Wednesday, August 03, 2011

Defining The Scoring Chance

The 6-Minute Abs?

By now, a hockey fan, you've heard plenty about advanced stats and the rise of the scoring chance.

When I was a much more naive stats geek, I used to make a lot of assumptions about what I was hearing and extrapolate that information to make my own (ultimately flawed) interpretations. I can tell you that when I first found out the working definition of scoring chance being used by the squadron of scoring chance trackers around the league, I was more than a little bit surprised.

I want to show you all some work I've been doing with scoring chances over the last couple of seasons. So as a service to those who are starting from where I was just a little while ago, I thought I'd clear up the assumptions before we start.

How better to do that than to ask the Patriarch of Habs Scoring Chances, Olivier of En Attendant les Nordiques, the question:

"Just what is a scoring chance anyway?"

I asked him more questions than that, and he kindly answered them all. But by the end of this, we should, agree or disagree about the methodology, at least be on the same page about the stat in question.

Without further ado:

What is a scoring chance?

My definition: a scoring chance is a shot on goal or a missed shot unleashed from the "home plate" area. The home plate area goes from goalposts, up to the face-off dots, then up toward the blue line to the top of the face-off circles.
(image taken from Copper and Blue, 2010)


So every shot is not a scoring chance?

No, when the shot is taken from outside the zone and doesn't result in a goal, it isn't a scoring chance.


And some non-shots are scoring chances?

That's right. Both missed shots and shots on goal from inside the home plate area are accounted for because I see them as successful attempts to challenge the opposition's goaltender. A blocked shot is not, in my opinion, a scoring chance because a defender actually thwarted the challenge.

I should say here that I give myself some leeway; if a guy floats a feeble backhander from the top of the plate and the goaltender easily follows it, I don't give a chance. A hard shot from the limit of the zone will be counted as a chance.



You say a shot can also be a scoring chance if it results in a goal?

A goal scored from outside that zone is counted as a scoring chance *if* it's an actual attempt to challenge the goaltender. So, Gill floating a wrister from the blue line is a scoring chance. Other guys recording scoring chances have some variations to their definitions: some count screened shots from the blue line, some never allow a chance from a shot from outside the home plate even if it's a goal.


So not every goal is the result of scoring chance then?

Not in my record of chances, no. Best example is a goal scored by Jeff Halpern early in the 10-11 season: he was behind the net and attempted a pass in front, only to see the puck deflect on a defender's skate into the net.

But that doesn't mean being behind the goal rules out a chance. I've also given a chance on a Moen Goal where he was behind the goal line and clearly shot it on the goaltender's leg to bounce it into the net.

Basically, to be a chance, there has to be an attempt at challenging the net.



So we are left with:

1) A shot released from inside the home plate or a goal, that
2) "Challenges the net"

is a scoring chance.




This is probably at odds with what a lot of people think at this point. It was definitely at odds with what I thought before knowing the definition being used by trackers myself. I used to think of a scoring chance as a shot that "challenged the net" basically from anywhere. I think I was also quite surprised to find out that missed shots were scored in the same way that shots on net were.

To me there still exists a massive undescribed hierarchy of scoring chances within the scoring chances that are being tracked around the league. Call them good scoring chances, great scoring chances, quality scoring chances, whatever you like. I think many like me are left thinking on the inherent difference between miss from inside the home plate and save on a breakaway or two on one.

I brought it up with Olivier and he was unfazed.


But with this definition all scoring chances are equal. Are all scoring chances equal?

As far as scoring chances are concerned, quantity *is* quality. A scoring chance is a shot attempt from a zone that leaves almost no reaction time for the goaltender. Whatever difference in quality one can find between scoring chances is, I think, drowned by the fact that you actually *got* a scoring chances.

Let me put it this way: over a normal game, I record around 35 scoring chances. Of those, maybe 25 are at even strength. One team getting 15 ES scoring chances in a game is huge, below 10 is feeble. So, is this scoring chance better or worse than that scoring chance? I think the important point is that scoring chances are very dangerous events, 15% of which ending in goals while maybe 4% of Shots+Missed shots are converted in goals.



Why not at least track whether the shot was on net or off net? That could be an interesting aspect of quality to look at down the road.

That might actually be doable trough a script. But I don't think it would tell us about the "quality" of the scoring chance.

What is more dangerous: a 20 foot wrister that Cammalleri sends 6 inches over the cross bar, or a 20 foot wrister Moen buries into the goalies crest?

Missed or not, the guy tried to bury it by unleashing a shot from a given spot. When you talk about "quality", it seems to me you see that said quality comes from the spot the shot was unleashed from, not what the shot looked like going toward the net. That's an important point.



I think there'd be disagreement about this from observers of the game. Just as the scoring chance people have broken away from simply using shots. In "Something About Mary" the hitchiker has just thought of 6-minute abs as a revolutionary idea to replace 7-minute abs. Ben Stiller's character asks what will happen when someone invents 5-minute abs.

