Wednesday, November 21, 2007

Driving Possession

This is a chart of all shots directed at net while a player was on the ice at even strength, excluding blocked shots. And since that takes a long time to say, I'll call these Fenwick Numbers from now on, since he was the lone lobbyist for this metric. This for all guys who have played three hours of EV ice time so far this year, and to keep it fair, everyone's ice time has been prorated to Staiosian levels.

It says a lot about which end of the rink the puck was at, and which direction the scoring chances were coming from.

Now clearly a guy like Hemsky is more likely to create high quality scoring chances, so he could be evens at this and still make a really positive difference. And a guy like Staios plays a lot of the shifts after Oiler PPs, has been out on the ice for a tonne of own zone draws, and has usually played a fairly high quality of opposition. And the same goes for most of the veteran players to one extent or another.

Regular Oilogosphere commenter Bruce recently quipped that Cogliano and Gagner create some great moments in the offensive end of the rink, and some bad minutes in their own end. There is a lot of truth in that methinks. And the people who agreed with him will feel vindicated by the graphic above. And considering that these two are the Oilers which are more likely to take an offensive zone draw than a defensive zone one, and MacTavish is loathe to play them against talent, and likes to get them on the ice with the puck moving north, and they are getting the primo "shifts after the opposition's PP" more often than not ... well they just aren't helping the Oilers win hockey games. Just aren't. The opposite in fact.

Terrific to watch though, if you can overlook the terrible giveaways going forward. And Gagner is barely 18 and has obvious skill, if you can play in this league at that age without embarrassing yourself, chances are you are going to be a damn good NHL player one day.

And like Lowetide, I'm really impressed with Cogliano, I think he's only a few years away from being a Horcoff type of player, in style and quality.

And while most of the Dmen are bunched together in the middle, Gilbert is driving zone time. He's like Pronger in the sense that he never dazzles you with a single play the way that Jovanovski does, he just never screws up, and like Pronger, because of this he separates himself from the pack in this measure. I'm expecting the bubble to burst to be honest, but lately he's been taking on more than his share of tough opposition and own zone draws, and he just keeps ticking along. Good night on the PP too in the last game. It's too early to start building statues, but damn, he's a rookie defenceman in this league and should be taking a pounding at this, especially considering the context of his ice time. But he keeps getting it done in an all steak and no sizzle kind of way. And even this stretch of results, I doubt that Eric Brewer will ever have a streak like it in his career. So even though I might regret saying this out loud; I think that the Oilers have fallen backwards into a real difference-maker here.

29 Comments:

Blogger Black Dog said...

Terrific stuff Vic.

I like your point about Gilbert. Just a really good hockey player in all facets of the game. He makes it look easy.

Here is a question - I know Cogliano is playing softer minutes for the most part but what do you think would happen if he and Nilsson were to switch - ie/ Cogliano got to play with Horcoff and Hemsky say. Also, are Penner's numbers a reflection of playing with Horcoff and Gilbert?

Thoughts? I guess breaking down each guy according to who he was playing with would be a lot of number crunching but I wonder what the result would be.

11/21/2007 12:57 pm  
Blogger Matt said...

Gilbert's results are nothing less than stunning. The difference between his !ROOKIE! numbers and those of the other D are absurd... I don't see how there can be any doubt that he's driving results.

11/21/2007 1:13 pm  
Blogger Vic Ferrari said...

Black dog:

Yeah, you're right with the idea for sure, and there is a pure way to do that and a couple of shortcuts. I'm just using Oiler numbers here though, and we all know the context of these guy's icetime. Matters less here.

So you still have to give Penner props I think, the Oilers have so few guys right now that can keep the puck low, he stands out, very different from the 05/06 that way.

And Nilson probably deserves more props than he gets from me. He's been getting icetime with good players for sure, I mean last night playing with 10 and 83 and not vs sedins ... that's a good gig. Still, considering he isn't exactly the king of keeping the play alive in the offensive end, the puck does keep finding a way to be in the right end of the rink when he's out there.

