Wednesday, October 07, 2009

Oiler Team Scoring Chances and Reasonable Expectations for 09/10

Since all points in the standings are derived from scoring goals and all goals emanate from scoring chances, the New York Rangers coaching staff emphasizes the ability to create scoring chances.

Kevin Leonard, NYR Database Designer, c.1992

I like rolling averages, it smooths out the bumps in the season that are caused by tough stretches of the schedule or puck luck. This spring I posted a few items on the Oilers' scoring chances, all concerned the individual player on-ice scoring chances. This post is the first looking at the team totals.

"ScoringChance%" is he percentage of the total even strength scoring chances in the game that were owned by the Oilers. So if, over twenty games, the Oilers racked up 200 EV scoring chances, and the opposition tallied 300 EV scoring chances ... then the Oilers would lay claim to 40% of the scoring chances over the stretch (200 of 500).

"Corsi%" is calculated using the same reasoning, as is EV "Goals%".

Corsi% is our best guess at the extent of territorial advantage held by the Oilers in the previous 20 games.


I think it speaks for itself. Here is an enlargement of the 40% to 60% band:


So, quite clearly Corsi% (or any of the shots metrics for that matter) ends up giving us a clear indication of scoring chance percentage. Using 20 game rolling averages as above, they are almost on top of one another.

On the year as a whole:
ScoringChance% = 47.7%
Corsi% = 47.8%
Goals% = 51.0%

The Oilers got a bit fortunate with the shooting/save percentages, mostly due to the fact that Roloson had a terrific year at even strength. (In fairness, that gets balanced out a bit by some bad puck luck on the PK, in a fairer universe the 0809 Oiler penalty kill would not have been nightmarish, merely terrible).

The Goals% (yellow line) may look volatile, but EV goals are rare birds in NHL, so it always will be similarly wiggly for all teams in all years. If you drew a straight red line at 51.0%, then coloured in the spaces in between the yellow curve and that red line, filled them in with black dots (Jim Albert's BLACKstat) then the black area would represent the streakiness or inconsistency of the Oiler Goals%. And in 753 of 1000 parallel universes the Oilers would be expected to be a touch MORE streaky with Goals% than they were in 0809.

The Oilers just weren't a very good even strength hockey team in 0809. And they aren't this year either.

There is a downward trend in the ScoringChance% starting early in the back half of the season, and it never recovered. I'm not sure when Hemsky got concussed or when Visnovsky's season ended, but it must be thereabouts. Both are terrific players when healthy, and neither appear to be fully fit at the moment, so this next stretch could be tough for the Oil.

On the other hand, Gagner and Cogliano should improve, Gilbert and Grebeshkov too, Penner even. Though, as RiversQ showed in an earlier post, that progression is wildly erratic. And the guys who are on the hard downslope of their careers (Moreau, Staios, Strudwick, Pisani) can't do much worse this season than last. Plus Comrie is a good addition to my mind, or maybe it's just the thrill of seeing the Oilers sign someone with a good chance of beating his contract price.

My conclusion: With good health, the Oilers should be better this year than last, maybe even become dead average at outchancing (50%). Let's dare to dream.

26 Comments:

Blogger Dennis said...

An interesting take because a few of us believed that it was EV play that killed the '09 playoff hopes (this was parroted by MacT on the TSN panel).

But perhaps Roli's play did skew the numbers and I'm sure we'll start to get an idea about 25% of the way through the season.

Regarding 71, I can't remember the exact numbers but I think there was a four or five chances per game swing once 71 went down. Though he most assuredly peaked counting stats-wise from 06-to-07, the guy is still a beast.

10/07/2009 1:26 pm  
Blogger Vic Ferrari said...

Maybe MacTavish was setting up Buchy to look like a star when the PK improves and the EV results stagnate or get worse.

I think the scoring chance numbers (and other underlying numbers) will improve for both EV and PK, if the key guys get healthy and remain that way. But the rest is in the hands of the hockey gods, and they dispense justice at random.

If I were to wager at even odds, I'd bet big that the Oiler PK %s improve, even if they don't add any proven PKers, and including the cruel bounces in those first two games.

