Thursday, December 20, 2007

Best In The Game


This is a photo of Zetterberg and Lecavalier from a game last month in Detroit. Vinny played sixteen shifts at even strength, all of them featured Zetterberg on the ice for some or all of it. Zetterberg played 17 minutes of 5v5 ice time, a cracking 15.1 minutes of that vs Lecavalier.

Of Vinny's 16 EV shifts, 11 of them came immediately after a Brad Richards shift, which is Tortorella's modus operandi from past years. Richards took ten draws in his own end, and one in the offensive zone. Lecavalier took 5 draws in the offensive end (all against Z's line) and 5 in his own zone. There just aren't that many offensive zone faceoffs available when you play Detroit.

Detroit won the game, in spite of the fact that Z had a rare minus night. -1 vs Vinny's line and -1 vs Richards. On this night Zetterberg, Lecavalier, and Richards all had Corsi numbers a touch in the red. They hurt each other's results a lot. And it's a certainty that if Babcock had gone old school with a checking line against Lecavalier, that both Z and Vinny would have looked a lot better to our eyes.

The reason I write this though, is that I saw a poll on TSN or Sportsnet the other day, one of these "who is the best player in the NHL" things. The poll of the hockey writers, one from each NHL city, ended up with Lecavalier as the runaway winner with 17 votes. Zetterberg got 4 votes, Iginla and Kovalchuk were next I think, with a couple each.

This surprised me, now maybe I have a Western Conference bias, but to my mind Zetterberg is a much better player. Last time I checked these two were nearly identical in goals and points, both are obviously terrific on the PP as well, and have the benefit of some good linemates. And I think that one of them makes a lot more difference to winning games than the other one does. So I'll attempt to build a case briefly here, with some context, numbers that represent the underlying results that really matter.

First up, shots for/against:

When Zetterberg is on the ice at even strength:
  • EV+ 30, EV -16
  • Shots for: 352
  • Shots against: 169
That's impressive, and the EVsave percentage behind him is a touch weak, and the EVshooting percentage of his linemates is good, but still a bit lower than you'd expect at 8.4%. i.e ... it isn't just a lucky streak, if anything he deserves to be even a smidge better.


When Lecavalier is on the ice at even strength:
  • EV+ 33, EV -29
  • Shots for: 261
  • Shots against: 260
This isn't new. Vinny was also on the ice for the second most even strength goals-against last season, of all the forwards in the league, only Martin St. Louis (72 GA) edged him out, and just by one goal. Admittedly the Lightning have poor goaltending, but 60 goals against at evens is the Bure line, these are madass big goals against numbers.

And now the faceoffs:

I wrote a php script a while ago to scrape info off of the play by play sheets. Since then I've been tweaking it to get different things. I'm getting a lot of mileage out of that crudely written page of code. :)

One thing I wanted to look at was changes in zone, the guys who are getting the puck going in the right direction. And since the NHL doesn't publish their zone time information any more, I just went by faceoffs. If a shift started in the defensive zone the the player gets a plus, and a minus for starting in the other end of the rink. Also if your shift ends in the offensive zone you get a plus, and you get a minus if your shift ends at the bad end of the ice.

If you do this for every player in the NHL, the worst player in the NHL by this crude possession metric is Vincent Lecavalier. A staggering -78. Also, Lecavalier was on the ice for 62 more offensive zone draws than defensive ones, the most favourable start to shifts of anyone on his squad.

The second best player in the NHL by this crude possession metric is Henrik Zetterberg, trailing only Bobby Holik. A whopping +80. Also, Zetterberg was on the ice for the same number of offensive zone draws as defensive ones, granted that's just average, but on this very strong DET team that's the toughest start to shifts of anyone on his squad. And a big part of the reason that everyone else on the roster starts more of their shifts in the offensive end of the rink is because the Zetterberg line keeps finishing there, and usually while playing against the other team's best.

My point:

As talented as Vinny is, as fun as he is to watch play the game, as much as he creates offensively, he gives almost as much back at the other end of the rink. And this isn't just a recent thing. Plus, if you're a Lightning player coming onto the ice as Vinny leaves it, chances are that shift won't start well. All of the opposites are true for Zetterberg, and this isn't a new thing either.

To me, Zetterberg is the MVP, and it ain't close.

.

Trivia #1: Lidstrom is +63 at this new made-up possession stat of mine (not including tonight's game) that is not the best defenceman number in the league though. To test your eyes, can you name the NHL defender who is edging out Nicklas? Hint: Plays for a Canadian team.

From the 'this can't be right' file: Marc Savard is +67 at this. And has been taking a ridiculous amount of own zone draws, and according to Desjardins has been playing against good players. I'll be damned, that's a leopard who I thought would never change his spots. Who's coaching the Bruins now, anyways?


Achilles heel: As good as DET is, their 10 through 15 forwards have been a gong show in every measurable way, much of what the Z line creates gets given back by this crew, and all with limited icetime.

Trivia #2: Red Wing tough guy Aaron Downey has been on the ice for 71 faceoffs at even strength so far this year. How many were in the defensive zone?

41 Comments:

Blogger HBomb said...

Hmmm, Trivia....

1) I'd take a stab and say Gilbert Gilbert, but I doubt the Oilers as a team are good enough to allow that in the first place, and he hasn't been THAT good. I'm going to take a stab and say Chris Phillips. I'll force myself to vomit if the answer rhymes with Leon Maneuf.

