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

[DrEasy]

Final round:

Tom Awad

Modern era of hockey stats started in 2007 with availability of real time stats: led to Corsi/Fenwick, Zone starts, QoC, QoT, etc. By now, most of lowest hanging fruit has been picked.

Still, stats more of descriptive nature than prescriptive. Still can't answer Taylor Hall's question: "How can I improve my Corsi?"

Recent progress due to manual tracking.

Unsolved mysteries:

- score effects: teams play differently depending on the score (number of shots taken changes based on score, etc) but we don't know why at the micro level.

- contribs of players on the ice (WOWY), still can't see who screws up etc just by looking at current data.

SportVu will give us data we need: defensive positioning, zone entry formation, etc.

---

Timo Seppa

Tracked microstats data for local NCAA team. noticed that zone entry numbers don't match eye-assessed quality of D, and that's because of the defensive scheme they were running. RD stopped entries but because LD was backing off to retrieve dump ins. So D partner has impact on "targeted" D.

---

Michael Schuckers

Nemisz picked 25th (Calgary), Eberle picked 22nd (Edm), yet Central Scouting ranked Nemisz higher in skating.

Teams outdo central scouting up to 180 overall draft in career TOI.

So teams do better, are their scouts worth the money? Answer is yes: Teams get $2M value from their scouts. Scouts value comes from 2nd and 3rd rounds.

Central Scouting does better than teams in late rounds.

---

panel: question to Schuckers: could it be that teams are invested in players they draft early and so give them bigger chance to succeed with more TOI?

other question: shift in strategy where teams take more chances late in the draft, and that can also skew the results.

Timo's answer: based on limited experience with one nhl team, early rounds were consensus based but 3rd round and onward is where they relied more on scouts.

question: which new stats get your stats juices flowing? Awad: Zone entry analysis. Schuckers: upcoming player tracking data is what he'd love to have, but for now zone entry as well. Vollmer: compared to a few years ago, new stats come at you faster; need time to wrap your mind around them. Timo: draft analytics, he'd like to do this for a team, where it counts!

===

Aaaaaand that's a wrap.

 

[StefanW]

Final round:

Tom Awad

Modern era of hockey stats started in 2007 with availability of real time stats: led to Corsi/Fenwick, Zone starts, QoC, QoT, etc. By now, most of lowest hanging fruit has been picked.

Still, stats more of descriptive nature than prescriptive. Still can't answer Taylor Hall's question: "How can I improve my Corsi?"

Recent progress due to manual tracking.

Unsolved mysteries:

- score effects: teams play differently depending on the score (number of shots taken changes based on score, etc) but we don't know why at the micro level.

- contribs of players on the ice (WOWY), still can't see who screws up etc just by looking at current data.

SportVu will give us data we need: defensive positioning, zone entry formation, etc.

---

Timo Seppa

Tracked microstats data for local NCAA team. noticed that zone entry numbers don't match eye-assessed quality of D, and that's because of the defensive scheme they were running. RD stopped entries but because LD was backing off to retrieve dump ins. So D partner has impact on "targeted" D.

---

Michael Schuckers

Nemisz picked 25th (Calgary), Eberle picked 22nd (Edm), yet Central Scouting ranked Nemisz higher in skating.

Teams outdo central scouting up to 180 overall draft in career TOI.

So teams do better, are their scouts worth the money? Answer is yes: Teams get $2M value from their scouts. Scouts value comes from 2nd and 3rd rounds.

Central Scouting does better than teams in late rounds.

---

panel: question to Schuckers: could it be that teams are invested in players they draft early and so give them bigger chance to succeed with more TOI?

other question: shift in strategy where teams take more chances late in the draft, and that can also skew the results.

Timo's answer: based on limited experience with one nhl team, early rounds were consensus based but 3rd round and onward is where they relied more on scouts.

question: which new stats get your stats juices flowing? Awad: Zone entry analysis. Schuckers: upcoming player tracking data is what he'd love to have, but for now zone entry as well. Vollmer: compared to a few years ago, new stats come at you faster; need time to wrap your mind around them. Timo: draft analytics, he'd like to do this for a team, where it counts!