So what if scoring chances as they are now is the 6-minute abs? What if someone comes along and ranks shots from the zone where people score most often calling them really good scoring chances?


Are Scoring Chances the 6-minute abs? Yes they are, I am absolutely certain of that.

I record scoring chances for three main reasons, and I list them in order of importance to me: 1) I enjoy doing this 2) bunched together in a multi-game data stream, they give a better idea of who's driving the bus 3) If I take said data, make cool table out of them and write my toughest about them on a blog, people come around and we end up having very enjoyable discussions about stuff I enjoy.

Can someone come up with a better ranking of where shots are coming from? Yes, probably. I have an iPad and am fairly convinced that I could, should I beat myself into building the right tool, use it as a too for recording scoring chances in a more interesting fashion. Somebody, somewhere, will eventually hammer out such a thing and we'll end up with richer data. The question is, where will this data come from?

The NHL is putting charts up on the game centers pages of each game where you have a goaltender's save% on shots from different spots in the defensive zone. The 5 minute abs may very well be just around the corner.



Anyway, agree or disagree about the detail of the statistic, they are still useful to fans looking for more, right? On their use, is a scoring chance a team or an individual statistic?

Both. Scoring chances are useful because they are a "purer" from of shot-based metric. That is, they give you a more precise understanding of who was more challenging to the opposition. Even tough they are a team metric, they are more useful as an individual measure. They give a more reliable understanding of who is generating offense, even over a short span of time (that is from 10 games to a full season) and help weed out the luck factor.

I see two levels of individual statistics: individual achievements (Andrei Kostitsyn had a shot from between the dots, accounting for a scoring chance) and on-ice events (AKost, Cammalleri, Pleks, Gill and Subban were on the ice for a scoring chance). From my perspective, scoring chances as on-ice events is where they are more revealing about a given player.

As I wrote earlier this summer, Scott Gomez was terribly unlucky this past season: he was on the ice for the same amount of time at even strength as in 09-10 (about 1190 minutes) and the team had about the same amount of scoring chances while he was on the ice, about 370. Yet, the team scored 20 goals less in 10-11 (35 to 55) and Gomez ended up with 20 ESPoints instead of 40 (his career norm up until then). That is bad luck, pure and simple. I think scoring chances tells you who's actually driving the offense, luck be damned. Obviously, wingers will get more individual scoring chances than centers, who will get more than defensemen, but that's a structural reality.



How do you think scoring chances will be used in future stats analysis (by fans)?

It's all about the data made available. You look at advanced stats analysis done by fans and they mostly use behindthenet.ca and timeonice.com's outputs as building blocks. The managers of those two sites are, by the way, to be recognized as the enablers of the whole scoring chance projects you see left and right. Desjardins, of BehindThenet, used his site as a hub of analysis of data collected by Chances Recorders and timonice is the site that gives public access to a script that allows anybody to punch in time codes and notes and get a nice set of HTML tables of scoring chances data. Without these guys, there is no "Scoring Chances" as we see them being batted around in the hockey blogosphere nowadays. Making multiple years of scoring chances data on multiple teams available to fans is what will allow them to take these data stores and cross reference them with other sources such as Behindthenet.ca and timeonice.com and come up with some new analysis.

On what? Off the top of my head: defining shot quality, quality of competition faced by goaltenders (who, of Halak and Price, saw the highest rate of scoring chances against in 09-10? If I'm not mistaken, Halak), luck-independent offensive and defensive output. Stuff like that, y'know.



When are fans on the laggard teams gonna get on this? Any word, because with the small sample now, it might as well be the Habs alone

The problem is, we aren't pooling the data already being collected so it's hard to tell who's on board. Of the 2010-11 season, I'm fairly certain the Habs, Leafs, Caps, Flames, Rangers, Flyers were recorded. I think some work was done with Chicago.

The next step would be some form of aggregation of data; seeing the data store would probably be an incentive for fans of uncharted teams to do the legwork.



The NHL teams record scoring chances, do you ever foresee these stats being available to fans?

No. Actually, I think NHL teams data stores are way more granular than anything we ever see out in the open (touches? Zone time?) and I think the NHL's data stores are way more refined than what we see on their website. But team's data stores are an asset that they manage so to give them a competitive advantage. They will *never* publish current data. The data published by the NHL is, if I'm not mistaken, the only data eligible for salary arbitration process. So whatever data we currently have from the big guys, it will all stay the same unless the new CBA states so or, maybe, if the NHL comes up with something akin to MLB's Pitch F/X. I'm not holding my breath.

What would be insanely cool is if some historian/researcher could convince old franchises such as the Habs, Rangers or Wings, to open up whatever archives they have so to give us a glimpse of how they approached stats in, say, the 40's, 50's or 60's.



Thanks again of course to Olivier for answering all of my basic questions so fully and clearly (even under challenges).

In our unending search for the 5-minute abs, and new and more interesting ways of crunching data in 6 minutes, we'll share some scoring chance extrapolations with you in the upcoming while.

At least now we can all start on the same page.

No comments: