Saturday, November 25, 2006

Strengths of Divisions

Just a simple check to see how this stuff is shaking out. By goal differential sans empty netters and shootout goals, here are how the divisions stack up.

Last year Toronto would have had a legitimate beef that the lopsided interdivisonal schedule cost them a playoff spot even with Belfour's poor season. I have a warm feeling that the same thing might happen to the Leafs again this season. Fingers crossed.

I would have guessed that the Northwest Division was still a fair bit stronger than the Pacific. By this measure it's about the same, this even with the two weak sisters, PHX and L.A, in the mix.

And how did the Atlantic turn to suck so fast? That surely can't all be pinned on Philly. Damn.


Blogger PDO said...

I'm not sure if this measure works this early.. how many divisional games have their been? Obviously any divisional game has a net impact of 0 at the end of the game.. if the NW has had twice as many divisional games as this point as Pacific, then it's not really a fair measure.

That make sense?

And, anyone else find it kind of weird that he NW has the most GF and 2nd most GA given the 'tenders in the division?


11/25/2006 12:35 pm  
Blogger mudcrutch79 said...

Is that ES numbers or something Vic? I added up the NW GF and it came to 309...90 odd ENG and shootout goals seems unlikely.

11/25/2006 1:06 pm  
Blogger JavaGeek said...

Just stick with: Sagrin Ratings.

Although, his ordering agrees with yours...

11/25/2006 3:08 pm  
Blogger JavaGeek said...

However, I used a matrix rank solver and got [includes strength of schedule]
Adjusted win percentages (division vs other division)
North East: 59%
Pacific: 54%
North West: 50%
Atlantic: 47%
South East: 47%
Central: 43%

11/25/2006 3:23 pm  
Blogger Vic Ferrari said...


Ya, all true, especially on the "it's early" bit. Still gives a decent indication. The NW has played 79 out-of-division games, the pacific 73. Normalize for the number of games and they are near enough the same in scoring rates and GA rates. NW leads the pack by the skinniest of margins over the pacific by that measure. As good as the NW goalies have been, Nabakov, Giguere and Turco have been pretty terrific as well.

11/25/2006 5:58 pm  
Blogger Vic Ferrari said...


Ya, I like Sagarin, interesting guy. Like a bookie from MIT. I've read quite a bit about him, he wrote a letter to an ESPN or SI journalist that had criticized him, it was priceless. He added on a postscript bit that I assume he didn't expect the recipient to publish in his next column, but he did :D good stuff.

I think I've got a very good idea what he has been doing for Mark Cuban over the past 6 or 7 years on basketball as well.

With hockey here he uses 4 as a base number, the difference between the last number (predictor) and 4 is his team value in more conventional terms. Usually you just read off of a table to get the gameline from there, based on historical results of teams running that diff. Though poisson on hitter-pitcher models and the like give you the same result anyways, within a % or so. In any case I suspect that he must be weighing recent games, because he's tied to pure math and can't nick teams for having injured players, or returning players, or trades and so on.

An off topic question though: have you seen anything on the likelihood of goals happening after own-zone draws? Something along the lines of "what happens within 20 seconds (or whatever time) of a faceoff in the defensive end"? ,,, from the POV of the defending team by this wording I suppose. It wouldn't be too tough to scrape that off the PBP sheets, but if somebody else has already done it, then I won't have to bother. That would be ideal. :)

11/25/2006 6:39 pm  
Blogger Vic Ferrari said...


Ya, I didn't include the head to head games between teams in the same division for those numbers. That's probably the difference, it would seem about right. Or maybe I made an error in the sums, I dunno.

To me the thing that becomes very clear, at this early stage, is that the Pacific is on the rise. But with the disparity in team quality in that division clearly Dallas (gets to play Phoenix 8 times!) has a softer sked than Phoenix (has to play Dallas 8 times :( ). In the Northwest it's all much more even, at least when Gaborik and Demitra are healthy it is.

It seems pretty clear that PHX, STL, CHI, L.A and CBJ are on the outside looking in. Probably one of these teams will catch lightning in a bottle and have a late season surge (my guess is L.A), and the strong teams in the PAC and Central should coast in unless the team planes crash. So, here we are in Novemeber and it looks like it's boiling down to 3 of 5 from the NW making the playoffs and just the seedings to be determined.

11/25/2006 8:17 pm  
Blogger JavaGeek said...

I can calculate those face off stats, but here's something to start with
21st Century Hockey Analysis Begins

It's a pain to calculate these things because you have to go through 74,000 faceoffs and see if there's a shot/goal after... You can't just write a quick only line SQL search...

11/25/2006 9:03 pm  
Blogger Vic Ferrari said...


Ya, and you have to separate out the game situations as well (pp, sh, ev, en) for it to be meaningful. I'll get around to it eventually I suppose.

The information in the link isn't useful to me, though it's cool to see more and more people taking a kick at this stuff.