2) Downey in the defensive zone out of 71 draws? I'll say 5. But that might be high.

12/20/2007 6:44 am  
Blogger Black Dog said...

That's terrific stuff Vic.

1) I would say Phaneuf. They send him and Iggy out there together, no? When the draw is in the Flames' end they would get sent out there and likely the puck ends up in the other end of the rink nine times out of ten.

2) Zero.

Where does Richards fit in using these metrics? And what about Horcoff? With the exception of PP roles I would like to think that Horc and Richards are comparable type players. Am I way out in left field here?

Agreed on Zetterburg by the way.

12/20/2007 7:52 am  
Blogger Steven said...

Wow that's some great stuff once again Vic.

1) I'll go with Ohlund only because I refuse to believe it's a Flame.

2) Pretty sure its got to be zero.

I'm curious about the Oilers with this metric as well.

This stuff is pretty interesting to me because it's a pretty big given that everyone thinks Vinny is top 5 in the NHL and would give up the moon for him.

Zets on the other hand doesn't get as much press despite the fact his numbers are outrageously good and I don't think people would give up as much for him as Vinny.

I'm not sure if it's a style thing (Vinny is more fun to watch) or if it's because he looks more like a better hockey player (Bigger, more flashy, etc).

How does Crosby measure in this as he often gets compared to Vinny for the best player in the league THIS year.

12/20/2007 8:53 am  
Blogger Hawerchuk said...

Great stuff here, especially since Richards' 5v5 stats look so bad relative to Lecav's. Possession and zone information is clearly key to understanding player value.

12/20/2007 10:08 am  
Blogger Showerhead said...

1) I won't stand for it to be a Canuck or a Flame so I'll go east and suggest Andre Markov.

2) Somewhere < 7

3) This was a simply fantastic article, following up a string of great pieces. Although... it seems you just had to choose my final exam time to do this in.

12/20/2007 10:26 am  
Blogger Jeff J said...

Awesome post.

1) Joe Corvo
2) One, just because he was on for an icing call.

Of course, Zetterberg > Lecavalier. In Lecavalier's defense, TB lacks a defense. Zetterberg has Lidstrom and Rafalski behind him 2/3 of the time. I mean... wow.

I think the real comparison in the East would be Crosby, who has even less to work with than Lecavalier.

12/20/2007 11:10 am  
Blogger Jeff J said...

Oh yeah - Claude Julien is coaching the Bruins now.

Marc Savard's only in this role because Bergeron's out. Last year Savard racked up points while the 21 year old had all the tough minutes. All the talk about Savard being the league's Rodney Dangerfield was pissing me off because he wasn't earning those numbers the same way other real 1st line players do. If he keeps it up this year, I'll be a convert.

12/20/2007 11:18 am  
Blogger choppystride said...

This comment has been removed by the author.

12/20/2007 3:44 pm  
Blogger Matt said...

I'll guess Robyn Regehr... he was struggling earlier in the year, but seems to get sent out with Iginla vs. power a lot, and the results have (duh) been pretty good.

As to the Red Wings 4th+ line... that's interesting. If they're garbage, then fwds 4 thru 9 are obviously dominating. I've only kept a specific eye on Cleary, that dude is ALWAYS in the black. (And scoring on the PP to boot).

I'm really starting to think that Dan Cleary is going to be this year's UFA bargain, if the Wings let it get that far.

12/20/2007 5:20 pm  
Blogger Matt said...

Sorry, still guessing Regehr, but to restate my logic in a way that makes more sense on the question:
1. Flames take more O-zone than D-zone faceoffs
2. Regehr is more likely than anyone to be on for the D-zone ones, and less likely than anyone but probably Hale to be on for the O-zone ones.

12/20/2007 5:24 pm  
Blogger Lowetide said...

Trivia 1: I know it's wrong, but the defenseman I like best on a Canadian team is Toronto's Kaberle so I'll guess him.

Trivia 2: Sounds like a trick question so I'll say 0.

12/20/2007 6:25 pm  
Blogger choppystride said...

This comment has been removed by the author.

12/20/2007 6:43 pm  
Blogger choppystride said...

Trivia:
1) I think it's gotta be a Sen. As Philips is taken, I'll go with Volchenkov
2) Absolutely no idea. How about 1?

As for Savard, I remember when Hitchcock was in Philly, he commented that Savard (then in ATL) had become a much better player who was competing a lot harder than before. So perhaps the transformation started in ATL and the credit goes to Hartley?

As for the faceoff metric...I actually have a long response written. But upon re-reading your article, I'm not sure that I've understood your definition.

So to clarify, I would like to ask you this: do you examine:
i) only those faceoffs that belong to a player's shift which must both begins and ends in faceoffs, or
ii) all of a player's faceoffs no matter how a shift begins/ends (i.e. begin on faceoff then end by changing on the fly)

12/20/2007 7:28 pm  
Blogger PDO said...

1) Phillips... plays with the big 3 and goes PVP, and I think the big 3 are better than anything in Calgary.

2) Sounds like a trick question, so I'll say 25.

Question:

What are the chances that a defensive zone faceoff leads to an offesnive zone faceoff? And to a defensive zone faceoff? And I suppose to a neutral zone faceoff as well, just so we have all 3 together...