I dunno, I'll have to look for defensive zone play from him more. See if he's consistently coming back down to the hashmarks when he's the weak side winger, and how he's handling himself as the 1F when that happens. Frankly, I haven't really noticed him much, good or bad, so far in his own end.

11/21/2007 1:34 pm  
Blogger Vic Ferrari said...

Matt:

Yeah, like I say, I keep waiting for the bubble to burst. But he's been getting more and more responsibility, in the last 6 or 7 games he's gone from being an offensive zone faceoff guy to the opposite, and there just don't seem to be any cracks showing yet.

And it's not like he takes a bunch of chances and happens to be on a streak where the dice are rolling his way (the legendary 'his only problem is consistency!' NHLer). Gilbert plays a simple game, and the chipstack keeps growing. What can you say?

11/21/2007 1:39 pm  
Blogger Vic Ferrari said...

Matt:

Just to add, these results don't include last night's game. Gilbert was +10 by this measure last night. Tarnstrom was the same of course, nobody else was close.

And I think Dennis was right that he was the guy that MacTavish wanted out against the Sedins, just that with all the Oiler PPs the second option (Staios/Smid) saw a lot of shift-afters, so it worked out fairly even between those two pairs.

Surely at some point this stops being luck. And if we're not there yet, we're close.

11/21/2007 1:48 pm  
Blogger Steven said...

Terrific stuff as usual Vic. Pretty much jives with what I've been seeing on the ice (other than Nilson which you pointed out. I don't know if his sample size is really that large though).

I was watching last nights game and I kept saying "Great play Gilbert" while my cousin beside me kind of looked at me funny. The guy consistently doesn't make any mistakes and always seems to be able to take the puck away from the danger zones.

The more amazing thing to me is his Penalty Minutes... 0! You figure he would have taken some kind of hooking call after seeing the type of players he's been playing against the last little bit.

It's a stark contrast to watching Matt the Greene Giant over there who seemed to take penalties every single shift.

11/21/2007 1:52 pm  
Blogger Black Dog said...

vic - I'd say that nilsson has done a decent job, just from eyeballing I mean - as you say you don't really notice him and for the most part that usually means good things from a defensive point - he's not cheating and when it comes up his wing its generally getting out

I guess he figures he better be doing that right if he's not scoring

11/21/2007 1:54 pm  
Blogger Matt said...

Oh, Vic, you're there. He can play. The risk is no longer whether he can get results, it's that (being 30-ish games into his career) he forgets what he was doing to get them, i.e. the simple game.

Your "chipstack" analogy is apt here, I think: if you've played poker at all, you'll have seen someone who built up a nice stack playing a very smart, disciplined game... and then burnt it all off because they got away from that game, started making riskier plays, and just generally thought that their metaphorical shit couldn't stink... and then they can't stop even once they realize they're blowing it!!

Gilbert should be relatively immune from this hazard, though, on account of (A) his advanced rookie age, and (B) the fact that the media are too busy fawning over Gagner and Cogliano to pay him much mind.

11/21/2007 2:05 pm  
Blogger mc79hockey said...

Anyone think that the Oilers could get Gilbert onto one of those MAB contracts, the kind that guarantee him $5MM over 5 years, or whatever MAB got?

Good chart by the way, Vic. It's a shame that the NHL doesn't have this info extending back five or six years. For as long as "we", in the sense of those of us here who've been exchanging internet missives since 2002 or so have "known" each other, it's seemed like the bottom of the roster has been an Oilers strength. By my recollection, they've gotten good mileage out of those guys and the quality of the top sixers has kind of been up and down. I'd imagine if we looked back, we'd see the bottom end guys doing better than the guys at the top of the roster playing tough minutes (at least since the demise of the MGM line as a unit).

I'm assuming that guys like Horc, Penner, Torres and Hemsky are here to stay playing tough minutes - Penner excepted, they're cheap top end guys who can fight the real elite to close to a standstill. This team will be ready to compete again when the bottom end guys can swing those numbers around.