A long while ago a heard or read a coach talking about this, he had a shitload of obscure stats but it was just for bantam and midget games. The goals in the shift that started with a faceoff in one of the ends was surprisingly high, and I was expecting it to be high. (As an aside: Everybody wants to be the next Hitchcock, they really should just let the kids play man on man and enjoy the game.) I assume that carries over to the NHL pretty well, hockey is hockey, and obviously bad things are much more likely to happen in your shift if it starts out with an own-zone faceoff, esp against Sakic or the like.

11/25/2006 11:21 pm  
Blogger JavaGeek said...

Vic Ferrari:

Using 8 seconds (2005-2006):
Faceoffs aren't recorded as EV/PP so I had to use my penalty prediction of PP/EV algorithm.
Even Strength:
win-shot: shots/faceoffs = pct%
Win-Shot: 4622/17334 = 26.7%
Win-Goal: 176/17334 = 1.02%
Loss-Shot: 698/17367 = 4%
Loss-Goal: 34/17367 = 0.2%
Power Play offense:
Win-Shot: 1399/5508 = 25.4%
Win-Goal: 74/5508 = 1.34%
Loss-Shot: 195/4460 = 4.4%
Loss-Goal: 4/4460 = 0.31%
Short handed offense:
Win-Shot: 122/616 = 19.8%
Win-Goal: 5/616 = 0.81%
Loss-Shot: 10/806 = 1.2%
Loss-Goal: 0/806 = 0%

- It's interesting but most shots after faceoffs are garbage (shooting % 1/2 of normal).
- PP faceoff win% = 55%. (Players who play more PP time will have high win %, PK players will be lower)
- shooting is high ~20%, even though unsuccessful.
- I hope there's no bugs in these results...
- The data is poor, it's possible for a shot to be displayed after a faceoff even if it occurred before.
- sorry no EN info...

11/26/2006 1:52 pm  
Blogger Vic Ferrari said...


Thanks, but that wasn't really what I was thinking of.

The general idea is that coming onto the ice with your team in possession of the puck is a lot better than the opposite. You're more likely to have a good shift, more likely to score a goal, and it's more likely for the next line for your team to come over the boards with possession as well. I think we all see that, and we all see the momentum swing in a hockey game that way.

Since that's damn near impossible to measure though, coming over the boards to a faceoff in the offensive end of the rink is a similar thing. And goals, as rare as they are, they are a hard measure. So if you're looking at shifts that start in the defensive end of the rink ... how much more likely is that to result in a minus than a plus?

I think I'll write something to run through the PBP sheets, starting at the bottom, and once it finds an even strength goal ... keep scrolling up to find the last faceoff before it happened. To see how long ago it was and which end of the rink. Makes sense, no?

11/27/2006 9:37 am  
Blogger Vic Ferrari said...

Okay, I wrote a little script to do just that, and it's interesting stuff. I just checked the Oilers for now, because they're the only team I watch regularly. So it all makes intuitive sense as well when meshes with memory (unless you imagined Peca taking all the key Dzone draws and Stoll playing tough minutes, in which case you should probably just clear your memory ;) ).

And, all other things being equal, it appears that you are very nearly 3x more likely to get a EV- than an EV+ if your shift starts with an own zone draw. And the reverse if it starts in an O-zone draw. And obviously the same should apply if your shift starts with possession.

My point is that this team needs a couple more Dvorak or Mike Johnson types. I'll get back to it one day when the mood strikes, whether it has a broad appeal or not, I personally think it's important.

11/27/2006 11:47 am  
Blogger JavaGeek said...

Face offs are interesting, because they have a rather short window where the results are different and then it jumps to the normal random process (pretty quickly), I personally estimate this time to be 15 seconds, but 1 shift is around 45.

I estimate the benefits of a offensive face off to be around 200 goals over the course of the season (above the normal process), Considering there are about 25000 offensive even strength draws, we can say 0.8% chance of getting an extra minus/plus from an defensive/offensive face off.

Since the largest difference for individual players offensive/defensive face offs around 400 face offs. So, I estimate the max cost/benefit to individual players who see a larger difference between offensive and defensive draws to amount to about 2 in terms of plus minus. Most players will see a difference of about 0.5.

When I say the effects are small, I mean it. Teams like Edmonton would be better off working on their power play, then trying to get a couple extra goals off the face off.

11/27/2006 7:20 pm  
Blogger Vic Ferrari said...

We really disagree here java. Frankly, and IMO, there is so much wrong with this comment that I don't know where to begin.

I know from your comments in a thread long ago that you think Jovo is a better defensive defenseman than Ohlund. (Gretz obviously agreed BTW, Jovo started the season playing tough minutes, and if it were The Masters our man Jovanowsky (sp?) would have been the odds on favourite to win the green jacket by the end of October, I assume that Barry Smith has taken over the bench there recently just going by the boxscores, but I haven't checked, and it's a subject for another day in any case). And from a comment by RiversQ in a thread here a few days ago that you think Hemsky is a very good defensive player.

You pick 45 seconds as the arbitrary shift length, which seems a bit fat to me for forwards at evens, but we'll run with it. In this case the over a quarter of the Oilers gaols came in the same shift as a draw in one of the ends, which is staggering considering that the vast majority of shifts do NOT start with a draw, and if they do it is often in the neutral zone.

