Northwest vs Southeast Division 2005-10

by Hawerchuk on Jun 9, 2010 9:07 PM EDT

There seems to be some incredulity about my statement that on the basis of their negative shot differential in head-to-head games, the NW division teams have not been better than the SE division teams. How can this be possible if, in 107 head-to-head games, the NW has a .570 winning percentage?

Well, it all comes down to unsustainable and uncharacteristic save and shooting percentages:

	NW vs All	NW vs SE	SE vs All	SE vs NW
S%	6.28	6.23	6.24	4.16
SV%	940	958	937	938
PDO	1002	1021	999	979

Both divisions played 2050 games over five seasons and ended up with roughly the same save and shooting percentages over all of those games. But in 100 head-to-head games in the regular season, the southeast division shot just 4.16% (and the northwest division got a 958 save percentage). Those rates have no relationship with the teams' true talent levels observed over thousands of games, but it is possible to sample 100 games where we see uncharacteristic percentages. In other words, sometimes strange things happen - but it doesn't mean that the NW division is better than the SE division.

Here's the division-vs-division PDO grid with the NW-SE pairs in bold:

	atl	ne	se	cen	nw	pac
atl	1000	1001	999	1015	989	1010
ne	999	1000	1005	1008	1002	989
se	1001	995	1000	1001	979	999
cen	985	992	999	1000	1000	1004
nw	1011	998	1021	1000	1000	998
pac	990	1011	1001	996	1002	1000

And the composite PDO, Sv% and Sh% for each team against other divisions since 2005:

	atl	ne	se	cen	nw	pac
oth div	1001	1002	996	1000	1002	999
oth conf	1005	1000	992	992	1010	1000

sv%	atl	ne	se	cen	nw	pac
oth div	946	945	945	942	943	943
oth conf	946	941	942	941	949	941

sh%	atl	ne	se	cen	nw	pac
oth div	5.57	5.66	5.17	5.73	5.84	5.91
oth conf	5.89	5.97	5.03	5.15	6.08	5.91

There is something real here - SE shooting percentage is 0.4% lower than the league-wide average (goaltending is exactly average.) But there's no reason to expect them to have a 979 PDO against the NW.

Tweet Comment 12 comments | 0 recs |

Do you like this story?

Comments

I haven’t measured if this is statistically significant, but a PDO difference of 3 means a goal differential of +3 every 1000 shots, or roughly every 30 games, which means +0.1 goal differential per game. This would mean that the NW should outscore the SE 2.8 to 2.7 (roughly), which would give them a head-to-head win % of 52%.

by Tom Awad on Jun 9, 2010 9:58 PM EDT reply actions

And the NW had 49% of the shots, so they’re basically 51% vs the SE

by Hawerchuk on Jun 9, 2010 10:03 PM EDT up reply actions

I don’t know Gabe, this suggestion strikes me as being a bit ridiculous. The NW goal differential since the lockout is +69. The SE is -321. That’s a crapload of ground to make up and over a five year sample goal differential should be pretty meaningful. Considering the NW plays the tougher schedule, I just don’t see how the SE could be better. Maybe the advantage isn’t expressing itself in EV shot differential w/ the score tied, but there was something separating those two groups of teams.

by Scott Reynolds on Jun 9, 2010 11:00 PM EDT reply actions

I don’t think anybody’s saying it’s better, but the difference is pretty close. There’s a roughly 400 goal differential, but 5 seasons times 5 tteams = approximately 5000 GF and 5000 GA. Let’s say the NW is 5069 / 5000, and the SE is 5000 / 5321. Then the NW would have 52% of the goals in games played between the two divisions. That’s pretty close to what I calculated.

The bottom line is we are all correct: Gabe is correct in saying that the two divisions are roughly equal in terms of Corsi. HockeyAnalysis is correct in saying the NW had better goaltending, which explains their better record. An extra 0.003 of save percentage on 75000 shots is VERY significant.

by Tom Awad on Jun 10, 2010 1:16 AM EDT up reply actions

I’ll agree that we are all correct in that sense, but I am not sure that Gabe was just trying to show that the divisions have similar Corsi numbers. I have never disagreed with that. Although I haven’t calculated Corsi numbers myself, I have been taking Gabe’s calculations as fact so I would never dispute that. The issue is the conclusions drawn from that observation. That being, that by the NW having a similarly bad Corsi% that makes them an equally bad division as the SE division, which is plainly not true. In fact, the evidence shows that the NW division is quite likely a significantly superior division when it comes to actual results on the ice.