If I was plunked into Lowe's chair tomorrow, with all the mess that this team is in, I'd probably spend the season trying to identify who amongst the pending UFA's would be cheap and could hold their own tough minute wise, even if they added no "O". I'm guessing that the quickest route to respectability owuld involve doing that.

11/21/2007 2:58 pm  
Blogger dubya said...

Great post. I've been guilty of accusing some of overplaying the 'rookies suck' card...but this clearly illustrates that even GOOD rookies tend to get their ass handed to them.

As for Gilbert, he's the exception. He made a play last night in his own end where he was looking to make a pass up the middle but the lane closed. He waited to see if his man would get open (he didn't), looked off the forechecker, then changed plans and made a smooth bounce pass off the boards to an open man at center. Haven't seen anybody with that calmness/decision making in our own end since Pronger.

11/21/2007 3:12 pm  
Blogger MattM said...

Great post, but the thing that most strikes me is the range of that graph. -60 to +20? God the Oilers are a giant bag of suck this year.

11/21/2007 4:15 pm  
Blogger PDO said...

Great chart Vic, just fantastic stuff.

Two points:

#1) When does Hemsky start getting the counting numbers? Seriously? The guy is playing great hockey by my eye, and now the underlying numbers are supporting it too. He's shooting, he's chipping the puck in at the right time and making the move at the right time, and is just generally reading the play better.

#2) What's the advantage/difference of this to a Corsi number? I just don't see how the inclusion or omission of blocked shots does anything but change the sample size...

11/21/2007 9:36 pm  
Blogger Dennis said...

Ty, if you did that after last season, guys like Nichol and Devereaux would be the guys to target.

Seriously.

BTW, I think you're right about the Oilers always being stronger at the bottome. Just remember back to '01 for instance, with Horc-Murray-BG on the 4th line. That's pretty damn impressive. Of course that meant Weight was carrying around Zholtok and Cleary but still, that was a great 4th line and our "third" line, which was what we called it back in the day;), has always been good. You always had Marchant and Grier to build around and then Moreau showed up later on and Voila.

As for what the Oilers cna do for their bottom six right now, look at what Thor did last year, 21/24 EV, and what he's doing in the A so far this season, 11 goals and +7 in 19 games, and you have to think he could right the bottom six ship somewhat.

BTW, great stuff, Vic. Finally, I've been liking Gilbert for awhile now. frankly, I've liked him since his first game but as of late he's making almost all smart plays coming out of his own end. Little stuff like hanging onto the puck juswt a little longer so that the forechecker will pass him and thus allow him more time and space to make the breakout pass.

I guess it's because the counting stats aren't there and the everyday newspapermen wouldn't know a good player unless he was scoring 70 goals, but Gilbert is as good a story as there is on this team.

11/21/2007 11:28 pm  
Blogger Matt said...

What's the advantage/difference of this to a Corsi number? I just don't see how the inclusion or omission of blocked shots does anything but change the sample size...

My argument is basically:
1. The whole (or perhaps best) use of Corsi is to have objective figures that can be used as a proxy for scoring chances (what else are you using it for?).
2. A shot that is blocked is either
a) not a scoring chance at all, or
b) on average from a worse scoring area than shots/posts/missed shots

Yes it affects the "sample size" but that only means anything if what you are sampling is relevant to what you are trying to represent. You could include Penalties Drawn (or hell, faceoff wins/losses!), but I'm not sure that the connection to a scoring chance is as obvious as for shots/posts etc.

That's my opinion anyway, take it or leave it.

11/22/2007 11:15 am  
Blogger RiversQ said...

My argument is basically:
1. The whole (or perhaps best) use of Corsi is to have objective figures that can be used as a proxy for scoring chances (what else are you using it for?).


We've been over this a million times, but I think the true value lies in zone time.