I have a loose theory about the flow of the game that may or may not explain the numbers a little bit and make Vinny look a tad better and Z a tad worse.

12/20/2007 7:40 pm  
Blogger Matt said...

I don't think Vic was trying to establish too much empiracal proof(!) here (reach beyond the grasp of the numbers), but anyway:

This metric is half about your results, and half about what your coach thinks of your abilities (good, bad, or varying). Someone who gets thrown out for a ton of Dzone draws is going to look good, or at least passable, by this measure, regardless of whether they are a good choice to be there or not.

Usually coaches are pretty smart about this stuff, but not always. Rating penalty killers based on how much SH TOI they have, or centremen based on how many faceoffs they take, will usually give you a good general idea of who's capable, but it will miss guys who are over- or under-rated by their own coaches at those skills. That said, it's interesting nonetheless.

12/20/2007 9:19 pm  
Blogger RiversQ said...

I'll go with Komisarek as the dman.

I'm sure Downey's had just a handful.

From an Oilers perspective, when Sykora played C last year I believe he had a grand total of 2 d-zone faceoffs in the first half of the year. Poor guy couldn't hack it for the Oilers last year despite their best efforts. I wouldn't say he was terrible, but it seemed to me it was either all or nothing with him. If they played him conservatively he barely crossed the other blue line, or he was just trading chances back and forth.

12/20/2007 10:31 pm  
Blogger Vic Ferrari said...

choppystride:

It is the second one ... All of a player's faceoffs no matter how a shift begins/ends (i.e. begin on faceoff then end by changing on the fly)

Just simple stuff, choppystride. I mean I think that, by and large, the guys that are sent over the boards for a large share of the D zone draws are the same guys being sent over the boards more often when the coach sees that the team is under pressure.

And the guys that are sent over for a lot of extra offensive zone draws are also getting the tap when their team has just had the bad guys pinned in their end for a long shift.

That's certainly the case with Jagr, Lecavalier and St. Louis in the games I've seen them play.

So if you're keen, you can see how shifts ended by typing this:
http://timeonice.com/xfaceoffs.php?team=CGY
That doesn't catch the shifts that end in goals though, for that you need to go here, and add them in.
http://timeonice.com/tshots.php?team=CGY

And for how faceoffs start, here:
http://timeonice.com/tfaceoffs.php?team=CGY

Just change the team name to suit.

If you look at the guys that are starting a lot in their own end, they are finishing a lot there too, on the whole. Opposite for guys who start a lot in the offensive end of the rink. And that makes sense.

So it is fairer to take that into account. And since this is just a bit of a general look-see at what is there, after five minutes of glancing at it, I decided that simply subtracting one from the other seemed fairest.

It certainly isn't bulletproof, but there is a thread of reason to it. Nothing more or less than that.

12/20/2007 11:05 pm  
Blogger Vic Ferrari said...

Trivia answers:

Downey has been on the ice for TWO own zone draws.

Cory Sarich has been sent out for a lot more defensive zone draws than offensive zone ones (a team leading 46 more). His shifts have ended in the offensive end of the rink 18 more times than the bad end.

I haven't watched enough Flames games to know why this is the case. Maybe he wasn't playing with Regehr when #28 was terrible early, but has been playing with him lately? Maybe he's just a very effective defenseman?

12/20/2007 11:16 pm  
Blogger Vic Ferrari said...

PDO: Fire away with the theory, the worst that we can do is criticize.

12/20/2007 11:18 pm  
Blogger Vic Ferrari said...

Rivers

The last time I checked the Oilers shots +/- last season was mid January, and Sykora was leading the Oilers at that point. Granted he was playing soft minutes after Christmas, I think he had a couple of brutal games in there, one was a wild affair against L.A, another high scoring one too, and MacTavish just lost faith in the guy.

I still like Sykora, smaller player than I'd realized though. And those years in N.J and then ANA really made him a passive player I think, his natural instinct was to fall back into position instead of pressuring, to a fault I think.

And we shouldn't forget that it was when him, Petersen and Hemsky started taking on some tougher competition that the Pisani/Stoll/Torres line started to have success.

12/20/2007 11:29 pm  
Blogger PDO said...

Vic:

Well, a big part of it would be a quality of competition thing. I mean, let's say you're Henrik Zetterberg, and you're playing the Oilers...

Would you rather start a defensive zone faceoff, with Gagner on the ice, or an offensive zone faceoff, with Horcoff on the ice? The position of the puck certainly matters... but wouldn't the quality of the players be more important?

A second, smaller point, would be that hockey games just to tend flow a bit, which would skew these numbers, I think. I mean, I don't have the means to check this, but I'd bet that if player X is on for an offensive draw and more than 45 seconds expire (average shift), the next faceoff is more often than not in a different zone. Simply because it's pretty rare for a team at 5v5 to keep a team pinned for a significant period of time, baring a few bounces or Joffrey Lupul trying to get the puck out.... isn't it?

I wouldn't be against laying a wager that most of the continual defensive zone draws happen in a 10 second window of each other, from a quick shot that is either A) stopped or B) gets deflected out of play.

Mostly though, it's the QoC. I mean, if Z is playing so many more defensive zone draws than offensive, and that's the way most of the true stars are played.. then they're also getting slightly easier QoC, because it's the other teams 3rd line taking the draw, right?

