Thursday, October 04, 2007

The NHL's divisional divide
How the schedule affects the standings

As preseason predictions have come in the past few days, there's been a little bit of hysteria lately. In Montreal and Vancouver, they're taking my name in vain. Toronto's planning a parade, and the fans in Florida haven't yet materialized.

The wonderful Edmonton folks, meanwhile, are telling me 11th place (!) is too kind.

It's true, my picks look a little nutty. And certainly different than most of those out there (Golbez is going to post a list soon). But what I was going with was the notion that division play affects the standings so much so that it actually makes these predictions easier, not harder.

For one, with 40 per cent of every team's games played within their own division, almost without fail, two or three teams will make the playoffs from each division.

This season, that likely means two from the Central, three from the Pacific and three from the Northwest. And three from the Northeast, three from the Atlantic and two from the Southeast.

So, even if you hate Detroit's competition in their division this season, one of Nashville/St. Louis/Chicago/Columbus is displacing one of Calgary/Colorado/Minnesota/Vancouver in the postseason. I'd bet big money on it.

Four teams making it from one division is an anomaly, an improbability and a near impossibility (I sound like Jackie Chiles here). It happened last season with the Atlantic only because Philadelphia, the worst team in the league, won just five of 32 games (15%) within its division.

That's not happening all that often.

What that means is that teams in strong divisions are going to pickup more of their points outside of their division, where parity will rule. Teams in weak divisions, meanwhile, will have a harder time winning outside of their division and pickup many of their points from division play.

A look at the points breakdown by division:

Rank Divisions Points from Div Games Points % from Div Games
1 Southeast 182 434 41.94%
2 Atlantic 182 454 40.09%
3 Central 179 448 39.96%
4 Pacific 176 459 38.34%
5 Northwest 176 471 37.37%
6 Northeast 176 475 37.05%

Eliminating the non-playoff teams from the equation opens it up even more:

Playoff teams

Rank Divisions Points from Div Games Points % from Div Games
1 Southeast (2) 83 190 43.68%
2 Central (2) 95 223 42.60%
3 Atlantic (4) 165 398 41.46%
4 Pacific (3) 129 324 39.81%
5 Northwest (3) 114 305 37.38%
6 Northeast (2) 80 218 36.70%

It was simply easier to make the postseason last year in the Southeast, Central and Atlantic, and the Northwest was incredibly difficult even with the Oilers tanking the final quarter of the season.

Here's a team-by-team look at points obtained in divisional play. Standings projections should be heavily based on changes you see in division dynamics — not simply which team you think is 'better' in the conference.

Unfortunately, it doesn't work that way anymore:

Rank Teams Points from Div Games Points % from Div Games
1 Dallas 49 107 45.79%
2 New Jersey 49 107 45.79%
3 Carolina 40 88 45.45%
4 Atlanta 43 97 44.33%
5 Detroit 50 113 44.25%
6 Tampa Bay 40 93 43.01%
7 Chicago 30 71 42.25%
8 NY Islanders 38 92 41.30%
9 Colorado 39 95 41.05%
10 Pittsburgh 43 105 40.95%
11 Nashville 45 110 40.91%
12 Minnesota 42 104 40.38%
13 Washington 28 70 40.00%
14 Montreal 36 90 40.00%
15 Anaheim 43 110 39.09%
16 Ottawa 41 105 39.05%
17 Calgary 37 96 38.54%
18 Boston 29 76 38.16%
19 NY Rangers 35 94 37.23%
20 Florida 31 86 36.05%
21 Columbus 26 73 35.62%
22 Los Angeles 24 68 35.29%
23 San Jose 37 107 34.58%
24 St. Louis 28 81 34.57%
25 Buffalo 39 113 34.51%
26 Phoenix 23 67 34.33%
27 Toronto 31 91 34.07%
28 Vancouver 35 105 33.33%
29 Edmonton 23 71 32.39%
30 Philadelphia 17 56 30.36%

The average NHL team picked up just more than 35 points from divisional games, and no playoff-bound team had less than that last season. (N.B. The average team had 91.3 points in 2006-07. The average Western Conference team had 91.9.)

Division play is why Toronto missed the playoffs, why San Jose didn't win its division, why the Islanders made the playoffs and New Jersey won the Atlantic.

Thirty-two games is an awful lot in an 82-game season, and if you can't win in your division, you can't win.

(This is probably all for me today ... I'm out with the flu and these meds aren't conducive to wielding a spreadsheet.)


At 1:19 p.m., October 04, 2007, Blogger Pete said...

Wow, at first you just had a good insight about why teams generally will make/not make it but then to back it up with those stats- kudos!

At 2:02 p.m., October 04, 2007, Anonymous Anonymous said...

Good post.

Hope you feel better soon. Just got over the creeping crud myself.

At 2:07 p.m., October 04, 2007, Blogger saskhab said...

I agree. Since the lockout, I've made sure in any predictions to predict the division order first, and work the rest out from there. Even though it did happen last year, I really don't think anyone should ever predict 4 teams from one division all make the playoffs... so many factors have to occur for that to come to fruition. Essentially, the Islanders played division spoiler all year and got into enough 3 point games to sneak out a win. The fact that they got in defied a lot of odds, and not just because they were a questionable team on paper.

At 2:35 p.m., October 04, 2007, Blogger Tom L said...

With waters that muddy James, I think you now know why I hate to make predictions based on anything more than pure 'gut' feeling.