2. A shot that is blocked is either
a) not a scoring chance at all, or
b) on average from a worse scoring area than shots/posts/missed shots


I don't even think that's true. (a) is definitely not true because that logic gets really slippery when applied to missed shots. (b) is a horrible generalization IMO. You can miss a shot from anywhere.

To me a scoring chance is a shooting opportunity from a certain area of the ice. Whether that shot is blocked, hits the post, the net, or misses shouldn't really matter IMO.

We're not going to agree on this. So far the most powerful argument has been that it doesn't really matter to the numbers, which supports neither argument and only lowers the sample size.

11/22/2007 12:56 pm  
Blogger Vic Ferrari said...

A lot of good comments here. I agree with the points everyone has added, so I won't argue.

And though I'd hate for this place to become a math forum, to respond to PDO's comment, here are the Pearson correlations of shots for/against while a player was on the ice to missed shots for/against (Fenwick) and blocked shots for/against and the sum of the two (corsi) respectively.

CGY 98.8% 97.1% 98.5%
DET 94.8% 92.7% 95.4%
EDM 89.0% 94.6% 95.6%
VAN 97.0% 94.8% 96.8%
COL 95.7% 91.0% 95.2%
MIN 93.6% 92.2% 94.7%
L.A 94.6% 96.6% 98.1%
ANA 95.3% 91.4% 95.1%

avg 94.8% 93.8% 96.2%
med 95.0% 93.6% 95.5%

The correlation is overwhelming in all cases. As the sample size gets bigger, we're at the quarter pole of the season now, these just keep getting stronger, and the blocked shots element really isn't adding much, if anything, at this point.

Over the course of a playoff series, you'd benefit from the larger sample size for sure. Go Corsi.

Over the course of a full season, probably just goals and posts is best. Because we have enough of a sample size that we don't need that other stuff, we are physically closer to what we're trying to measure (impact on creating and preventing goals) with just that.

I mean just looking at this stuff, I'm sure that if you went back to 02/03, and JUST looked at the guys who had a great "shots blocked while I was on the ice" ratio, compared to their teammates ... and of those guys, if you picked the ones that had relatively shitty counting numbers and EV+ and EV- rates that year ... you'd have made a list of guys who would have had much improved 03/04 seasons at even strength by the counting numbers and the EV+ and EV-.

That's flagrantly obvious to me now, looking at this stuff.

What's really spooky is some fucker pulled this idea out of the air. We've all watched a lot of hockey, has anyone ever heard anyone mention this stuff.

I mean right now Pronger is playing very tough minutes in Anaheim. And 71 Ducks shots have been blocked by the opposition when he's out there, the Ducks have only blocked 47 shots when he was on the ice.

I think that most Pronger bashers would use that stat to discredit him. But the fact of the matter is that it really shows just how good he is. And is surely a better predictor of where Pronger's even strength points, EV+ and EV- will end up in April.

There are some buggers with convoluted minds in Buffalo.

11/22/2007 12:58 pm  
Blogger Vic Ferrari said...

Edit for my preceding comment:
...
And is surely a better predictor of where Pronger's even strength points, EV+ and EV- will end up in April than any stat on TSN's website.
...

Also, the shots-blocked-while-on-ice numbers for Pronger are just for even strength ice time, of course.

11/22/2007 1:05 pm  
Blogger Matt said...

Good stuff Vic... I think we're zeroing in on a rough consensus:

I'm starting to understand RQ's stance better (though I still disagree). Like you say, over the course of a short sample (like a playoff series), a big imbalance in BS For & Against is probably indicative of an imbalance in scoring chances.

My objection would be using BS once you have a larger sample size. Take the case of 2 defencemen: Jason Smith and (imaginary) Dale Durgeon. They have exactly the same effect on a game (zone time for & against, etc.) with the notable exception that Jason Smith has shown a persistent and repeatable ability to block shots.

To me, this difference should be reflected in any metric that is trying to measure impact on creating & preventing goals -- it certainly shouldn't be *deliberately excluded*. Same goes for a player who is skilled at getting his shot through traffic, or makes good decisions about whether to shoot through traffic or pass/skate/try something different.