12/20/2007 11:39 pm  
Blogger choppystride said...

It is the second one ... All of a player's faceoffs no matter how a shift begins/ends (i.e. begin on faceoff then end by changing on the fly)

Thanks for the reply. That's the case I was expecting.

I certainly understand your reasoning behind the metric. But I see some issues. Anyhoo, here's what I wrote:

It seems to me that the concepts of (1) where you end up when a faceoff is called, and (2) where you are to start a faceoff are two different types of measures such that straightforward summation of their data points may not be totally suitable.

Type (1) is related to a player's ability to influence zone/possession time much like one of those shots-directed-at-net (SDAN) measures.

In fact, I would incorporate it into those SDAN metrics (Corsi? Fenwick?). Theoretically, this should strengthen the aim of those SDAN metrics by including additional "pressure points" that represent cases where the defending team is forced into d-zone faceoffs (i.e. icing) which did not result from a goalie stop.

As for type (2), its meaning is kinda fuzzy. Consider the following cases:

(a) player not on ice before faceoff called, but goes on ice to take the faceoff:

This is most likely a coach's decision and therefore should not be associated with a player's ability to influence zone/possession time (i.e. he did not force/yield the faceoff). This is a measure of a player's circumstance. In fact, I think this is really a toughness-of-icetime measure.

(b) player was on ice before faceoff called, then stays on for the ensuing faceoff:

Obviously, the two scenarios are:
- the player forced a faceoff in o-zone, then takes the faceoff in o-zone.
- the player yielded a faceoff in d-zone, then takes the faceoff in d-zone.

This is where things get really murky. Should the emphasis be on causing the faceoff or on taking the faceoff?

If I understand correctly, your metric simplying sums up the two sequential events, which means that they are nullified. I think this may create a bias against the offensive players and in favour of the defensive ones.

For instance, you're one of the team's better offensive players and your line just pressured the opposition into freezing the puck its own zone. If you continue on with your shift, according to the metric's formula, your good work would be negated. I can definitely see a big minutes offensive guy like Lecavalier who plays on a top-heavy team like TB suffers in these scenarios.

Conversely, the top offensive guys on a team with good depth will benefit. I think that's especially true if the team wins a lot of games by big margins.

As for the defensive players, if you're one of your team's go-to defensive players, you'll probably get a lot of freebie do-overs even as you keep getting pressed into d-zone faceoffs. And if every once in a while you manage to force an o-zone faceoff, there's a reasonable chance that your coach would substitute you for an attacker.

Or, looking at it another way: a player who begins his shift with a d-zone faceoff will never be tagged with a minus when his shift is over. OTOH, a player who begins his with an o-zone faceoff can never end up on the positive side of the ledger. The metric is implying that the "even up" outcome (the zero value case) of a play starting in an end zone will end up stopping in the same end zone. I think it may be underestimating the efficiency with which a defending team can clear the zone.

12/21/2007 7:09 am  
Blogger MetroGnome said...

Maybe he wasn't playing with Regehr when #28 was terrible early, but has been playing with him lately?

Ding! Ding! Ding!

You are correct sir.

12/21/2007 10:12 am  
Blogger Vic Ferrari said...

As you say, and Matt as well, it does err towards the guys who are sent out for more defensive zone draws.

The thing is, when you look at the guys who are ending their shifts in the offensive zone, there is an intuitively obvious connection with how much they are favoured to take offensive zone draws. And to get mathy with it, a Pearson correlation of 75%.

So it is impressive that 54 more of Jagr's shifts have ended in O-zone draws than bad ones. But the fact that he's been on the ice for a whopping 55 more O-zone draws than bad ones .. that takes a lot of the shine off of it. Though Jagr is still obviously a very good player. And if you've ever seen the way that Renney plays Jaromir, it would be damn near impossible for Jagr to end up in the red with Corsi numbers or shifts-ending-in-the-O-zone.

So while subtracting off the Dzone draws may skew things a bit the other way, it's much better than ignoring them. And with something that's really just a simple indicator, I didn't want to get to complicated with it. Multiplying the faceoffs taken difference (D.zone - O.zone) by 0.8 gives the best overall result I think. Just through trial and error.

But if you do that Zetterberg still finishes second overall and Lecavalier last. Nick Lidstrom does distance himself from the defenseman pack though. Which would have changed my trivia question to "Who rates second at this behind Lidstrom?".

So if you just look at 'where the shifts ended' ... it would look like several Flames forwards are better at determining the zone than Iginla. Subtract them and Iginla becomes the guy driving the bus at this in Cowtown. Use the "0.8 rule" and it's even moreso. In fact you could pretty much estimate how much the other guys played with Iggy just by looking at their number here, I'll call it the ZoneNumber. And since that makes intuitive sense, for this and other teams that I know a bit about, I'm comfortable with that.

It is what it is, choppystride. It is a general indicator, not a magic bullet.

I didn't know that Z would finish second and Vinny 700th when I ran them, but I knew that there would be a hell of a lot of separation, because Zetterberg does a lot of little things that help you win hockey games, and Vinny, for all his talent, just never has except for a good stretch in 03/04. And even then I predicted he'd fall back to his old ways. Leopards rarely change their spots.

