A Metric for individual possession

[Micklebot]

So, I was thinking about players that always seem to have the puck on their stick, and decided to try and see if I could come up with a way to measure it.

So far, I've taken individual Corsi + takeaways + giveaway as they are all indicative of a instance in time where the puck was on your stick.

Here's the top 10 Dmen ranked by touches/60, with 700+ mins this season:

Rank | Name | Touches/60
1 | Brent.Burns | 23.24
2 | Erik.Karlsson | 18
3 | Dustin.Byfuglien | 17.16
4 | Jeff.Petry | 17.14
5 | Jake.Muzzin | 16.64
6 | Victor.Hedman | 16.62
7 | Torey.Krug | 16.48
8 | Johnny.Boychuk | 16.45
9 | Morgan.Rielly | 16.38
10 | Kris.Letang | 16.03

Here's the bottom 10 Dmen

Rank | Player |Touches/60
10 | Paul.Martin | 7.69
9 | Adam.Larsson | 7.62
8 | Willie.Mitchell | 7.41
7 | Tobias.Enstrom | 7.03
6 | Ryan.Murray | 6.73
5 | Andrew.Greene | 6.55
4 | Christopher.Tanev | 6.14
3 | Mark.Borowiecki | 5.97
2 | Fedor.Tyutin | 5.46
1 | Niklas.Kronwall | 5.28

Rank | Player | Touches/60
1 | Alex.Ovechkin | 25.78
2 | Jason.Spezza | 24.34
3 | Nazem.Kadri | 23.59
4 | Max.Pacioretty | 23.58
5 | Jeff.Skinner | 23.37
6 | Rick.Nash | 23.36
7 | Vladimir.Tarasenko | 23.26
8 | Brent.Burns | 23.24
9 | Dustin.Brown | 22.6
10 | Filip.Forsberg | 22.54

Rank | Player | Touches/60
10 | Paul.Stastny | 9.99
9 | Chris.Vandevelde | 9.86
8 | Valtteri.Filppula | 9.81
7 | Mike.Ribeiro | 9.37
6 | Henrik.Sedin | 9.32
5 | Stephen.Gionta | 9.27
4 | Jori.Lehtera | 8.96
3 | Cody.McLeod | 8.33
2 | Alex.Tanguay | 7.71
1 | Jarret.Stoll | 6.11

Thoughts on whether this passes the eye test? or potential improvements?

My initial thought is it doesn't seem to work consistently with the forwards.

Click to read more...

 

[Johnny Engine]

Is that Corsi for, or Corsi +/- you're using?

A few issues:
- Giveaway and takeaway stats are unreliable and vary from arena to arena. But if you're aware of those limitations, it's still fun to play around with.
- If you're looking to quantify "touches" the way someone viewing the game would, this doesn't entirely work. A player could steal the puck, direct a shot at the net, and have it blocked resulting in a turnover, and be credited with 3 puck touches, when they really had it once. And a player could tick-tack-toe it around with his linemates for ages and never get credited with anything. As long as you're aware that this is something of a proxy, a "who's doing stuff on the ice" stat, then you're fine.

Edit: regarding the second point, I see that many of the worst-performing forwards are pass-first guys. You've set up a situation where you lose points when you pass the puck off successfully. You're going to have to address this to make this a useful metric.

 

[Michael Farkas]

Interesting. As an "eye test" guy, I can say that your "top" group of d-men does feature a lot of three-zone puck carriers...so there's that...that's a plus for it.

And your "bottom" group of d-men does feature a non-NHL caliber player in Borowiecki - who for some reason refuses to ever put a second hand on his stick, see if you can measure that...I bet he has one hand on his stick more than any non-goalie in the league...and he has no puck skills anyway, so it doesn't really matter...Mitchell and Tyutin are not very talented and often stay out of the way on the breakout...

Then on the flip side, you have a guy like Paul Martin who is a three-zone carrier but has one of the league's weakest shots. These shooting statistics, as I think is obvious, are very biased in favor of players that shoot from anywhere - whether it's the best play or not - so as a result, you end up with guys like Martin and Larsson and Greene on there, all heady players without big shots and guys that tend to not make mistakes or take a lot of poor shots...

