Greg writes:
>One possiblity (let's just give a little more ammunition to the ECAC
>detractors among us) is that their are a disproportionate number of
>weak teams, both offensively and defensively, in the conference. The
>goaltenders of the best teams would then tend to have artificially
>depressed GAAs.
>
>So, would somebody with a decent stats package give us the standard
>deviation in team scoring for each of the four Div. I conferences? If
>the above hypothesis is correct, we'd expect a significantly higher
>variance for the ECAC.
OK, I'll bite... I put together the goals-for and goals-against for the
teams in all four conferences, and here are the numbers I came up with
(which I'm pretty sure are accurate :-) The variances and standard devi-
ations are listed below the overall conference numbers.
(I've Emailed John Haeussler about a few discrepancies between the numbers I
got and the UPSIDA stats)
Conference standings: (GFA = Goals-For average, or the number of goals
scored per game. GAA = Goals-Against average)
ECAC
Team Games GF GA GFA GAA
-----------------------------------------------
Harvard 27 103 78 3.81 2.89
St. Lawrence 34 160 111 4.71 3.26
Clarkson 33 165 100 5.00 3.03
Yale 27 121 110 4.48 4.07
Cornell 29 92 78 3.17 2.69
Brown 30 120 143 4.00 4.77
Vermont 31 116 95 3.74 3.06
Colgate 31 152 151 4.90 4.87
Princeton 27 107 115 3.96 4.26
RPI 33 124 128 3.76 3.88
Dartmouth 26 70 142 2.69 5.46
Union 25 82 138 3.28 5.52
Hockey East
Team Games GF GA GFA GAA
-----------------------------------------------
Maine 37 203 90 5.49 2.43
New Hampshire 37 168 127 4.54 3.43
Providence 36 175 134 4.86 3.72
Boston University 35 161 124 4.60 3.54
Boston College 35 116 140 3.31 4.00
UMass-Lowell 34 127 155 3.74 4.56
Northeastern 35 142 167 4.06 4.77
Merrimack 34 128 149 3.76 4.38
CCHA
Team Games GF GA GFA GAA
-----------------------------------------------
Michigan 44 215 144 4.89 3.27
Lake Superior 43 203 103 4.72 2.40
Michigan State 44 199 143 4.52 3.25
Western Michigan 36 140 130 3.89 3.61
Miami 40 165 183 4.13 4.58
Ferris State 38 119 148 3.13 3.89
Illinois-Chicago 36 123 152 3.42 4.22
Ohio State 38 162 211 4.26 5.55
Bowling Green 34 133 165 3.91 4.85
WCHA
Team Games GF GA GFA GAA
-----------------------------------------------
Minnesota 44 215 133 4.89 3.02
Wisconsin 43 175 142 4.07 3.30
Northern Michigan 42 238 165 5.67 3.93
Colorado College 41 170 175 4.15 4.27
Minnesota-Duluth 37 144 162 3.89 4.38
Michigan Tech. 38 131 175 3.45 4.61
St. Cloud 39 175 197 4.49 5.05
North Dakota 37 145 161 3.92 4.35
Denver 36 124 187 3.44 5.19
Means, variances, standard deviations:
ECAC
Mean Variance Std. Dev.
GF GA GF GA GF GA
------------------------------------
3.96 3.98 0.47 0.93 0.68 0.97
Hockey East
Mean Variance Std. Dev.
GF GA GF GA GF GA
------------------------------------
4.29 3.86 0.44 0.49 0.66 0.70
CCHA
Mean Variance Std. Dev.
GF GA GF GA GF GA
------------------------------------
4.10 3.96 0.30 0.81 0.55 0.90
WCHA
Mean Variance Std. Dev.
GF GA GF GA GF GA
------------------------------------
4.22 4.23 0.45 0.47 0.67 0.68
Some observations on all of this number-crunching:
The ECAC's 3.96 average goals scored per game is the lowest of the four
conferences. Since the variances and standard deviations for goals scored
are quite consistent across all four conferences, this pretty reliably
indicates that the ECAC was the weakest as far as offense was concerned.
