Sunday, August 10, 2008

Zone Time

From 1999/2000 through the 2001/2002 seasons, the NHL recorded zone times and published them on their game sheets. I don't know why the NHL stopped making this information publicly available, then again I don't understand a lot of the things that they do, I suppose they have their reasons.



Zone time and possession are terms that are often used interchangeably by hockey coaches, and I think we all realize why; having a bit more than your share of touches in the neutral ice pachinko game doesn't represent much of an advantage. But having meaningful possession moving forward, and in the offensive zone, that often ends well for you. When Mike Babcock says "Possession is everything." we know that it's hyperbole, and we know that he's not talking about one of his defencemen standing behind his own net with the puck.

Unfortunately the NHL never broke this down by game situation, so the PP stuff is in there too. Still, it's better than nothing.

I thought it would be worth looking at a season to see how shots, and shots directed at net, meshed with the zone times for the teams. Now obviously some teams play more of possession game than others, and unbalanced scheduling means that some teams played this type of opponent more than others. Still, the connection between shots and zone time will still shine through, as will the link between the ability to outscore opponents and these two things.

So I wrote a simple little Excel macro to scrape the zone time info off of the NHL.com game sheets for 2001/2002, and the shots stuff from the NHL.com event sheets. Laziness prevented me from filtering out the empty net goals, and the data is missing for 21 of the 1230 games, c'est la vie.

The results for the teams are listed in the table below. By way of example, for Atlanta the puck spent 355 more minutes in their own end of the rink than in the good end. They were outscored by 101 goals, outshot by a margin of 927 shots, adding in the missed shots they were beat to the tune of 1276, and for all shots directed at net they were outdone by 1532. And witness the Oilers reaching the very heights of mediocrity.



It's self explanatory, and as always, readers are encouraged to add common sense. A couple unusual seasons for teams in there, but for the most part it rolls out as you would expect. Clearly Babcock has a point, better teams, by and large, have better zone time, though it is a surprisingly small total over the course of a season, even for the terrific teams. Having said that, if I were to make a sweeping statement I probably would go with "Goaltending is everything."

And to try and pin values on it, the Pearson correlation coefficient between zone time and the three different shots metrics:

Shots +/-: .86
Fenwick +/-: .87
Corsi +/-: .90

Much of a muchness between the three really, this is a larger sample than a guy is usually looking at, though. And just generally the Dallas and Colorado seasons look like quirky ones this year, and if I gave even half a damn about either team, I'd probably have a closer look.


Update: In response to a comment by Traktor, the table has been changed to reflect which teams made the playoffs. The playoff team names are highlighted in light green.

20 Comments:

Blogger Slipper said...

Obviously some of the teams at the top of that chart were employing the old "let the other team tire itself out in your own end" strategy.

You'll never be able to encapsulate THAT with a spreadsheet, Ferrari!

Otherwise, good stuff.

8/10/2008 6:52 pm  
Blogger Traktor said...

Half of bottom 16 puck possession teams made the playoffs, so no correlation between possession and wins?

8/10/2008 7:03 pm  
Blogger Vic Ferrari said...

I think you're looking at the wrong season, Traktor, this is 2001/2002.

Of the teams that averaged more the a minute extra in their own end every game, only Montreal and Colorado made the dance. MTL had horrific underlying numbers, but presumably either a hot PP or a hot goalie (was that theodore's big year?).

Colorado I can't explain by memory. How do you lose the territorial battle that badly and still outshoot your opposition by 75 shots? And why weren't they getting few shots blocked?

Of the teams that averaged an extra minute of zone time per game, all but Dallas made the playoffs. Was that the year they canned Hitchcock? A shame if it was, because pretty clearly by these numbers they were still playing good hockey (and also by my memory, it was much discussed on OilFans at the time). Just wasn't their year.

8/10/2008 8:02 pm  
Blogger Showerhead said...

I think it's great work. Zone possession is one of those facets of hockey that seems so obviously tied to winning that it would have been alarming if things were especially out of whack. It would really fuck with a big part of how I see the game, anyhow.

The problem with proving this is that it doesn't immediately scream practical value. Even if contribution to zone possession were easily quantifiable, it's hard to think players who performed well in that regard wouldn't also be getting attention in terms of counting numbers or +/- results... Add a little context and even while ignoring Fenwick numbers or your own location-of-faceoffs-beginning-and-end-of-shifts, how many good players are left?

