription to The Athletic, there were two really good articles recently on advanced stats usage:

- https://theathletic.com/435131/2018/07/17/bourne-what-does-an-nhl-statistical-analyst-do-how-do-the-numbers-shape-strategy-and-where-is-this-all-going/
- I can't find the better one. I'll keep looking.

Thank you. I've found all of those, but was wondering where to find the advanced stats about first pass completion percentage etc. Obviously, that's going to any further depth.

I'm not sure if that exists right now. I know there are several people out there working on passing projects, zone entry/exit projects, but not sure how far along they are.

EDIT1: As soon as I started writing this reply, I remembered who wrote the 2 Athletic aricles. They are from Ryan Stimpson, who has been around for awhile in the advanced stats community and was just hired by the Athletic this summer. He did a 2 part series on advanced stats (the first part co-piece with naturalstattrick):

Part 1: https://theathletic.com/415611/2018/07/05/an-advanced-stats-primer-with-naturalstattricks-brad-timmins/
Part 2: https://theathletic.com/408421/2018/07/06/advanced-stats-primer-part-2-why-does-passing-matter/

Consequently, Ryan is also one of the main guys working on the passing project with Corey Sznajder. Ryan published a link to his tableau workbook that shows some of his work, including stats on passing:

https://public.tableau.com/profile/ryan.stimson#!/vizhome/PlayerPassing/ComparisonDashboard
FLPensFan
AHL Hall of Famer
AHL Hall of Famer
 
Posts: 9,427
Joined: Sun Jan 29, 2006 1:30 pm
Location: South Florida

Re: Advanced Stats

Postby Jesse on Thu Jul 19, 2018 6:58 pm

My advice to you would be expand your advanced metrics beyond puck-possession. Look at deployment. Where did the player start their shifts: OZ, NZ, or DZ? Who did they play with? Who did they play against? How was their performance relative to the rest of the team? How was their performance in expected-goal shares? Did they generate them/prevent them really well? What about scoring chances instead of expected-goals? Any differential there?

Make an expansive case and then do what you did here; supplement it with video. It's one thing to talk about what the data says, it's another thing to show on-ice examples of why the data might be presenting itself in such a way. I don't think Corsi is a bad stat, I just think it's very bland and there have been developments that have given us more predictive data. Expected goals, for example, give us a much better idea about shot quality. While there's still a long way to go before we can nail down shot quality more, it's the best tool we have for it right now.

Advanced metrics are like building a puzzle. You can't slap down 20 pieces of a 100 piece puzzle and say you've figured out the big picture just because of the 20 pieces you used. Use all of them. Look at every metric available. I think that, more often than not, you'll find that it lines up pretty well with the video analysis. There's a decent marriage between on-ice results and on-ice strategy.
Jesse
NHL Healthy Scratch
NHL Healthy Scratch
 
Posts: 10,660
Joined: Tue Jan 31, 2006 9:07 am
Location: The Land of Condescending Posters

Re: Advanced Stats

Postby OKpensfan247 on Sat Jul 21, 2018 12:27 pm

Great OP longtimefan, and some great responses in here.

Jesse - would love to hear your thoughts on starting forward lines. Do you have an opinion on Fwd lines yet?

Ive been banging the Guentzel-Sheahan-Kessel 3rd line idea on here, but the more I analyze their 5-on-5 corsi, they hovered around high 40%. So not as strong as my eyeballs thought they were seeing as far as possession although we won most of our games and they recorded a lot of points in Jan-Feb (like 7 games together).

You make a great point on analyzing the entire puzzle. My advanced stats knowledge is reliant on corsi and zone starts currently.

A few observations from naturalstat website (thx FLPF):

- Sheahan's best corsi was with Hags and Rust.
- But Hags and Geno have outstanding corsi together, so I think Hags sticking with Geno makes a lot of sense.
- As many have pointed out, Simon has outstanding corsi with Sid... But also Brassard in much less sample size.
- Hornqvist has good corsi with Sid and Geno.
- Surprised to see Rust had good corsi with Sid.
OKpensfan247
ECHL'er
ECHL'er
 
Posts: 685
Joined: Sun Apr 22, 2012 10:51 am
Location: Tulsa, OK


Return to Pittsburgh Penguins

Who is online

Users browsing this forum: No registered users and 9 guests


e-mail