That is an impressive amount of work for results that yield different causes for different situations as to why some teams did or did not make the playoffs.

I think this analysis works really well for the teams at the margin, 8th to 10th place in the conferences, to explain what happened to them, e.g. Toronto vs. the Isles.


At 2:39 p.m., October 04, 2007, Anonymous Anonymous said...

When I made my predictions I also predicted the total points that a division gains.

The last two seasons the central division had 426 and 439 points. This upcoming season I have them at 426 points. The last two seasons the pacific had 479 points and 459 points, this upcoming season I have them at 453 points (I think generally they lost talent). In the northwest I predicted 481 points after seasons of 471 and 469 points. With those 481 points I have 4 teams making the playoffs with essentially Colorado getting in and Nashville dropping out. The question that needs to be asked is can Nashville lose enough within division points to St. Louis, Chicago, Detroit and Columbus to fall out of the playoffs without one of those teams gaining enough to get in. Well, St. Louis missed the playoffs by 15 points and Nashville was in the playoffs by being 15 points ahead of Colorado.

So yes, it is certainly possible, even if the majority of the Predators point loss gets tagged onto the Blues. I don't think it will be easy, but it is certainly possible and I believe has a good chance of happening.

It should be noted that my point total for the northwest division is 481 points which is only 2 off of what the pacific division had in 2005-06 and one off what the northeast division had in 2005-06. It is also only 10 points higher than last year which is 5 more wins that northwest teams have to win over non-divisional opponents. That is 5 wins in a total of 250 non-divisional games. That isn't all that significant when you consider that for the Blues to make the playoffs they have to win on the order of 8 more games over their 82 game schedule.

At 2:43 p.m., October 04, 2007, Anonymous Anonymous said...

It's not just predictions -- analysts often fail to account for the divisional games when citing stats. For example, we keep hearing about how bad the Leafs' defense and goaltending are, but if you take any mediocre team (of which the Leafs are clearly one) and put them in a division with both Ottawa and Buffalo, they were going to get shelled last year (and probably will this year, too).

The Leafs' stats would look a lot better if there was a way to adjust the numbers to account for those 16 games against the most potent offenses in the league.

At 2:50 p.m., October 04, 2007, Blogger Earl Sleek said...

One curious note about your table (though I wouldn't read too much into it):

The two SC Finalists, Anaheim and Ottawa, were the two teams that had nearly exactly the same level of success within their division and outside their division. The Ducks got 1.34 points-per-game and the Sens got 1.28 points-per-game, whether they played in the division or not.

Just a fluke, I think, but still a pretty noteworthy one.

At 1:17 a.m., October 05, 2007, Blogger Kel said...

The whole thing is further complicated by the existence of the extra point (and SO that increases the frequency of such point occuring). Without 3-points games, all intra-division games should produce 160 points. What is a little bit surprising is that the two divisions considered to be toughest last season produced the fewest 3-point games. I am not sure whether there's an explanation but this outcome could be one of the biggest advantages those teams gave up when every point is worth the same in the standings. I don't understand why a fairer ranking system is not invented to go with the unbalanced schedule when comparing teams from different divisions. A simple way to do so is to weight-adjust by dividing divisional game points by half, because each team plays divisional rivals twice often as non-divisonal rivals in the same conference.

At 1:16 p.m., October 05, 2007, Anonymous Anonymous said...

Nice thought provoking material here. It shows well the overall strength of the divisions. But this doesn't really address the question about what allows 4 teams from one division to make the playoffs.

I suppose the conclusion here is you have to win inside your division to have a chance, but I find that kind of lame because arguably the 50 games outside your division are more important because there are more points to be had.

I would think that 4 teams making the playoffs should be a result of a strong division (relative to other divisions) with parity among those teams (splitting intra-division points and especially gaining those tie points). Last year in the atlantic was because of primarily parity and also because of one horribly crappy team - but their overall record outside the division was only average. The horrible team helped pad the other atlantic teams stats giving them enough points to get 4 qualifiers - but they also didn't have any standout teams like Buffalo.

At 1:21 p.m., October 05, 2007, Anonymous Anonymous said...

noah said...
The Leafs' stats would look a lot better if there was a way to adjust the numbers to account for those 16 games against the most potent offenses in the league.

Maybe the Sens' and Sabres' offenses were potent because they got to play the Leafs, Habs, and Boston horrid D eight times each?

At 5:06 p.m., October 05, 2007, Anonymous Anonymous said...

I'll be happy when they scrap divisions, go to a schedule where you play home and away vs. the opposite conference, and 2 home, 2 away vs. the teams in your own conference. Adjust that 86 games down to 82 (if you want to) in some fashion that's random - remove 2 home and 2 away games within the conference.

Everyone's happy, right?

At 5:08 p.m., October 05, 2007, Anonymous Anonymous said...

well, not random, I guess, some sort of pattern where you drop 4 different teams each year to get down to 82 games.


At 11:54 p.m., October 09, 2007, Anonymous Anonymous said...

James you might want to try looking at more than one year. For example in 05-06 Carolina had a better winning percentage against the other divisions than they did against their own. Counting OTLs as losses. The Canes were 18,14 vs the SE with 3 OTLs They were 14,6 vs the NE with 1 otl and 13,7 with 3 otls. I don't know if this holds up if you use more than one season.


Post a Comment

<< Home


Free Page Rank Checker
eXTReMe Tracker