Friday, January 02, 2009

Corsi on the PK

HBomb and I proposed alternate theories in the comments to my previous post with regards to a team's corsi on the PK. While both of us had an opinion, we were both aware that we didn't have numbers to support ourselves.

So I went hunting.

The following picture shows the Corsi #'s for Edmonton's top penalty killers, sorted by minutes per game. 2008/2009 is on the left, 2007/2008 is on the right.

The difference from year to year is astounding. This season's best penalty killer, in terms of Corsi, would be last season's worst. Steve Staios is more than twice as bad. Matt Greene would be this year's best. ;)

All I can say is that this is a clusterfuck, and yes Corsi rates can change dramatically from year to year. I don't have a sense of context in that I haven't compared other teams to previous editions of their pk units or anything like that, but does this boggle the mind of anyone else out there? I'm sure I'm not alone. And I'm as sure as ever that there's more to it than faceoffs or quick fixes. Positioning and pressure levels, shot blocking and winning battles - every one of these needs to get better. On a MacT team, how has this even happened?


Blogger Jonathan said...

I'd be curious to see how the Fenwick numbers compare to the Corsi numbers - as it stands, those Corsi numbers are so bad that shot-blocking wouldn't appear to be an issue (because Corsi treats blocks the same as misses and shots that make it to the net).

The problem is that the Oilers are allowing too many shots; and since Corsi shows territorial advantage, I'd suggest this means that the opposition is simply spending a much greater percentage of their time in the Oilers' zone as opposed to last season.

1/02/2009 11:41 pm  
Blogger Lowetide said...

Question: What are the odds that this is just a terrible, random stretch and airtight days are ahead? I mean, we can talk about active sticks and faceoffs and positioning and not losing the box but it also seems like the Oilers are in the middle of a mountain of bad luck.

So, is it POSSIBLE that luck is responsible for a decent percentage of this?

Also, are the 3 goalies similar in terms of PK SP?

1/02/2009 11:42 pm  
Blogger HBomb said...

Christ almighty....I thought there might be a bit of difference but that it would be negligible.

Instead, we've got no clue as to know if the Oilers are worse at shot blocking or not, because it appears they're just giving up too many shots.

I think the idea of looking at the Fenwick number might be a start. And I'm also wondering how much of an effect the rule change to all PK's start with a defensive zone faceoff has had (I'd bet every team has seen some degree of PK Corsi drop....but are the Oilers in their own league of suckitude in this regard). Lowetide's suggestion is interesting, as this COULD be dumb luck, but from my eyes, the Oilers PK simply doesn't look as competent as in past years - to go from pretty damn good to downright godawful in one season is mistifyting. Were Stoll and Reasoner really THAT important? It sure as shit wasn't Matt Greene making the difference.

This really is an absolutely fascinating discussion.

1/03/2009 1:38 am  
Blogger Vic Ferrari said...

I think that Gabe might have used Fenwick numbers instead of Corsi numbers last year, Showerhead.

Note also that Gabe's "SHOTS" columns are actually "SAVED SHOTS" on all tables. I see people getting that wrong all over the place.

These two links will give the numbers for 5v4 PPs only, and with both goalies in their nets.

And you can get the 5v4 PP ice times from BTN, just multiply TOI/60 by GP.

Then if you rebuild your table it will still surely show an increase in shots against, but not an alarming change.

The PPsave% for the Oilers has fallen through the floor though. At 5v4 Roloson is a terrible .856, and Garon (one of the league's best last season at 5v4 puck stopping, .913! Damn, that was never going to hold up. I think Tyler pointed that out in the summer BTW) is a ridiculously bad .789 this year.

I caught a clip of Moreau on TV yesterday talking about the hot topic, the PK. His main point was that 'everything they shoot on net seems to have a chance to go in'. I think he is probably right. While the PK looks measurably worse in terms of zone time and not blocking as many shots and not winning as many faceoffs ... the big apple near the bottom of the tree is PPsave%.

Some of that save% downfall is the effect of the players on shot quality, but the vast majority is surely just shooters making their shots, the goalies missing a few that they shouldn't and some bad bounces around the net.

1/03/2009 7:39 am  
Blogger Vic Ferrari said...

Just to add, on a brighter note, the Oiler's 5v4 PP shots rates and Corsi+ rates (per game) are both up a whopping 23% from last season. That bodes well.

And sooner or later more pucks are going to hit the Oiler goalies when they are PKing. It looks like terrific save% may have masked some PK problems last season. And the 17% increase in 5v4 PP shots against per game, and 12% increase in 5v4 PP Corsi- per game are bad signs for sure, though always starting in your own end has surely had some effect leaguewide ... the save% is the killer here.

I think the smart money is rding with Lowetide's assertion.

1/03/2009 7:59 am  
Blogger Showerhead said...

Quick comment before I have to run off for a bit - I ran those links Vic and both appear to be for EV situations only. Gotta run but let me know if I'm wrong and I'll give it all another look tonight at some point.

1/03/2009 11:49 am  
Blogger Vic Ferrari said...

No Showerhead, I just changed the EV scripts for my own use so I never bothered to change the text at the top. Those numbers are for 5v4 with both goalies in.

1/03/2009 12:03 pm  
Blogger Showerhead said...

Just starting to look at the data now and it seems I have more questions now than before. First off though, thank you Vic for noticing and pointing out my mistake. And my apologies to HBomb for choosing the wrong numbers and skewing my argument. Whoops! :)

Now, on to your links Vic. I am looking at the 07/08 data and am having a hard time reconciling a few things. First off, the basics. Corsi should be everything but blocked shots, right? IE goals, iron, and saves? Fenwick the same but including blocked shots?

Second, I want to be sure and say again that this is indeed Edmonton's pk I'm looking at? Then why the positive Corsi/Fenwick for a guy like Horcoff or, heck, Nilsson? That part had me wondering if this was 5v5. Staios with Corsi of -437, so this must be totals as opposed to rates, this much is true?

When you have two columns under a heading, such as "missed shots", I assume the first is for and the second against. Is this correct? And under "EDM shots that were blocked" would the first column be Oilers blocking opposition shots or vice versa?

I can see the whole comments readership collectively facepalming right now as perhaps I'm missing a whole lot of obvious but I need to understand everything I'm looking at before I can make an informed decision/opinion on any of it.

1/04/2009 12:39 pm  
Blogger Showerhead said...

And finally, a thought on Lowetide's comment:

To start, it would appear that Vic has already proven you right with regards to luck. The save percentage is a question that's been in my mind and others have posted about as well (just today I see that Mudcrutch put up the numbers and indeed they're a lot worse than last season and perhaps we can expect a bounce back to at least average).

Second, it might speak to you having more perspective than me but when my eyes see systemic problems and the pucks are going in, I usually assume the systemic problems will continue. There's a chance that my mind is making up reasons why pucks went in after the fact but I still think that things look ugly enough (other than last night, incidentally enough) that there's got to be something there. This leads me to an interesting question though:

If positioning and pressure and stick positioning are all choices made by each individual skater on the pk... and each choice is made based on the circumstances they see in front of them... and let's say Oiler penalty killers are all very bright so they make the right choices 90% of the time... over enough repetitions there will be enough sequences of that 10% of wrong choices that a stretch this ugly could theoretically be a result of normal variation, no? Completely excluding luck with regards to posts, save percentages, and saucer passes that find their target more often than they should? Purely from a theoretical standpoint does anyone think this holds water?

1/04/2009 12:50 pm  

Post a Comment

<< Home