In other words, once the sample size is large, that's when it's most important to exclude BS.

Over the course of a full season, probably just goals and posts is best.

This is essentially correct... goals and posts will track *quality* scoring chances very closely in most cases... but at the same time, these metrics will be very useful too. Out of 750 players, there's going to be a non-trivial number who got lucky & unlucky. Boersma of Hockey Numbers once posted the 95% confidence interval for a goalie's "true" SV%, and it was surprisingly wide, even for 2000 shots against.

An awful lot of mysteries can be solved with some authority using this stuff... Owen Nolan says he's playing the same way, but now the pucks are going in: true or false? Is this 33-year-old starting to go downhill, or is it just a stretch of bad luck? etc.

Thanks for all the work, Vic.

11/22/2007 2:09 pm  
Blogger Vic Ferrari said...

A bit off topic I know, but I think Nolan is right. The underlying numbers are all very strong for him at evens, it's just the poor PP production from him so far that has hurt.

Looks like a guy the coach trusts to be out there for own zone faceoffs as well, granted CGY has a lot more offensive zone draws.

I never could stand Nolan, something about the look of the guy. He doesn't drive the reults like he once did, and he's near the end of his career, but he's still useful I think. Surprisingly good EV numbers on PHX last year as well, downright impressive considering the team and his injury history.

11/22/2007 2:36 pm  
Blogger Matt said...

I never could stand Nolan, something about the look of the guy.

Rudy at BoCali wrote last month that, "He looks like the kind of guy who would drive on the shoulder of the freeway during a traffic jam." I can't exactly disagree, although for some reason he's growing on me lately...

11/22/2007 3:01 pm  
Blogger Vic Ferrari said...

Matt

I should have said "goals, post & saved shots" above. Jesus, I really should have a read over before I post.

And yeah, I'm a big believer in the EVsave% behind you, relative to the goalie's avg, being mostly just luck.

We can all look at these things and rationalize them by the individual player, but the repeatability from year to year is piss poor. It's almost all coincidence IMO.

The top 200 NHL players, by EV shots against last season, they woulod almost always be the guys playing against talent, and are largely defencemen. They had a combined 916.8 EVsave% behind them, the rest had a combined 917.1 EVsave% behind them.

So for the guys that see less icetime, that's about one fewer goal against per 3000 shots, which is several seasons worth.

If you can spot that by watching the games ... you've got an eye for the game that are fucking magic. :D

11/22/2007 3:12 pm  
Blogger Bruce said...

A very interesting discussion, I am watching and learning, but for now I just have a deliberately naive question: how do you square the results of the graph with the following plus/minus figures:

Cogliano even (team-leading!)
Hemsky -4

I realize those are gross numbers, not 5v5. I'm sure EN GA and SH GA will have some effect on Hemsky's +/-, as will SH GF on Cogliano's, but the flow of play metric revealed by your study surely suggests there should be a large differential in Hemsky's favour which in reality does not exist. At least, not yet.

11/22/2007 4:04 pm  
Blogger Tyler said...

I think Vic has access to this Bruce but I'll suggest that it's probably to do with the shooting/save percentage for the Oilres when they're on the ice and I suspect that Cogs has that way more in his favour right now.

11/22/2007 4:33 pm  
Blogger Bruce said...

... which leads to the follow-up question, is the apparent lack of relationship between SF/SA vs. +/- due to hot goalies? dumb luck? or does it speak to the quality (as opposed to quantity) of scoring opportunities being created and allowed when the relevant players are on the ice?

11/22/2007 5:18 pm  
Blogger mc79hockey said...

... which leads to the follow-up question, is the apparent lack of relationship between SF/SA vs. +/- due to hot goalies? dumb luck? or does it speak to the quality (as opposed to quantity) of scoring opportunities being created and allowed when the relevant players are on the ice?

I won't speak for Vic but I doubt that there's enough data available to answer this at the moment. I would guess that there are aspects of all three but that if I had to guess which was most important in a given year, it's probably luck.