8/10/2008 10:59 pm  
Blogger Black Dog said...

That's terrific stuff Vic. Thanks so much for the work you put in.

8/11/2008 7:56 am  
Blogger Kent W. said...

It's shame this stuff isn't available for more recent seasons.

Thanks for more great insights nonetheless Vic.

8/11/2008 9:26 am  
Blogger Big T said...

Terrific stuff Vic,

I wonder if someone has already done somehting testing the correlation of corsi, fenwick, etc.. to WIN%?

If it's been done, I wonder if you could point me at it... if not, it might be of some value. no??


T

8/11/2008 11:54 am  
Blogger Schitzo said...

One note, Vic - Unless I miscounted, you missed one of the playoff teams. I'm counting 15 in green.

8/11/2008 11:55 am  
Blogger Showerhead said...

T: not speaking for Vic, who tends to be the digger-upper of this info, but I think some of the generally accepted wisdom around these parts involves goal differential being a more accurate suggestion of a team's caliber than win %. Seems to scrub away some of the luck, which I'm sure you are aware of... buy anyway, hence the GD column and not win%.

8/11/2008 12:17 pm  
Blogger Vic Ferrari said...

Schitzo:

Thanks, I'd missed Montreal, and this after I'd commented on them in this thread, d'oh!

That was the year of Theodore's .931 save%, and the Hart and Vezina trophy wins. He was a bit good not long ago, strange the way he just came off the rails.

8/11/2008 12:35 pm  
Blogger choppystride said...

Very cool. Didn't know that zone times were given back then.

Since you're assessing the accuracy of SE (shot events) as a proxy for zone times, wouldn't it be better to do a proportional comparison instead?

i.e. perhaps we should be checking if:

(corsi for)/(corsi against) == (off zone time)/(def zone time)

In the past, I've also suggested incorporating other events besides shots to increase the sample size. Perhaps this idea can be tested to see if it enhances accuracy? If it works, then perhaps these other events could be used to estimate the neutral zone times as well?

8/11/2008 12:37 pm  
Blogger Vic Ferrari said...

T:

What Showerhead said, I listed the goal diff because I'd like to avoid the stuff that I think is mostly just 'shit happening'.

Bear in mind that this includes the PP stuff as well, and I didn't filter out the empty netters, so it's a bit of a mess to start with. Still, some trends are pretty obvious I think ... everyone has looked at the goal diff and the other four columns and spotted the obvious, surely.

How much of that is coming from the PP? A good question that I don't know the answer to.

8/11/2008 1:28 pm  
Blogger Vic Ferrari said...

choppystride:

I didn't really scrape the data off that way, so I'd have to do it again, Wouldn't take much, I can do that and post it as an HTML table on the web for others to take a kick at, for other seasons too. I won't rush though, every time I've done this over the years barely anyone downloads it.

On the second point, you're absolutely right, I mean you watch a Hartley or Wilson or Carlyle team play, and clearly they aren't using the point shot to get the puck into the house at evens as much. Offsides would be gold, the separation of faceoffs+/- from corsi+/- will happen too (just as it will for the Leafs this year). % of EV shots taken by defensemen might be a way around it too.

In any case we're trying to find a decent measure for meaningful possession here, and zone time is only one of them anyways, though it would be damn nice if the NHL gave it to us.

8/11/2008 1:38 pm  
Blogger Vic Ferrari said...

As I'm procrastinating on other things, the answer to choppystride's question, FWIW:

Correlation coefficient of Zone Time ratio for the 30 teams in 2001/2002:

To Goals for/against ratio: .74
To Shots for/against ratio: .84
To Fenwick for/against ratio: .85
To Corsi for/against ratio: .89

And to answer T's question, in a half-assed way (i'm not bothering to look up win%, goal differential will have to do), with this data for this season at least:

Correlation coefficient of goal differential for the 30 teams in 2001/2002:

To Zone Time +/-: .79
To Shots +/-: .82
To Fenwick +/-: .78
To Corsi +/-: .78

8/11/2008 3:19 pm  
Blogger Big T said...

showerhead;

Agreed on GD being the better indicator of the quality of the team, just looking at it a different way - something that might be more convincing to those who are less sabremetric-ally inclined.

vic;

Thanks for that. Tough to miss the relationship there. I can use pythegorous to get an estimate from there I suppose anyway.