And right now I'd take the under on the Oilers scoring 51% of the EV goals that happen in their games this year.

I think that they should be in the thick of it for a playoff berth, it will probably be a tight grouping again this year though. We'll see.

10/07/2009 2:08 pm  
Blogger Jonathan Willis said...

Question for Dennis/Vic:

Given the similarities (on different scales) between Corsi rating and scoring chances, doesn't the former make the latter rather irrelevant?

10/07/2009 5:16 pm  
Blogger Kent W. said...

I was counting scoring chances just tonight (and not enjoying it) and was going to ask the same question as Jonathan...is there a point, because every time we've looked the correlation between the two has been pretty strong.

10/07/2009 9:37 pm  
Blogger JLikens said...

In response to Jonathan and Kent:

While corsi and scoring chances tend to converge in the long run, the two do not necessarily correspond over smaller sample sizes, such as a single game, or even a collection of games.

To give a concrete example, if two teams are relatively evenly matched, and one team establishes a lead early in the game, and retains that lead for the remainder of the game, there's a good chance that the trailing team will establish a sizable corsi advantage by the end of the game. However, the scoring chances will likely be more or less evenly distributed (after all, the teams are evenly matched).

I suppose my point is that, relative to corsi, scoring chances are more reflective of the balance of play over shorter periods, and therein lies its utility (at least from my standpoint).

10/07/2009 9:53 pm  
Blogger Jonathan Willis said...

JLikens:

True enough, although I'm guessing voer short spans most of us trust our eyeballs.

I just kept looking at the scoring chances (and thanks for those, Dennis) last year to use them for something and over the long haul I couldn't see any reason not to use Corsi instead.

This study has, if anything, only proved to me that Corsi is the most relevant metric of individual ability (taking context into consideration, of course).

10/08/2009 12:38 am  
Blogger Scott Reynolds said...

I think it's important to remember that we only have one season's worth of data from one observer for one team. I think it would be unwise to make sweeping generalizations. Maybe all teams behave the same. Maybe they don't. JLikens has pointed out in the past that the Avalanche are a team that has a much better shot differential than they do Corsi or Fenwick. Maybe this is something that will persist this season. If it does, maybe scoring chances will help us to sort out that discrepancy (i.e. perhaps they have more scoring chances per Corsi than average). It could also help with the shot quality debate. Will all of the teams we're looking at have the same and/or similar scoring chance per Corsi rates? Or will they be different? With respect to the Oilers, the personnel on the team has not changed drastically but the coaching has. Maybe we learn something about coaching from looking at scoring chances. I think there's a lot to look at with scoring chances. Maybe you end up being right and they're not worth the effort but I don't think we're in a place to make that conclusion yet.

10/08/2009 5:48 am  
Blogger Jonathan Willis said...

Scott:

Those are good points, plus I imagine we'd need to look at different teams and coaching strategies (the change to Quinn this year should help make that interesting) to come to any conclusions.

Still, intuitively, wouldn't it be just a little shocking if Corsi and scoring chances weren't proportional on the large scale?

10/08/2009 8:39 am  
Blogger Scott Reynolds said...

Oh, I definitely think that the charts will be similar regardless. I think we suspected that before we began. Corsi isn't just about "shots," it's about controlling the play and, fundamentally, scoring chances are measuring the same thing. But how closely they match in general is an important question. What each of Corsi and SC say about a specific player or team might well be a little bit different and might help us to understand the game a little bit more.

The quote Vic opened the piece with could well read, "Since all points in the standings are derived from scoring goals and all goals emanate from scoring chances and all scoring chances require a shot directed at net..."

That doesn't mean that each step along the way (standings points, goal differential, SC differential) doesn't have something to tell us about how the game works.

10/08/2009 9:13 am  
Blogger Olivier said...

Off the top of my head, I'd love to see what an expanded scoring chances dataset tells us about the variations in s% and save% from team to team...

And I certainly didn't enjoy scoring last night's 3rd period between the habs and Vancouver. Ungrateful bastards.

10/08/2009 2:19 pm  
Blogger Olivier said...

Hm. I was referring to the habs, of course.

(second language and all that stuff, y'know...)