And by this arbirtary 45-second shift rule of thumb ... Oilers collectively scored about two goals to every one they surrendered when it came at the offensive end of the rink, and the exact opposite at the other end (i.e. surrendered two goals for every one that they scored when they started the shift with a defensive zone draw). Of course this becomes emphasized much more when you have a guy out there just until the puck hopefully until you gain possession and get it going north, then Hemsky or Torres (both great defensive numbers!) or similar jumps over the boards in your place.

So when you say that the difference between the number of Horcoff's EV defensive zone draws (224 of them) and Peca's (83 of them) is immaterial, and bear and mind that most of the people who post here notice the level of competetition ...then forgive me if I say that that I think you've been smoking the fucking drapes.

Details: At evens, not including the last 2 minutes of the third period, the Oil were +13 and -25 in the 45 seconds after a defensive zone draw. And +24 and -14 after an offensive zone.

To add: The important thing here, imo, is that this little snippet in time where the NHL stats takers write everything down, it's indicative of the value of starting your shift with possession of the puck.

11/27/2006 9:31 pm  
Blogger JavaGeek said...

We really disagree here java. Frankly, and IMO, there is so much wrong with this comment that I don't know where to begin.
I'm not sure what part you're referring to in your response though, where did you begin?

- Go ahead and discredit me with my opinions on players from a while ago. Jovo doesn't take face-offs last I checked.
- 45 seconds is a sort of universal agreed shift length value (look at Nike's commercial).
-- The official SUM(icetime)/SUM(shifts) = 46 seconds (most goalies excluded) Of course you want EV, which is 42.4, so 45 isn't that off...
- Horcoff: About 2% of those 224 face offs would result in a goal, half of those goals occur as "normal process goals" the other half occur as extra goals. = 2.24 extra minuses. (not sure how many extra pluses because I don't know his # of offensive draws). We probably disagree here, I have reasons for this conclusion: that is to say I compared the time after the 45 seconds vs. the time before the 45 and the only scoring difference occurred when the team won the face-off in the offensive zone the absolution quantity of goals was 200.
- a 2% vs a 1% isn't a big difference in my opinion, but you see 2% = 1%*2 so twice as many pluses as minuses. I see a scoring rate that is higher in the first 15 seconds and then goes back down to normal after that.
- Your details agree with my theory +13, -13 normal process goals plus the -12 extra goals or
- +14, -14 plus 10 extra. Basically exactly what I'm saying.
- RE: To add: huh?
the vast majority of shifts do NOT start with a draw, and if they do it is often in the neutral zone.
- Not sure I agree, I have 35% of hockey occurs between 0-45 seconds after a offensive/defensive zone face-off. Of course you have to subtract the effects of pp and sh...
-- Considering there are about 75000 face-offs in 73800 minutes, there should be a lot of shifts with face-offs...

I'm not saying there's no effect, there are 200 goals as a result of offensive draw wins, that's a lot, but in comparison to other factors I don't know if picking up a player for just this effect is beneficial. Maybe better stated: I'm not sure how Dvorak or Mike Johnson would make Edmonton go from, for example, a 50% team to a 51% team...

11/28/2006 12:14 am  
Blogger Vic Ferrari said...


You're good at digging up data, you should stick to that. Mudcrutch doesn't have time to update that stats section of his, and I'm sure that's the kind of thing you could do easily, you should make yourself useful and do that.

Horcoff played against good comp on own zone draws, and he was -8 in that time, that kicks the shit out of your numbers. Kids like yourself that just count the numbers without recognizing what is really happening in the game, they are going to be mislead. If he's the guy that you're throwing out there for that gig then he's also the guy that you're throwing out there at the first chance when your current line is in "whoa fuck, finally a chance to get off the ice" mode and Modano is storming back the other way. And in fact he was, and still is.

Interesting that you say that Edmonton is a 50% team. The oddsmakers disagree (they had them at 16.5 points as an O/U for November in 13 games, granted the late goal in the ANA game makes they won't cover. Which makes them a smidge wrong, and you appear, on the surface, to be a little dumbass.

So while you confidently throw out data that says (indirectly) that the oddsmakers are fools, you seem to lack the courage to put your money where your mouth is. Why? Nutless? Just askin'. I guarantee you that if you use your pure analysis you will lose well outside the post (theoretical hold or margin), i.e. a guessing monkey will do better the vast majority of the time. Tis true. Try and use your own horseshit analysis to prove me wrong. Set your tiny balls square, javageek. Be a man.

And when you have time, you should help mudcrutch strip data from into MySQL, so his stats sheets update automatically. He seems to struggle with that, i know buggerall about it myself, but you're gold with that sort of thing.

11/30/2006 4:51 pm  
Blogger JavaGeek said...

Peca didn't take 83 D-zone face-offs (more like 160).

Anyway I liked this discussion and so I made a post on my blog that has a few more details

12/16/2006 8:39 am  

Post a Comment

<< Home