Similarly, you end up with just shooters (and Spezza, somehow) on the forward side (and Brent Burns twice, btw). And a ton of pure setup guys and playmakers on the "bottom" group...H.Sedin, Paul Stastny, Filppula, Ribeiro, Lehtera...I mean, they have a lot in common in that they generally play with guys that shoot and they don't do it...even though some of those guys are actually two+ zone carriers, they defer once they cross the blueline, so they get no "action" according to Corsi, but they're the ones that made the whole play in a lot of cases...

Of course, you have donkeys like Chris VandeVelde - ubiquitous clod - and Stephen Gionta - glass-licking enthusiast - in there, so that makes sense. CVV never touches the puck it seems like unless it clicks off of him in front and Gionta's line never appears to even have a goal of shooting, they only want to go and cycle it to keep the other team from getting it...they cycle and never rotate the puck to the front it seems like...but not even because they can't, it literally looks like they don't want to, and their d-men don't join it, they just hold the line, so they never out-number...

I don't really know what you have to work with in total, but if you want to work all those playmaking centers off the bottom I might just add assists...it will keep fork-and-spoon operators like Stoll, CVV and Gio on the bottom but lift those very talented players out of the weeds and it won't damage your top group - I would surmise - if you weight it appropriately because like I said, you have almost nothing but pure goal scorers in the "top" group...

The interesting thing or test might be, and I have no idea what his underlying (or over-lying for that matter, I'm not a big stat watcher despite playing fantasy hockey haha) numbers look like, but find a way to elevate Nick Leddy...Leddy touches the puck a ton on the Islanders, but I don't know what "action" it leads to, I can't image he has a lot of points...but if tertiary assists were given out, I'd bet he'd be off the charts, same with Doughty for that matter...

 

[white galaxy*]

what about a chip in the puck and sensors on stick blades , pucks exact movement charted entire game by computer

 

[Johnny Engine]

...find a way to elevate Nick Leddy...Leddy touches the puck a ton on the Islanders, but I don't know what "action" it leads to, I can't image he has a lot of points...but if tertiary assists were given out, I'd bet he'd be off the charts, same with Doughty for that matter...

You'd want complete passing data to capture that (and to make this project work to its potential, for that matter).

At this point, I think if the NHL is going to bother with tracking giveaways and takeaways and whatnot, there's no reason why they shouldn't be able to fully divide up possession changes into a few categories, and make them balance each other:

- passes completed should balance out among teammates with passes received.
- Corsi events and Dump Ins (any time when a player willingly surrenders the puck, from chipping it off the glass, to point-men with no play, etc.) should balance out between teams with Pick ups, Saves and Goals.
- Stoppages would cover any loss of possession cause by a whistle - offsides, delayed penalties, the end of a period, etc.
- Giveaways should balance out between teams with Takeaways, and the definition of those terms should be any possession change not covered by the above.

I feel like that would remove some of the subjectivity from current real-time stats, and go a long way to that fabled "statistical mapping of the game."

The fact that the OP is missing the data for some of those factors limits what he's trying to do.

 

[Michael Farkas]

Oh of course...this type of data, to varying degrees, is tracked internally by some teams...but what the NHL provides is not really "advanced" stats...shot attempts, I mean, seriously, there's nothing "advanced" about that...that's the low-hanging fruit...but that's what the league provides...

Teams track this type of stuff, they break it down using film as opposed to this sterile, league-wide stuff...but these are the tools that micklebot and others have to work with, so I mean, you can only play the hand your dealt...

These guys sit here rubbing two sticks together unfortunately, the NHL clubs - with abundant resources - have a blow-torch. But like a scout, knowledge is your currency...same thing for these stat guys, the data is their currency...you can't just give it all a way...

 

[Ogopogo*]

Why the hell would you use Corsi if you want to have anything close to accurate? Do it like they do in Australian football and give players credit for touches. You want to know how much a guy possesses the puck - count how many times he actually has it on his stick. Not rocket science.

 

[Doctor No]

Why the hell would you use Corsi if you want to have anything close to accurate? Do it like they do in Australian football and give players credit for touches. You want to know how much a guy possesses the puck - count how many times he actually has it on his stick. Not rocket science.

Given the context of this thread, how do you intend for the original poster to do this with the data currently available TO THE PUBLIC?