We really shouldn’t be surprised by this Corsi% vs Win% disconnect. Gabe himself said Corsi only explains 40% of winning. If you ignore 60% of the factors that go into winning hockey games, you can expect to find errors in your conclusions.

----------------------------------------------------------------------------------
HockeyAnalysis.com - Taking a Deeper Look at the World of Hockey

by HockeyAnalysis on Jun 10, 2010 5:49 AM EDT up reply actions

Point taken. Luongo, Backstrom and (oddly) Kiprusoff skewed the NW’s goaltending.

But the NW is not .600 vs the SE, nor is it .570. It is, as Tom points out, probably around .520 (when you factor in goaltending) and, in terms of puck possession, not even 50%. There’s a huge amount of luck that made the NW’s record that good. There may even be some small shooting component. But if you had these stats and you were betting on future matchups, it would be a very bad idea to assume a .600 baseline.

I haven’t seen any evidence that it’s a good idea to let yourself get outshot in close games, even if you think your goaltender or your team shooting talent is better than your opponent’s.

by Hawerchuk on Jun 10, 2010 10:39 AM EDT up reply actions

The only real mistake in the article is this one:

“Both divisions played 2050 games over five seasons and ended up with roughly the same save and shooting percentages over all of those games.”

This is actually false. The save and shooting percentages are VASTLY different, at least as much as could be expected of an amalgalm of 5 teams over 5 seasons. 999 vs 1002 of PDO means that for every 20 goals you score, your opponents will score 21, at even shots. You need 51.2% of the Corsi events just to break even.

That being said, it’s still 0.520 head-to-head.

by Tom Awad on Jun 10, 2010 10:59 AM EDT up reply actions

Of course PDO doesn’t take into account all aspects of the game either (special teams, propensity to take/draw penalties, etc.) so it has its flaws as well, even if used in conjunction with Corsi numbers so I still am not certain how confident I should be in that .520 number.

----------------------------------------------------------------------------------
HockeyAnalysis.com - Taking a Deeper Look at the World of Hockey

by HockeyAnalysis on Jun 10, 2010 12:04 PM EDT up reply actions

I agree that if we assume that these numbers came from two imaginary teams over a very long period and that these entities could play future games against one another, expecting a 60/40 advantage from the NW would be a bad idea. Having said that, I think I’d end up coming out ahead in the long run by betting on the NW if you were to give me even odds.

by Scott Reynolds on Jun 10, 2010 11:38 AM EDT up reply actions

I think there is too much unbalance in the schedules between the SE and NW conferences to be able to directly compare save and shooting percentages with and significant degree of confidence.

I pointed out in your other post that the NW systematically has a better record vs SE, ATL, NE and to a lesser degree central division (haven’t checked pacific) than Corsi% would indicate they should. Do you witness the same save% differences with those divisions as you did with the SE. I suspect the NW has just had superior goaltending (and maybe even overall defensive play) that isn’t getting picked up by Corsi%.

----------------------------------------------------------------------------------
HockeyAnalysis.com - Taking a Deeper Look at the World of Hockey

by HockeyAnalysis on Jun 9, 2010 11:20 PM EDT reply actions

You’re on fire lately, Gabe. Terrific stuff.

And though none of the other NW fans are putting up a defense of the division, I’ll make a feable attempt:

1. Most of the head to head games have occured recently, there were none at all in 05/06 iirc. Since the NW has gotten worse, culminating in this year’s shitshow, the head to head numbers are unflattering.

2. As has been mentioned by others, there are reasons to believe that the goaltending is better in the NW. If you looked at the variance in 150 EV road save% team-seasons for 05/06 thru 09/10 … and compared that to what we’d expect from random chance. The difference is very small indeed, and most of it is consumed by the NW and SE divisions. You’re using Fenwick shootting% and svae% here, no, Gabe?

3. The NW has better special teams play, esp PK (not including goaltending here).

Any road, next year the SE will almost certainly outchance the NW at even strength in head to head games. We’ll probably have about 30 games to count if your scoring chance project gets off the ground. If you can find someone foolish enough to take that wager at even odds, grab it Gabe, and goad them into as large of a wager as possible.

The opposite applies to special teams. The NW will almost certainly outchance the SE with the man advantage next year.

by Vic Ferrari on Jun 10, 2010 1:55 PM EDT reply actions