12/21/2007 10:19 am  
Blogger Vic Ferrari said...

metrognome:

Thanks. I'm assuming that Regehr was hurt early, he was spectacularly bad by the early underlying numbers. So bad that it was surprising he was able to separate himself from his teammates so much. And that if he was anyone other than a star player he'd have been in the minors or the pressbox.

Then I check the other day and he's clawed his way up into a the black with Fenwick and Corsi numbers, and he's being played for defensive zone draws a lot more, and this zone-time number has soared as well. And he was in a really deep hole with this stuff not long ago.

Sarich is a very, very good NHL defender imo, but I find it easier to believe that Regehr is the guy driving it.

Hannan had a brutal startas well. Spectacularly so by every measure, and he's really started to turn it around in the last couple of weeks.

Pronger, on the other hand, has recently seen his underlying numbers start dropping like a rock.

I'm a big believer that it's injury related stuff, when guys get bad in a hurry, and it's not just the bounces. Especially when their icetime gets easier at the same time.

I remember in 05/06 we were theorizing why Pisani's line (Reasoner, Moreau mostly then) went from one that owned the puck and outshot everybody, they went from that to line that was getting owned. Then when the Olympic break came Pisani was one of the guys that never practiced, made sense that he wa hurt. I don't think that Dr House could have guessed colitis, though.

12/21/2007 10:33 am  
Blogger Bruce said...

Very interesting discussion, I've just caught up with the whole thread in one go. You guys are good.

Or, looking at it another way: a player who begins his shift with a d-zone faceoff will never be tagged with a minus when his shift is over. OTOH, a player who begins his with an o-zone faceoff can never end up on the positive side of the ledger. The metric is implying that the "even up" outcome (the zero value case) of a play starting in an end zone will end up stopping in the same end zone. I think it may be underestimating the efficiency with which a defending team can clear the zone.

Well said, choppystride, I had a similar concern when reading the original (excellent!) article. You start in your own end and finish in the other and get two plusses? And you're very right that the faceoff position at the end of a shift says a lot more about the guy's actual effect on the game than its position where he came out. So I was thinking how about a weighted average, and of course Vic was all over that.

Over-valuing where play begins favours good players on bad teams, or good lines and/or defence pairs following bad in the rotation. Lidstrom is "only" +63 because there's lots of times he steps over the boards and the puck's already in the offensive zone. So he's already -1. Better that he's -0.5, or even -0.8, but can still have a positive shift.

As for my man St.Louis leading the league in ESGA, a little context please? He also led the league's forwards by a wide margin (over 70 minutes) in ESTOI, and to some extent the two will go in lockstep. Where do MSL and Vinnie and Richards rank in ESGA/60? The facts will still be on your (and Zetterberg's) side -- the extra ice time also explains some of Vinnie's gaudy offensive numbers -- but you don't really need to paint the Tampa guys in such a bleak light to make your case. They're being asked to do too much, and for the most part they're actually doing it. Besides, I've seen better goaltending in table hockey games than Tampa's been getting these last years.

12/21/2007 3:38 pm  
Blogger Jeff J said...

- Wait a second. If you look only at EV faceoffs, doesn't every single shift that ends in the off/def zone also start another shift in the same zone? Looking at an entire team, shouldn't all the plusses and minuses negate each other - shouldn't it be zero-sum? And if it is... doesn't this number just tell you how things are going within a team, and maybe which teams are top-heavy? Maybe I'm just missing something here.

- Taking a closer look, some teams are, as a whole, in the black (Leafs, +46) and others are in the red (Habs, -26). That's even if you exclude goals. Could the discrepancy just be stoppages for coincidental minors in the O/D zone?

- So, if it's zero sum, perhaps a way to compare across teams would be to count shifts that start and end in the same zone as non-zero. Starting in the Off zone and ending in the Off zone is a good thing while starting in your own zone and finishing there is bad. Your tally counts both as zero. Maybe just counting faceoffs? By this measure, the Wings have 1.6 Off zone draws for every def. zone draw, compared to 0.8 for the Oilers. It's something thst can be taken to the player level too.

- Again, if it's zero-sum, the Lecavalier number goes from surprising to mind-blowing. How is his line so much worse than the others on TB? Is it the ridiculously long shifts?

- Bruce: "You start in your own end and finish in the other and get two plusses?"

I think this makes sense. If you start in your own end and finish in the neutral zone, it's worth only one plus.

- Just an observation: The ratio of shifts ended to shifts started on a faceoff is around 0.9 for teams as a whole* - the difference is due to period-starting faceoffs (which happen in the NZ so the metric should still be zero-sum). It can very widely from player to player. For Crosby, it's 0.8, for Malkin it's 1.0. Horcoff is 0.77, Cogliano is 1.08. Generally, the top guys have lower ratios and the kids and fringe players have high ones. Probably because the big guns start periods and start after every TV timeout. Somehow, Zettersyuk is high - 0.98. Gagner is 0.97. It just can't be for the same reason. Is it possible that the Wing's big line outshoots everyone so drastically that it cuts down the number of times they have to change on the fly?

12/21/2007 10:33 pm  
Blogger choppystride said...

VIC: The thing is, when you look at the guys who are ending their shifts in the offensive zone, there is an intuitively obvious connection with how much they are favoured to take offensive zone draws. And to get mathy with it, a Pearson correlation of 75%.