Indeed, when we look at individual teams, we find that of the four worst
goal-scoring squads listed above, three reside in the ECAC: Dartmouth
(2.69 goals/game), Cornell (3.17, which makes me REALLY appreciate Parris
Duffus!), and Union (3.28), The fourth team is the CCHA's Ferris State
(3.13).
So now we come to the UPSIDA stats, which show that seven of the top 15
goaltenders in terms of GAA played for ECAC teams. Do these seven have
artificially lower GAAs? Based on the above, I'd say yes -- not neces-
sarily because the goaltenders for the good ECAC teams get to beat up on a
bunch of weaker squads, but because the league as a whole is relatively
weaker on offense.
This is one of a number of reasons why I don't consider GAA a particularly
good indicator of how well a goalie is playing. GAA can be artificially
depressed by playing against a lot of low-scoring teams or by playing behind
an outstanding defense. My own opinion is that some combination of save
percentage and number of shots faced (or saves made) would be the best way
to statistically compare the performances of different goalies.
Consider, for example, the cases of Maine's Garth Snow and Clarkson's Chris
Rogles. Snow's 2.44 GAA ranked him second in the NCAA, while Rogles, at
3.07, was a bit farther down the list (tied for tenth). However, Rogles'
save percentage was .904 (sixth), while Snow's name does not appear among
the save percentage leaders. If we look at the number of shots faced by
each goaltender, we find that Rogles faced 31.98 shots per game (this comes
from the fact that the .096 percentage of shots he missed per game resulted
in 3.07 goals), while the Maine defense allowed somewhere around 21 shots
per game on Snow. In other words, all other things being equal (same save
percentage, etc.), if the Maine defense had let Snow face the same number of
shots per game that Rogles did, Snow's GAA would have been about 3.70. In a
comparison based on these stats, [asbestos suit on] Rogles would emerge as
the better goalie. Of course, this comparison doesn't take into account the
fact that Maine had a tougher schedule, or that Snow appeared in about twice
as many games as Rogles did. Still, if comparisons need to be made between
goalies, the GAA alone is not the best way to do it.
Another note from the above stats concerns team defense. For Hockey East
and the WCHA, the variances and standard deviations for average goals
allowed per game are pretty much in line with those for average goals scored
per game, indicating that in these two leagues, there is not a huge gap
between the top defense and the worst one. This is not the case in the
ECAC, which has a relatively large variance of 0.93 and a standard devia-
tion of 0.97. It has been mentioned on occasion that the ECAC is not a
well-balanced league -- the top teams are usually quite good and can compete
with anybody, while the bottom teams often perform like sacrificial lambs --
and the defensive numbers tend to support that point.
But surprisingly enough, the CCHA appears to have been almost as unbalanced
as the ECAC this season. Although the average goals allowed per game for
both leagues are almost identical (3.98 for the ECAC, 3.96 for the CCHA),
the variance and standard deviation for the CCHA (0.81 and 0.90, respect-
ively) are almost as large as those for the ECAC. Thus, the gap between the
CCHA's top defense and its worst is quite a significant one. Now, the CCHA
overall was somewhat stronger than the ECAC (witness the two leagues' per-
formances in the NCAAs), but the stats seem to indicate that there were some
relatively weaker teams that the league powerhouses could beat up on (al-
though not always -- Ohio State did sweep Lake Superior). Before anyone
asks, no, I'm not saying that the weak teams in the CCHA are as bad as the
weak teams in the ECAC... but it's interesting to note that the ECAC was not
the only unbalanced conference last season.
Well, I suspect a fair bit of discussion may come out of all this, so I'll
stop blabbing for now...
--
Bill Fenwick
Cornell '86 and probably '94
LET'S GO RED!!
I know some people believe in astrology -- but do they follow it this closely?
"Balance that checkbook; it's easy, now that Mercury's touring Taurus..."
-- Joyce Jillson, horoscope for May 11, 1992
|