10/08/2009 2:20 pm  
Blogger Sunny Mehta said...

I've said this before, but coming up with some sort of basic definition of "scoring chance" would be good, not only to get everyone on the same page, but also to open up discussion on whether the definition should be revised.

After the Devils' opening night game on Saturday (a blowout loss to Philly that was actually very evenly played by my eye - mostly just a few bad bounces and some questionable goals let in by Brodeur), when prompted "what went wrong" Lemaire basically said something like, "nothing. we had 11 scoring chances which is right in line with where we should be. last season the team averaged between 10 and 14 per game."

Anyway, I noticed that number seems a little lower than what you guys are averaging. It's possible your scoring chances are correlating so highly with Corsi because you are counting them that way.

10/08/2009 5:53 pm  
Blogger Olivier said...

There is no doubt in my mind I am currently being over-generous when scoring chances. Desjardins commented on one of my posts that teams actually qualify chances as A B or C or something like that.

Then there is the fact that the NHL is actually keeping tab on scoring chances by itself (which is where networks like Versus take their stats from; I saw RDS use them once or twice too).

Kent Wilson actually pretty much summed up how I score them in a post at matchsticks and gasoline.

I suspect Lemaire is drawing his stats from the NHL; I used to score chances for the latter half of last year's habs, and I remember Carbonneau and some journalists quoting totals that were sometimes inferior to mine (but at other time pretty near what I got). My guess is the difference stems from a mix of two things:

- It's possible to get a better grasp of the actual quality/existence of the chance taken when you are sitting in the pressbox instead of on your couch.

- The league actually grade scoring chances and the team only keep up with grade-a chances.

Either way, I'm flailing around in the dark here... At the end of the day, I do this because I enjoy the project (spending too much time in universities / colleges does that, I guess). Also, finding "nothing" (ie: Corsi = SC) is already finding something; one data set from one coder isn't enough to jump to conclusion, I think.

10/08/2009 6:12 pm  
Blogger JLikens said...

I'm fairly confident that the 'official' scoring chance numbers -- that is, those referenced in the media or displayed during a broadcast (such as on versus or Fox sports net) -- are low because only chances that result in a shot on goal are recorded.

So, for example, a goalpost, which often results from a quality opportunity, wouldn't be defined as a scoring chance according to the definition employed by the networks.

10/08/2009 6:36 pm  
Blogger Sunny Mehta said...

Yeah, just to be clear, I wasn't trying to criticize - I think what y'all are doing is really fantastic.

I was just thinking out loud that if we repeatedly find very little difference in the scoring chance numbers and corsi, we could bump up the minimum threshold of quality for "scoring chance" to see if that reveals anything.

10/08/2009 8:01 pm  
Blogger Scott Reynolds said...

For me the scoring chances include a clear play directed toward the net from a dangerous scoring area (loosely defined as the top of the circle in and inside the faceoff dots, though I admit I'm sometimes slightly more generous depending on the movement of the puck and the player) or a screen shot that reaches the goaltender (and again sometimes I'm more generous with these once taking puck and player movement into consideration which sometimes results in a higher number of PP chances).

10/08/2009 11:27 pm  
Blogger spOILer said...

I used to count scoring chances for visiting TV for the Oiler games way way back in the day (with other stats), and from my memory, which is admittedly suspect, Scott's definition is pretty damn close, although I was usually pretty picky on counting screened shots especially from distance, even if the shot gets through. To me those were too low percentage. A screen however would be counted if the puck carrier is driving the slot unhindered and using a screen to his advantage.

But I AM pretty certain that I did not have to limit my counted chances to a shot hitting the net. In fact, my impression at the time was that scoring chances were recorded to account for times when the opportunity was great but the finish was not.

I would have defined a scoring chance as having time and space in a dangerous area and using that time, space and proximity (shooting).

10/10/2009 3:32 pm  
Blogger Sunny Mehta said...

Vic,

How does Zone +/- (or Zone%) do as a proxy for scoring chances? Better or worse than Corsi?

10/11/2009 11:12 am  
Blogger Colby Cosh said...

That doesn't mean that each step along the way (standings points, goal differential, SC differential) doesn't have something to tell us about how the game works.