11/22/2007 5:47 pm  
Blogger Vic Ferrari said...

Bruce:

MC79 is right on all counts. Important to remember that unlike in Europe, the MHL +/- is a strange conconction that includes +s and -s for shorthanded and empty-net goals as well as even strength play.

PPs and PKs are obviously very important, and shorties count just as much on the scoreboard of course. But we're just talking about even strength play here, it would be a nonsense to not divorce them.

And of course the Oilers will be a struggling team this year, so the guys who are out there trying to score a goal with the goalie pulled will probably take a shitkicking to the tune of -10 just for that gig.

At evens, with both goalies in their nets, Cogliano is -2 (and he's a little bit lucky it isn't 1 or 2 worse) and Hemsky is 0 evens (a touch unfortunate I think, should be a couple better).

11/22/2007 6:46 pm  
Blogger Bruce said...

MC79 is right on all counts. Important to remember that unlike in Europe, the MHL +/- is a strange conconction that includes +s and -s for shorthanded and empty-net goals as well as even strength play.

Indeed, I alluded to this in my own "naive" comment. No doubt +/- is flawed as a stat, a PK poacher like Pavel Bure can score five shorties and be credited with +5, even if the other team scores 100 PP goals against him. And empty net goals drive me crazy with how they fuck up stats.

A two-way player like Horcoff (or Cogliano?) who plays both special teams and both EN situations will see such numbers balance out, but an offensive guy like Hemsky will always take a bit of a hit on the minus side, and a reliable PK and defensive type will always get a bit of a boost in the plus column.

Your mention of "unlike in Europe" is bang on; I had a look at the Elitserien web page a while back and was blown away by how superior it is to the NHL stats package. Just a whole different approach, actually applying logic.

At evens, with both goalies in their nets, Cogliano is -2 (and he's a little bit lucky it isn't 1 or 2 worse) and Hemsky is 0 evens (a touch unfortunate I think, should be a couple better).

Thanks, Vic. That seems about right.

11/23/2007 12:32 pm  
Blogger Dennis said...

Not really a surprise here, because we seem to agree on most things for some reason, but I agree with Riv's idea on what's a scoring chance, ie it's decided by where you find yourself with the puck.

Vic, I watched some of the Yotes/Ducks yesterday, Pronger was playing with O'Donnell and Schenider, speaking of guys I can't stand, was playing with Beauchemin. Then Huskins/Hnidy rounding out things.

My point was that the Ducks were in charge when Pronger was on the ice. Granted, he scored his ES goal off an O-zone faceoff that was won by Pahlsson and tipped back by Moen but when Pronger was on, so were the Ducks. BTW, something to watch for if you check them out, Sutherby played with Perry-Getzlaf and really did a nice job rounding out that trio. He's a guy I really liked as a Cap. He was 22/33 and played in the toughest top six min for the Caps last year, according to Desjardins.

He's not a guy that I openly pimped like say Nichol or Devereaux, but I got into watching the Pens and Caps last year when the Oilers really went into the toilet and Sutherby was both big and fast and he was also tough. It will be interesting to see what the Corsi numbers are with that new line. Granted, the other two kids are gonna drive the bus but I think Sutherby will certainly be a help.

Couple of other things:

- Marchant and Parros took the gaspipe in the third period. No surprise for Parros and the Ducks don't use their 4th line fuckall anyway and didn't have a chance to in this game becuase the Bad Miller got hurt, but I still thought it was odd that Marchant took the pine while guys like Sutherby played.

- Burke gave up a 2nd rounder to the Caps for Sutherby. Couldn't the Oilers have offered up Nilsson, or say Pouiliot or Jacques?

11/24/2007 7:38 pm  
Blogger Rick Marnon said...

I hope that they do not change the size of the net. Over the last two years there has been a significant increase in goals scored. This would be terrible for the game.

11/25/2007 1:40 am  

Post a Comment

<< Home