===

Aaaaaand that's a wrap.

Good writeup! Thanks for taking the time to do this.

I wish I knew you were there, I would have kept an eye out for you and said hi. I was the one who asked the question about draft pick quality dropping and teams more likely to take chances in later draft rounds.

IMO the conference itself was excellent. I've been to many non-hockey conferences in my life, and very few have matched this in terms of top to bottom quality. It was exceptionally well done.

 

[silverfish]

Great recaps. Thanks a bunch!

If they do another one in DC, I'm hoping to get down there for it.

 

[StefanW]

Great recaps. Thanks a bunch!

If they do another one in DC, I'm hoping to get down there for it.

My understanding is that it will probably happen.

 

[DrEasy]

Good writeup! Thanks for taking the time to do this.

I wish I knew you were there, I would have kept an eye out for you and said hi. I was the one who asked the question about draft pick quality dropping and teams more likely to take chances in later draft rounds.

IMO the conference itself was excellent. I've been to many non-hockey conferences in my life, and very few have matched this in terms of top to bottom quality. It was exceptionally well done.

Thanks for the nice words, everyone! I hope I didn't miss or misquote anything important.

I couldn't see you from where I was, that is at the other end of the classroom. That was an excellent question that you asked BTW! It would explain why Central Scouting does better in later rounds with diminishing returns.

I agree, it was a nice little conference. I wish they would have gone a little more into the gory math details, but I understand that they had a diverse crowd in front of them. Was there an official Call for Papers for this conference, or did they just assemble a bunch of speakers with invitations? Is there a proceedings available anywhere? It still very much feels like the early days of hockey analytics, so much to do to organize the community a little better as it grows.

I wonder if the Sens sent anyone there. I also bet Yost was there, but I don't know what he looks like.

When and where is the next one?

 

[Do Make Say Think]

Glad to hear people were happy with the event

I really wish I could have made it

 

[StefanW]

Thanks for the nice words, everyone! I hope I didn't miss or misquote anything important.

I couldn't see you from where I was, that is at the other end of the classroom. That was an excellent question that you asked BTW! It would explain why Central Scouting does better in later rounds with diminishing returns.

I agree, it was a nice little conference. I wish they would have gone a little more into the gory math details, but I understand that they had a diverse crowd in front of them. Was there an official Call for Papers for this conference, or did they just assemble a bunch of speakers with invitations? Is there a proceedings available anywhere? It still very much feels like the early days of hockey analytics, so much to do to organize the community a little better as it grows.

I wonder if the Sens sent anyone there. I also bet Yost was there, but I don't know what he looks like.

When and where is the next one?

There was no call for papers, the organizers invited individuals to speak. And it was a good list of presenters so good on them.

I don't know if another one is formally in the works. We are seeing the first ones popping up in cities like Calgary, Pittsburg, Ottawa, and soon Washington. It remains to be seen how many of them repeat. I hope they all do.

 

[StefanW]

Last post that ties it all together is now up at www.integratinghockeyanalysis.com It has been a fun almost-year since I started. From now until the blog runs out in 4-6 weeks I will be working on my tank-a-lytics post that I started a while back (look for the tab at the top left on the main page to find it).

Thanks to everyone who supported this project by visiting and commenting!

 

[18Hossa]

Last post that ties it all together is now up at www.integratinghockeyanalysis.com It has been a fun almost-year since I started. From now until the blog runs out in 4-6 weeks I will be working on my tank-a-lytics post that I started a while back (look for the tab at the top left on the main page to find it).

Thanks to everyone who supported this project by visiting and commenting!

Awesome job!!!

 

[starling]

Missed it unfortunately due to family plans. Thanks for the write up.

I wonder if the Sens sent anyone there. I also bet Yost was there, but I don't know what he looks like.

Yost lives in Vegas I believe, doubt he went there.
Jamie McLennan (TSN color commentator) mentioned on broadcast he was there apparently.

 

[DrEasy]

The conference web site now has links to various write-ups. I found this one to be the most complete. But comparing to my own quick notes it doesn't look like I missed much.