On the team level it opens the door to a superior component analysis of standings points. We got x marginal points last year from good timing/"clutchness", x from offensive finish at evens, x from territorial dominance at evens, x from goaltending, x from the shootout, x from the special teams...

10/19/2009 2:11 am  
Blogger hunter1909 said...

Oilers should put this blog on it's payroll.

10/19/2009 9:33 pm  
Blogger Hawerchuk said...

Incidentally, this is Kevin Leonard:

http://www.hpme.utoronto.ca/about/faculty/list/leonard.htm

10/20/2009 9:32 pm  
Blogger Vic Ferrari said...

JLikens/Scott/spOILer:

Yeah, I agree with all that. In general, I count any time when, if the shooter had made his shot, it would have been a goal. An unstoppable shot, as Ron Wilson likes to say.

Screened shots are the fuzzier areas, for sure. Sometimes it doesn't look like a chance live, but the view from behind the net afterwards shows the goalie coming across and the forward jumping in front of him just as the Dman rips a slap shot. There is a lot of net available, if the puck hits it then it's a goal. And the goalie is in the butterfly hoping it hits him, doesn't even flinch until the puck is past him or hits him.

If you don't get the video replay from behind the net, though, it's tough to say.

As long as people are consistent and honest, it doesn't matter much anyways.

11/02/2009 12:26 pm  
Blogger Vic Ferrari said...

Sunny:

I've never looked at faceoff zones, Sunny. It's not a stat I use for predictive value at the team level. Feel free to do it though, be sure to account for icing rates though.

On your theory that scoring chances are inherently linked to shots-on-net or shots-at-net. To fit math to JLikens' response: If that were fully true, we'd expect the Pearson correlation of total shots and total scoring chances to be perfect (1.00) in fact. And given sample size constraints (77 games) the real world level should be r=.90. And between .88 and .92 in almost all parallel universes.

It was r=.48 for last season's Oilers. The odds of that happening by coincidence alone are staggering.

It's actually worked out well that I was so late to get back to this thread. A bunch of folks are recording scoring chances, and it is commonplace for the scoring chance % and shots% in any one game to be wildly different.

11/02/2009 12:43 pm  
Blogger Sunny Mehta said...

Vic,

Wow, yeah r=.48 is lower than I would've expected. Is that for shots or for Corsi? Either way, it's good news, and it quells my skepticism. :)

11/02/2009 5:18 pm  
Blogger Vic Ferrari said...

That's for Corsi. Best to stick with that instead of shots imo because it kills the effect of scorer tendencies.

Otherwise you end up erroneously thinking, for example, that Chicago is a shot happy team. That their low chance/shots ratio is a result of their system and that their goaltender save% is better than it really is. Really it's just a matter of everything being "on net" in the eyes of the NHL's Chicago shot scorer. This for both teams on the ice.

It matters little what Joe Fan thinks. Pat and Lain (bless them both) ride that yellow curve above with reckless abandon, at least compared to most around this corner of the Internet. And they would still be the voices of reason if they were talk radio callers. Grabia used to be manic that way, now he's as cold hearted as Tyler or Dennis. I suspect that he was a happier fan before.

It DOES matter that the Oilers have a grasp on these ideas, though I don`t expect them to bust out math or anything. Unfortunately there isn`t a hell of a lot of evidence to support that. Even when they do well it seems like a coincidence to me. At Gospel of Hockey Scott transcribed a radio interview with Lowe after the Visnovsky trade. Lowe thought he was getting a powerplay specialist. With the benfit of hindsight, Scott probably would have renamed the article "Blind Squirrel Finds Nut".

It's not easy being an Oiler fan, Sunny.

11/02/2009 6:04 pm  
Blogger Sunny Mehta said...

Haha, it's not easy trying to negotiate the world (and people in the world) using foundations of logic and reason, is how I would've said it.

(Ps - I missed you, Vic. Stop spending so much damn time away from the internet. And for fuck's sake are you gonna get those awesome playershots, playershotstied, teamxfaceoffs, etc scripts working for this season's stats or am I gonna have to work on my fucking programming skills?) :)

11/02/2009 6:23 pm  

Post a Comment

<< Home