So it is impressive that 54 more of Jagr's shifts have ended in O-zone draws than bad ones. But the fact that he's been on the ice for a whopping 55 more O-zone draws than bad ones .. that takes a lot of the shine off of it. Though Jagr is still obviously a very good player. And if you've ever seen the way that Renney plays Jaromir, it would be damn near impossible for Jagr to end up in the red with Corsi numbers or shifts-ending-in-the-O-zone.

So while subtracting off the Dzone draws may skew things a bit the other way, it's much better than ignoring them. And with something that's really just a simple indicator, I didn't want to get to complicated with it. Multiplying the faceoffs taken difference (D.zone - O.zone) by 0.8 gives the best overall result I think. Just through trial and error.


75% correlation does not seem surprising. But for what we're trying to measure here, I think this may not be the most relevant stat.

Instead, I think what we need to know is: when a guy takes an O.zone draw, what's the probability (i.e. league avg) that he'll end his shift with his team winning another draw in the O.zone (or scoring a goal)? Unless I've misinterpreted your statement, this is not the same as correlation.

We have the formula:

ZoneNumber = (O.Zone.Earned - D.Zone.Yielded) + (D.Zone.Faceoff - O.Zone.Faceoff)

For simplicity sake, let's look at just the O.zone part:

O.Zone.Earned - O.Zone.Faceoff

It's basically saying that: any O.zone faceoff that you take would grant you a big advantage in earning another one. Therefore, for every O.zone faceoff that you've taken, we'll deduct an O.zone faceoff that you've earned in order to balance your score.

But is a 1-to-1 deduction ratio fair? You surmised not and suggested a fudge factor of 0.8. This becomes:

O.Zone.Earned - 0.8 * O.Zone.Faceoff

Personally, I think that the fudge factor should be equivalent to the aforementioned probability number.

With this in mind, does 0.8 look right?

Putting it another way: if a guy takes a draw in the O.zone, is there an 80% chance that his team can force another draw or score a goal?

I don't have the hard data to back it up but it does seem like that the correct number should be a bit lower.

Of course, the same logic applies to the D.zone part.

I know that you would like to keep the indicator simple and I certainly understand as I'm always too lazy to even attempt any such analysis on my own. But I think in this case, the fudge factor should probably be fudged a bit lower still ;-)

12/21/2007 10:39 pm  
Blogger Vic Ferrari said...

choppystride:

I'm not following your reasoning. And anything more than a 20% knick on D zone draws and it will just get worse.

Really, the numbers in the original post are the best representation, with the caveat that you should discount the guys who took a bundle of own-zone draws by a shade. Having said that, it doesn't apply much to either of the players that were the subject of the post anyways, just a bit of an effect on one of the trivia answers.

If the origianl post is worded like this does the perception change?:

Zetterberg finished his shifts in the offensive zone 73 more times than he started there.

Zetterberg finished his shifts in the defensive zone 7 fewer times than he started there.

Lecavalier finished his shifts in the offensive zone 16 fewer times than he started there.

Lecavalier finished his shifts in the defensive zone 59 more times than he started there.

.

It's a matter of putting a number on an aspect of Z's and Vinny's games that is painfully obvious. In as simple and rational way possible.

12/22/2007 12:01 am  
Blogger Vic Ferrari said...

Jeff J said:

Just an observation: The ratio of shifts ended to shifts started on a faceoff is around 0.9 for teams as a whole*

Good find. I hadn't really thought about it, but I suspect it's because if a shift ends in a goal, the faceoff after it is in the neutral zone.

And also if a shift ends in a penalty, then the following faceoff is during a PP, so my script ignores it.

That would account for more extra EV shifts ending in one of the ends of the rink than starting there, wouldn't it? About 3 or 4 inside the bluelines per game ... seems reasonable, no?

12/22/2007 12:48 am  
Blogger Vic Ferrari said...

Bruce:

I don't see how the ice time totals make much difference here, unless your arguing that Vinny is getting tired.

Still, the fact remains that he gives back almost as much as he creates while generally being put on the ice in most of the offensive situations.

The EV+ and EV- rates from this year, and the fact that they echo last year's results for both players was probably enough on it's own. But I put up the shots numbers to show just how wide the gulf between these players is. Z could have a drunken Racicot behind him and he wouldn't see those GA rates.

The zonetime stuff is more about the "how" than the "how much", which isn't my usual interest, but it illustrated it pretty well I think.

If someone is a truly committed fan of Lecavalier, or Jagr, or Bure, or whomever, then there is no point in arguing. But if you are flipping through the channels and see one of those teams in action, and you stop for a moment and pay attention to the way that a lot of Z's shifts end and the way that a lot Vinny's shifts end ... you can't miss it.

This reminds me of some of the Forsberg vs Naslund for MVP debate of a few years ago. One of them did everything to help his team win except sharpen the skates, and the other scored a lot of goals.

12/22/2007 1:08 am  
Blogger Bruce said...

I don't see how the ice time totals make much difference here, unless your arguing that Vinny is getting tired.

I'm just saying if you';re going to cite a (truly!) gross number like ESGA, as you did for Lecavalier and MSL, that of course the excessive ice time they are getting will have an impact on that. If they are getting 25% more ice than Zetterberg and giving up twice as many goals, not enough as an impact to affect your argument, it just seems like quoting gross numbers instead of your usual /60 is piling on a tad.