 

[chipsens]

NHL enhanced stats (Sens)

The NHL just released a new page on their website that includes shot-attempts (SAT aka Corsi) & unblocked shot-attempts (USAT aka Fenwick). These are also shown for even-strength and ahead, behind re lead in games.

http://www.nhl.com/stats/advancedst...mptsEven&viewName=advancedStatsSkaterShooting

These stats shed some interesting new info about player performance...such as:

1) Wiercoch leads the team in SAT & USAT...BigRig is last.
2) Ryan is surprisingly low on shot-attempts at even-strength...likely due to his broken finger!?
3) ...(add your insight from enhanced stats here)

Just to stir the pot...

 

[BonkTastic]

Just to stir the pot...

Here's something that will surely stir the pot:

1) before anyone points at Phillips's team worst USAT/Fenwick, I'd like to point out that Bobby Ryan has the 2nd worst spot on the team in that regard.

2) That thing I just pointed out about USAT/Fenwick? It's the same for SAT/Corsi: Phillips last, Ryan 2nd last.

3) It goes without saying that stats like these two (Sat/Corsi, USAT/Fenwick), stats that do not take into account player usage/deployment/quality of minutes and, is slanted heavily towards players whose usage/deployment/quality of minutes is heavily managed. The three Senators whose minutes have been managed the most this season - Wier, Lazar and Hoffman - are all in the top 3 in both categories. Zero surprises.

 

[chipsens]

Here's something that will surely stir the pot:

1) before anyone points at Phillips's team worst USAT/Fenwick, I'd like to point out that Bobby Ryan has the 2nd worst spot on the team in that regard.

2) That thing I just pointed out about USAT/Fenwick? It's the same for SAT/Corsi: Phillips last, Ryan 2nd last.

3) It goes without saying that stats like these two (Sat/Corsi, USAT/Fenwick), stats that do not take into account player usage/deployment/quality of minutes and, is slanted heavily towards players whose usage/deployment/quality of minutes is heavily managed. The three Senators whose minutes have been managed the most this season - Wier, Lazar and Hoffman - are all in the top 3 in both categories. Zero surprises.

Damn lies & statistics, eh?! What you say is true in terms of player usage being different...however stats don't lie. The interpretation of those stats is what's key...including other qualitative info about how different players were used (Ryan vs Condra), or other key info (injuries affecting play, eg Ryan's broken finger means he is shooting less than you might expect).

So IMO, let's not throw the Fenwick baby out with the bath water there's gold in them thar stats! Note: Ryan is not 2nd last in Fenwick if you adjust for amount of time on ice (change "Report" on drop-down box to "skater shooting / time on ice".

http://www.nhl.com/stats/advancedst...pts60&viewName=advancedStatsSkaterShootingTOI

 

[chipsens]

Sample size matters. Prince leads all Sens in unblocked shot attempt, adjusted for time on-ice (USAT/60) but he has only played 2 games, similarly Puempel is relatively low, based on a single game. Most other players have at least 30 GP, which enuf data for a reasonable sample size (approaches enuf data for normal distribution, aka bell curve).

 

[chipsens]

Despite scoring only 1 goal in 20 GP, Greening is actually fairly high in terms of USAT/60...which suggests he is a good player when it comes to puck possession. He may not be a scoring threat, but perhaps he can be packaged as a defensive specialist?!...or simply thrown out with the Bingo bathwater

 

[BonkTastic]

So IMO, let's not throw the Fenwick baby out with the bath water there's gold in them thar stats! Note: Ryan is not 2nd last in Fenwick if you adjust for amount of time on ice (change "Report" on drop-down box to "skater shooting / time on ice".

Chip, I have to say you're doing a good job of preventing me from getting anyone's jimmies rustled here. Well done.

 

[Vesa Awesaka]

The NHL just released a new page on their website that includes shot-attempts (SAT aka Corsi) & unblocked shot-attempts (USAT aka Fenwick). These are also shown for even-strength and ahead, behind re lead in games.

http://www.nhl.com/stats/advancedst...mptsEven&viewName=advancedStatsSkaterShooting

These stats shed some interesting new info about player performance...such as:

1) Wiercoch leads the team in SAT & USAT...BigRig is last.
2) Ryan is surprisingly low on shot-attempts at even-strength...likely due to his broken finger!?
3) ...(add your insight from enhanced stats here)

Just to stir the pot...