It's never rocket science if you don't consider the difficult aspects.

 

[Michael Farkas]

I disagree, Doc...all micklebot has to do is watch and track every single player in every single game over the course of several seasons...it's not rocket surgery...

But seriously, I think micklebot is a Sens fan (maybe?)...it would be an interesting start to track something as simple as zone entries/exits and/or the amount of touches it takes to get out per player on the ice, or the amount of clean zone entries that yield a scoring chance or something to that degree. Junior teams do this, and they have much more limited resources than NHL clubs.

If he is a Sens fan, it would have been easy this year with the d-men, as there were only a couple that belonged in the NHL anyhow...

I could probably even give an estimate:
Ceci: 1.6 touches to get out on average, 36% clean zone entries
Karlsson: <1.6 touches to get out, the rest of the team's clean zone entries
...
Borowiecki: N/A (never gets out of the zone)
Cowen: N/A (never gets out of the zone)

 

[Doctor No]

Agreed - if you're trying to do this for one team, it would be tedious but not ridiculous. When I do similar exercises for goaltenders, it usually takes me about 2.5 times the length of the game to record the things that I record (and I imagine that if I had the time to do this more than a few times per year, it would get easier).

Given the size of these Boards, we could probably even crowd source something across the league - that's something Bill James started with "Project Scoresheet" in the 1980s. The data wasn't available, and so let's create the data.

Of course, then you run into problems of bias, where Mike sees a gray area as white, and Norm sees a gray area as black, and you end up with the New York Islanders having the top four guys in the NHL in hits or something.

 

[Micklebot]

Agreed - if you're trying to do this for one team, it would be tedious but not ridiculous. When I do similar exercises for goaltenders, it usually takes me about 2.5 times the length of the game to record the things that I record (and I imagine that if I had the time to do this more than a few times per year, it would get easier).

Given the size of these Boards, we could probably even crowd source something across the league - that's something Bill James started with "Project Scoresheet" in the 1980s. The data wasn't available, and so let's create the data.

Of course, then you run into problems of bias, where Mike sees a gray area as white, and Norm sees a gray area as black, and you end up with the New York Islanders having the top four guys in the NHL in hits or something.

Ryan Stimson had a passing project where he and a few others tracked passing data for the 2014-15 season. If that was available for every season, it would certainly help the data.

In the end, I'm just toying around for fun, and haven't the resourses to start tracking custom numbers.

as for the suggestions above, adding assists isn't a bad idea at all, as it's one more case of a confirmed touch of the puck that doesn't overlap with any of the other ones I already have.

 

[CpatainCanuck]

So, I was thinking about players that always seem to have the puck on their stick, and decided to try and see if I could come up with a way to measure it.

So far, I've taken individual Corsi + takeaways + giveaway as they are all indicative of a instance in time where the puck was on your stick.

Here's the top 10 Dmen ranked by touches/60, with 700+ mins this season:

Rank | Name | Touches/60
1 | Brent.Burns | 23.24
2 | Erik.Karlsson | 18
3 | Dustin.Byfuglien | 17.16
4 | Jeff.Petry | 17.14
5 | Jake.Muzzin | 16.64
6 | Victor.Hedman | 16.62
7 | Torey.Krug | 16.48
8 | Johnny.Boychuk | 16.45
9 | Morgan.Rielly | 16.38
10 | Kris.Letang | 16.03

Here's the bottom 10 Dmen

Rank | Player |Touches/60
10 | Paul.Martin | 7.69
9 | Adam.Larsson | 7.62
8 | Willie.Mitchell | 7.41
7 | Tobias.Enstrom | 7.03
6 | Ryan.Murray | 6.73
5 | Andrew.Greene | 6.55
4 | Christopher.Tanev | 6.14
3 | Mark.Borowiecki | 5.97
2 | Fedor.Tyutin | 5.46
1 | Niklas.Kronwall | 5.28

Rank | Player | Touches/60
1 | Alex.Ovechkin | 25.78
2 | Jason.Spezza | 24.34
3 | Nazem.Kadri | 23.59
4 | Max.Pacioretty | 23.58
5 | Jeff.Skinner | 23.37
6 | Rick.Nash | 23.36
7 | Vladimir.Tarasenko | 23.26
8 | Brent.Burns | 23.24
9 | Dustin.Brown | 22.6
10 | Filip.Forsberg | 22.54