Zetterberg finished his shifts in the offensive zone 73 more times than he started there.

Zetterberg finished his shifts in the defensive zone 7 fewer times than he started there.

Lecavalier finished his shifts in the offensive zone 16 fewer times than he started there.

Lecavalier finished his shifts in the defensive zone 59 more times than he started there.


Now that is a convincing case.

12/22/2007 7:17 pm  
Blogger Vic Ferrari said...

Bruce:

Really that's just a different, slightly more detailed, way of presenting the same information. But I agree that it makes my point more intuitively obvious.

As for the ice time argument. Unless you're arguing for Zetterberg I have no idea why you would bring it up.

Z plays 22.8 minutes per game on average, Vinny 22.9 per game. Nothing there on the whole.

Z plays an extra shift on the PK on average, especially if there are 5on3 PPs to defend, where he is Babcock's first choice. Vinny plays about 45 seconds more EV minutes per game, and I'm sure that if you looked (and I haven't) you'd find that averages close to one more of the primo shift-after-the-PK shifts per game than Z.

So if you prorate Z's EV icetime up a whisker to Vinny's, it separates them a smidge further. But since the gulf is plenty wide enough to start with (in terms of creating more than they surrender), it's hardly worth sacrificing clarity to punish Vinny even further.

Thing is, all this is available at NHL.com. It takes seconds to check Bruce, as long as it took you to type that "maybe Vinny plays 25% more!" you could have used that time to check and see that this wasn't the case. And that's a reasonable expectation from sphere commenters IMO.

12/23/2007 12:13 pm  
Blogger Bruce said...

Thing is, all this is available at NHL.com. It takes seconds to check Bruce, as long as it took you to type that "maybe Vinny plays 25% more!" you could have used that time to check and see that this wasn't the case.

Thanks for the lecture, Vic. I feel duly chastised.

If you took the time it takes to write: "And that's a reasonable expectation from sphere commenters IMO" to actually read what I wrote, you'd see that I was writing about gross ESTOI in 2006-07. You were the one who brought it up -- Vinny was also on the ice for the second most even strength goals-against last season -- provoking my initial and follow-up responses.

So I followed your advice and did look it up, and here's the gross ESTOI:

Lecavalier 1340:19
Zetterberg 874:37

So Vinnie actually had 53% more ES minutes last year, and I actually understated the margin by quite a bit. On a per-game basis, accounting for Z's injury, Lecavalier still had the edge by 18%, 16:20 to 13:53, so I was still in the ballpark. The point being Vinnie and Co. had an outrageous amount of ice time last year -- Tampa had four guys over 16:00 ESTOI per game, the entire rest of the league had ZERO such forwards -- and of course it affected all of their gross totals, for and against. It's not rocket science. So if you cite a gross total that happens to suit your purpose in making an otherwise perfectly good case, it's apples and oranges and I'm going to call you on it.

You're right I should've been more thorough, or perhaps more clear on my point, but I'm not so completely out to lunch as you suppose. I'm no particular fan of Vinnie, either, btw, I think Zetterberg is a far better player. But you knew that already. Along with everything else.

12/24/2007 1:14 am  
Blogger MetroGnome said...

Vic:

Regehr's season has been a strange one. He started out playing with Aucoin a lot of the time at ES and wasn't even facing the tough assignments a lot of the time...and he stunk (so did - and does - Aucoin). In addition, his penalty killing was abysmal. To be fair, Kippers SH SV% was something like 66% after October, but there were doubtless team effects contributing to that number. Including, but not limited to, Regehr's struggles.

Oddly, he really started to turn it around after getting a puck in the ankle and suffering a deep tissue bruise. He missed a period or two, then came back during the road trip and looked like the Regehr of old. He was (finally) consistently paired with Sarich and they've been the best pairing since.

12/24/2007 11:50 am  
Blogger rananda said...

am i the only one that doesnt think this stuff is particularly great? the overall point, that zetterberg is a better two-way player than lecav is obvious enough and though i think youre right, i dont think your argument is all that convincing. the most glaring problem is that zetterberg's ev goals f/a and shots f/a are clearly inflated by playing so much with lidstrom and datsyuk (who are actually probably even better players than zetterberg himself). contrast that with lecavalier who gets to ride shot gun with st. louis and i imagine kuba, good players to be sure but no niklas and pavel. i dont see how you can separate zetterberg's success from the brilliance of his teammates, not to mention the style and strategy employed by his coach.

i'll also add that your shift direction stat, no doubt impressive in your ability to obtain it, is probably the most useless stat i've ever seen in the sport. just because we have the ability to quantify and monitor a stat, it doesnt mean it tells us anything. tom benjmamin once made this point with respect to faceoff winning percentage, and i think the point is infinitely more salient in this context. is there any way you could prove that players who get the puck going a particular way when their shift ends (it's hard not to laugh as i type the words and imagine that people actually think this is determinative of anything relevant or useful in hockey) are actually helping their team win by doing so? it could be that lecav spends the majority of his shift in the o-zone and leaves the ice when the puck is harmlessly cleared down the ice. it could also be that lecav and torts know their team isnt very good and know that lecav has to try to exploit any scoring opportunity they have, thus he stays on anytime the puck is going forward and only goes off when the puck is going backward. to try and use this silly stat to impugn lecav's game is beyond laughable. i guess it's easier for you to do this than actually spend any time watching the guy play.