Wier is good because hes used more in offensive situations against weak opponents. I would assume he doesnt have to transition out of his zone as much as phillips had to or at least it was easier for him.

Wier is good in offensive situations against weak competition. This is something people have been saying a long time. He dominates in those situations. In general though i would only consider SAT part of the puzzle... deployment, linemates and opponents are also huge factors.

Worth noting that the team played terrible under maclean for all these stats. I imagine a lot of the core players were dug into deep holes they only started to pull out of under Cameron

 

[Micklebot]

Here's something that will surely stir the pot:

1) before anyone points at Phillips's team worst USAT/Fenwick, I'd like to point out that Bobby Ryan has the 2nd worst spot on the team in that regard.

2) That thing I just pointed out about USAT/Fenwick? It's the same for SAT/Corsi: Phillips last, Ryan 2nd last.

3) It goes without saying that stats like these two (Sat/Corsi, USAT/Fenwick), stats that do not take into account player usage/deployment/quality of minutes and, is slanted heavily towards players whose usage/deployment/quality of minutes is heavily managed. The three Senators whose minutes have been managed the most this season - Wier, Lazar and Hoffman - are all in the top 3 in both categories. Zero surprises.

The question is how much does managing their mins affect their ability to drive shot based stats?

A difference between 45% OZ starts (Gryba) and 55% OZ starts (Wiercioch) is one extra start in the OZ, and one less start in the DZ for Wiercioch per game. Quality of Competition stats might make a small impact, but the reality is because the opposing team want's to take advantage of weaker players just as much as you want to shelter them, the difference is less than you might imagine.

I'd also point out that since guys like Phillips, Cowen, and Gryba are far more prone to icing the puck, the are the architechs of their own DZ starts.

Now, where I do see some unexplored potential for how guys are sheltered, is how many of those DZ starts are against touch competition and vice versa?

In the end, I think Wiercioch (for example) would still have good corsi numbers and Gryba/Phillips bad ones even if their roles were reversed (though maybe a little less extreme), the issue with a guy like Wiercioch is all the stuff that doesn't make it into those numbers.

 

[Flamingo]

That's an interesting point about d-zone starts, R2-beer2. I wonder if icing face-offs are discounted. They should be.

 

[starling]

Two words: Quality of Competition.
Just throwing SAT/USAT out there does absolutely nothing, it's even more meaningless than +/-
Sadly, NHL.COM is yet to understand how to work with advanced stats. There are way better websites for that: war-on-ice.com, behindthenet.ca, etc.

 

[Micklebot]

That's an interesting point about d-zone starts, R2-beer2. I wonder if icing face-offs are discounted. They should be.

Well, if the goal is to know how a coach wants to deploy guys, discount icing faceoffs. If the goal is to know what situations they actually played in, they need to stay.

The reality is icing generally leads to the toughest deployment you'll see: tired group, in the DZ, against the oppositions top choice of attackers, so there is value in knowing who got those starts.

 

[God Says No]

Phillips, Gryba and Cowen are the kings of icings. Is there icings per 60 min? (I know there isn't. Joking)

 

[Icelevel]

Phillips, Neil, Smith, Gryba

are the guys I would first like to see replaced/moved out. More or less consistent with the bottom of the pile advanced stats wise. Course these guys are some of our more physical players which obviously has value not represented totally in shot attempts.

 

[Micklebot]

Phillips, Neil, Smith, Gryba

are the guys I would first like to see replaced/moved out. More or less consistent with the bottom of the pile advanced stats wise. Course these guys are some of our more physical players which obviously has value not represented totally in shot attempts.

I think Smith has traditionally done fairly well in terms of adv stats.

While his corsi numbers are down this year, he's generally been above 50% (50.8% over his career). He's done so while often being used in a shutdown roll to, starting under 42% of his shifts in the OZ.

He's also been stuck with a declining Neil quite a lot.

My biggest issue with smith is he's been terrible at the dot this year (not sure why), which has to change if he's to be effective for us.

All that said, I'd rather have Pageau in the lineup.