I'm going to have to break out my stats textbook and bone up on this so I can add a little more to the discussion.

Thanks again,


T

8/11/2008 4:32 pm  
Blogger choppystride said...

Those numbers look good to me.

I just made a quick visit to NHL.com and realized that they didn't give play-by-play events back then - only game summaries and event summaries were given, it seems. So I guess that incorporating other events into corsi/fenwick is not doable for that season.

But still, thanks for all the work!

8/12/2008 12:02 am  
Blogger PerformanceOil said...

Interesting analysis, and this is the kind of thing that needs to be done if someone is going to say Corsi ~ possession.

There are a few problems though.

First, three smaller things: I don't get the same correlation between possession and Corsi that you do - I get ~0.813. Still strong though, whichever value it is, I would say it is providing a good way of estimating possession. Another complaint is that we have no idea how zone time was determined. Given that they have abandoned reporting the stat, it may be that there was a problem. Not something to abandon the analysis over, but certainly something to be cautious about. Finally, this is again at the team level, and the ability to extend the conclusions to the level of the player is unproven.

The more major (and easier to quantify) criticism I have is due to the season. 2001-2002 is some time ago, and is only one season. There is no way to be sure that it is representative of the NHL now or even then.

In fact, I have some evidence to suggest things have changed since then.

In 2001-2002, when Corsi and possession showed a very strong correlation, shot differential (SD) and goal differential (GD) also correlated quite strongly (r^2=0.671). In 2002-2003, this correlation dipped down to ~0.4 and then in 2003-2004 it was back up to 0.6.

This was all pre-lockout.

Post lockout, the last three seasons have shown correlations of 0.491, 0.232 and 0.287, between SD and GD.

Overall correlation during the last six seasons: 0.44.

Pre-lockout vs. post-lockout: 0.558 vs. 0.313.

This is some evidence that things have changed (considerably!) since the lockout, judging by these values.
Now, this fluctuation may not be meaningful (as far as pre-lockout vs. post-lockout), and time may make things look less divergent. Nevertheless, it shows quite clearly that 2001-2002 cannot alone be used as a representation for what happened in 2007-2008.

2001-2002 may have been an aberrant season. Or, perhaps things have changed post-lockout. Another possibility is that even though SD and GD do not correlate so well anymore, Corsi and possession time still do. There are other explanations as well, I'm sure. Unfortunately, without more modern possession time data, these possibilities are difficult to distinguish.

Bottom line to me is that we still can't be sure how good of a measure Corsi is of possession at the team level (in 2007-2008), and there is even less certainty at the player level.

BTW, I replied to your posts over at LT's. The issue of Shots for correlating well with goals for and SA with GA is an interesting one, given their differences don't. There's a few reasons I can come up with for that, but I still don't have the full solution.

8/12/2008 5:32 pm  
Blogger Vic Ferrari said...

there is a lot of stuff that is in there that has been talked about quite a bit on the internet already.

The fact that the NHL stopped publishing the zone times, they stop and start publishing things often. In this case I expect that journalists started writing "neutral zone time" articles. Because there were some boring games watched by journalists with this sheet right in front of them, and human nature is what it is.

And anyone who has ever coached loves this type of data, above everything. It is unimaginable that the NHL coaches were willing to let this slip away. Completely and utterly.

Also, I know you haven't been around here long, but the NHL's collection of data differs from what it makes available to the public. Occasionally they feel moved to give us data from years back that we never knew existed.

You can look that up yourself, easy enough to do, PO.

8/13/2008 3:00 pm  
Blogger Vic Ferrari said...

Just to add:

Traktor is a moron, did anyone else miss that? Dude didn't even read the post.

The Oiler blogging community is a very special place IMO. Special enough that it deserves protecting.

The OilFans bridge quietly evaporated, that was good. But Lain is frightened of confrontation at all costs, just his nature, so the bridge between HF and us may never be burnt as it should.

Call me a socialist, but I think that everyone has a duty here. Bloggers and regular commenters alike.

Am I wrong?

8/13/2008 3:17 pm  
Blogger Jim Philips said...

I think that it is great that they are doing something like that. It makes things more interesting and provide data that it can help a lot of stuff. For me it is the price per head free demo

5/30/2013 10:55 pm  

Post a Comment

<< Home