Rank | Player | Touches/60
10 | Paul.Stastny | 9.99
9 | Chris.Vandevelde | 9.86
8 | Valtteri.Filppula | 9.81
7 | Mike.Ribeiro | 9.37
6 | Henrik.Sedin | 9.32
5 | Stephen.Gionta | 9.27
4 | Jori.Lehtera | 8.96
3 | Cody.McLeod | 8.33
2 | Alex.Tanguay | 7.71
1 | Jarret.Stoll | 6.11

Thoughts on whether this passes the eye test? or potential improvements?

My initial thought is it doesn't seem to work consistently with the forwards.

I don't think your metric shows what you think it shows. Henrik Sedin is, by your measurement, is the 6th worst forward for posession in the league.

Trying to measure possession by shots for and against is a mistake.

 

[Johnny Engine]

Ryan Stimson had a passing project where he and a few others tracked passing data for the 2014-15 season. If that was available for every season, it would certainly help the data.

In the end, I'm just toying around for fun, and haven't the resourses to start tracking custom numbers.

as for the suggestions above, adding assists isn't a bad idea at all, as it's one more case of a confirmed touch of the puck that doesn't overlap with any of the other ones I already have.

There would be frequent overlap. A shot from the point that gets tipped into the net would show up as both an assist and a Corsi event. If you want to include Stimson's data, that'd be much better, even if you only get one season out of it.

As for public stats vs. proprietary stats - yeah, it's the given the teams have better stuff. In my opinion, though, a clean breakdown of all possession gains and losses would not only be nice to have, but a short hop from what the NHL already tracks. In the end, the NHL is an entertainment conglomerate, and I know lots of its consumers are like me - we just love information, and want more of it! Then again, Gary Bettman thinks nobody liked CapGeek, so I'm not expecting much good.

 

[Micklebot]

I don't think your metric shows what you think it shows. Henrik Sedin is, by your measurement, is the 6th worst forward for posession in the league.

Trying to measure possession by shots for and against is a mistake.

It's Not using shots for and against. The only thing being used is personal shot attemts, so how many times did Sedin shot the puck himself.

As I said, it doesn't seem to pass the eye test for Forwards, probably because a forward will do things like cycle the puck, which won't really show up in shot attempts, usually would not be a give away if the puck is turned over, and isn't a takeaway.

With Dmen, it seems to work better but still clearly has it's failings (a guy always passes instead of shooting could be touching the puck just as often as a guy that shoots a ton, but would not show as well.

As for whether it measures what I think it does, well it measures exactly what I think it does; it's just not a very reliable proxy for personal possession time which was the goal.

I think that anyone who ranks highly likely does touch the puck a ton, but guys like Leddy and Sedin and Doughty who don't rank as highly as the eye test would suggest are doing something different.

One clear issue is the problem with RTS stats like giveaways, and takeaways. Using Away data only would probably help, but there are still issues.

 

[Legionnaire11]

Digging an old thread, but a couple of days ago someone posted a tweet that listed the top 5 "offensive zone puck possession" players in the NHL and listed by minutes per game. Patrick Kane was #1, I think Barzal was #3 on this list...

anyway, I asked this poster and in general if anyone knew where to find those stats for the entire league and got no response. So I figured I'd come here to see if anyone has an idea on if these stats are available anywhere?

 

[93LEAFS]

Digging an old thread, but a couple of days ago someone posted a tweet that listed the top 5 "offensive zone puck possession" players in the NHL and listed by minutes per game. Patrick Kane was #1, I think Barzal was #3 on this list...

anyway, I asked this poster and in general if anyone knew where to find those stats for the entire league and got no response. So I figured I'd come here to see if anyone has an idea on if these stats are available anywhere?

I've seen Mike Kelly tweet stuff similar to this. I don't find it that helpful because they never make all the information public, just bits and pieces of it, so it is hard to tell how valid the stuff actually is.

 

[Legionnaire11]

Thanks.

I'm actually making player ratings for sim leagues. I'm looking for useful stats that I can relate to the "Puck Control" rating.

Similarly, but less pressing is for skating and defense.

Anyone with suggestions? Even a combination of different advanced stats.