12/27/2007 4:37 pm  
Blogger Vic Ferrari said...

I'm a bit disappointed that none of the guys in a heavier weight class tackled the obvious, that Iginla is driving this stuff for the Flames, as well as the PP, as well as scoring, as well as the underlying stuff, and with lesser teammates. Dude is a beast.

Instead I get this rajeev/rananda nutter chewing up hundred of words of precious internet space (which I haven't read yet, but probably will eventually). Jesus, now i know how the Cino boys feel.

MG: Thanks for the info on Regehr and the CGY D. I like Sarich as well, I started paying attention to him in a few games last season after a rare knowledgable Isles fan had a jones going for the guy as a potential off season target.

12/28/2007 2:08 am  
Blogger Vic Ferrari said...

Bruce:

That's more like it, now you're bringing something to the table.

Of course Zetterberg missed a chunk of last season, and Lecavalier did not. Of the guys who played regular top minutes, did anyone have a worse GA rate, i.e. EV GA per hour of EV ice time, than Vinny? Get back to me on that.

See how annoying that is Bruce? :D

I remember hearing Feaster on The Team last season, a good interview as always, and he spoke about how Lecavalier had grown as a player, how he was doing all the little things. Getting the puck in deep at the end of a shift, staying high if he was the last forward going in, grinding out the play if his teammates were out of position ... Feaster can talk, he's a guy to follow, but this was a load of crap IMO. And it got on my tit a bit.

You're a good guy though, Bruce, sorry that I offended you. Rereading that, my comment above was a bit pissier than it needed to be.

12/28/2007 2:22 am  
Blogger Hawerchuk said...

Just to add to your statement that the Lightning have poor goaltending:

http://www.behindthenet.ca/blog/2007/12/zetterberg-vs-lecavalier-defensively.html

12/28/2007 11:08 am  
Blogger rananda said...

i am a nutter because i question the validity of using a group of stats that are team based (whether or not z's line gets a shot against or for has a lot to do with what his linemates do, no?) without mentioning how to control for such linemates?

think i misunderstood your shift direction stat as i thought you were including changes on the fly and not just faceoffs, so any allegation of me being a nutter re that and only that part is probably justified. that makes infinitely more sense. the former would be completely insane, the latter is actually interesting and could be of use. i think it's still a bit of stretch though. i wonder why holik has such a high number, i cant imagine he's getting nearly as many o-zone draws and d-zone. and i would think the opposite is true of lecavalier. again this is a team derived stat used only in the individual context, very problematic. it's probably something that's useful, but it is as near useful as g, a, evp/m, corsi numbers?

12/28/2007 5:22 pm  
Blogger Bruce said...

Hey Vic, thanks for the response. This was a "hockey play", you took a run at me and I brought my stick up, let's take our coincidental minors and move ahead with the game.

Of the guys who played regular top minutes, did anyone have a worse GA rate, i.e. EV GA per hour of EV ice time, than Vinny? Get back to me on that.

Well, I actually asked that question first, about halfway up this thread:

Where do MSL and Vinnie and Richards rank in ESGA/60?

Not sure where all to look for the numbers, I know nhl.com has ice time for last season but good luck finding anything more than a net +/-. I'd be willing to bet that Joffrey Lupul and his -28 in limited ice time would have a worse GA/60 rating than pretty much anybody who played a couple hundred minutes. I'd also bet Ilya Kovalchuk would have bad defensive numbers.

So rather than just whine about it I went to Desjardins site and checked, and sure enough for the current season he has a very nice list of sortable numbers including 5v5 GA/60. When I limit it to guys with 20 GP to lose the Brian Youngs of the world, the very worst defensive player in the NHL would seem to be (drum roll) Ilya Kovalchuk, another "MVP candidate", at an execrable rate of 4.26 GA/60. Linemate Todd White is almost as bad (without the offensive upside), then the third worst is one Ryan Smyth at 4.14. The Tampa guys are ranked 7th, 14th and 22nd worst on a per 60 basis, with Richards the worst defensively at 3.86 GA and by far the worst offesnively at only 2.08 GF/60. The other guys at least score as much as they give up, but BR is not earning his $7.6 MM if that league-worst -20 means anything. And sorry, Vic, I don't care how many faceoffs he takes in his own end, that's just brutal. Is his idea of clearing the zone asking the ref to pick the puck out of the net and carrying it to centre?

One more thing: when you anointed Zetterberg MVP based on a head-to-head comparison with Vinnie, did you conduct a similar comparison with, say, Dany Heatley? How about Jarome Iginla? Those guys strike me as contributing far better two-way hockey to their teams than Lecavalier or Kovalchuk do. The Heatley comparison should be particularly interesting, since both play on great lines and on great teams.

I'll give you a head start. According to Desjardins Zetterberg is +3.90/-1.89 per 60; Heatley +5.29/-2.25. Which by one crude measurement makes Heatley a goal a game better than your MVP. Z's linemate Pavel Datsyuk (+4.40/-1.69) also looks pretty darn good in this metric, as does my personal half-season MVP, Nicklas Lidstrom (+4.10/-1.43).

As for last season, Desjardins has some of the same stuff but it's not sortable, so uh, let me get back to you on that.

12/29/2007 3:49 am  

Post a Comment

<< Home