The College Hockey Computer Rating
compiled by Keith Instone
NOTE: TCHCR is *not* used in conjunction with the NCAA selection process.
Includes games on 1/4/92. Last week's ranking includes games on 12/25/91.
Last Division I Schedule Schedule
Rank Week Team Record Rating Strength Rank
1 2 Michigan 13 3 3 87.21 66.38 4
2 3 Lake Superior 13 4 1 84.63 63.90 12
3 6 Minnesota 14 6 0 82.51 66.05 5
4 5 Northern Michigan 13 5 2 81.01 63.76 13
5 8 Michigan State 9 4 4 73.99 63.91 11
6 1 Maine 14 2 0 73.24 44.12 34
7 14 Wisconsin 12 7 1 71.89 64.17 9
8 13 Western Michigan 7 6 2 68.11 65.48 7
9 9 Alaska-Anchorage 12 4 0 67.30 48.68 22
10 20 North Dakota 12 8 0 64.85 57.85 18
11 10 St Lawrence 11 2 1 64.81 39.37 42
12 18 Miami 9 7 2 63.57 59.09 17
13 7 Clarkson 12 3 0 62.99 39.19 43
14 15 Boston University 11 4 0 62.24 46.61 27
15 16 Providence 12 5 0 61.96 46.05 31
16 19 St Cloud 7 8 1 61.96 65.49 6
17 17 Minnesota-Duluth 8 9 1 61.84 65.14 8
18 4 Yale 6 2 2 61.39 47.21 26
19 22 Michigan Tech 8 11 1 60.37 67.80 2
20 12 New Hampshire 11 5 0 57.37 42.94 36
21 11 Harvard 6 3 2 56.64 43.53 35
22 27 Illinois-Chicago 5 8 3 54.18 63.09 14
23 25 Colorado College 5 10 4 53.23 62.76 15
24 28 Ferris State 2 9 5 52.05 68.58 1
25 33 Ohio State 6 10 2 49.84 60.98 16
26 30 Bowling Green 4 10 2 49.64 64.04 10
27 24 Cornell 4 4 2 48.51 46.59 28
28 21 UMass-Lowell 6 8 0 48.27 54.39 19
29 36 Denver 5 14 1 48.22 66.45 3
30 26 Boston College 7 9 1 45.06 50.71 20
31 23 Vermont 5 7 1 43.05 46.21 29
32 34 Alaska-Fairbanks 6 8 0 42.01 46.19 30
33 31 Northeastern 7 8 0 41.88 45.98 32
34 32 RPI 7 8 0 38.90 42.50 38
35 29 Brown 4 6 2 36.46 40.85 39
36 39 Alabama-Huntsville 4 9 0 35.12 49.74 21
37 37 Merrimack 8 9 0 32.44 33.48 45
38 35 Princeton 4 9 0 30.11 42.92 37
39 38 Kent 2 10 1 27.59 48.61 23
40 40 Colgate 3 8 1 27.50 40.27 40
41 42 Notre Dame 3 8 0 19.07 39.17 44
42 41 Dartmouth 1 9 0 15.19 48.50 24
43 43 Air Force 1 10 0 10.89 39.89 41
44 45 Army 0 10 0 8.32 48.19 25
45 44 Union 0 10 0 7.59 45.32 33
The College Hockey Computer Rating uses a connected schedule graph
and a least squares optimization to rate the 45 Division I college
hockey teams. This method considers a team's opponents, and their
opponents, etc., until all 343 Division I games are included.
In addition to rating the teams, a strength of schedule is calculated
for each team. The schedule strength can be computed by averaging
the ratings of a team's opponents for each game.
For every game, the teams are evaluated based on: the outcome (win,
lose, tie); the margin of victory; percentage of goals allowed;
the site of the game (home, away, neutral ice); and overtime or
regulation time.
Only games against other Division I teams are considered.
-----
Here is a table of how the leagues have done against each other so
far this season. Although TCHCR does not directly use statistics
like these (it only considers individual game outcomes), I have
found that I can explain some trends in TCHCR by looking at summaries
such as this.
CCHA WCHA ECAC HEA Indep Totals
CCHA --- 4-4-1 4-0 1-1-1 4-1 13-6-2
(.500) (1.000) (.500) (.800) (.667)
WCHA 4-4-1 --- 1-0 4-4 6-1 15-9-1
(.500) (1.000) (.500) (.857) (.620)
ECAC 0-4 0-1 --- 10-16 6-3-1 16-24-1
(.000) (.000) (.385) (.650) (.402)
HEA 1-1-1 4-4 16-10 --- 25-5 46-20-1
(.500) (.500) (.615) (.833) (.694)
Indep 1-4 1-6 3-6-1 5-25 --- 10-41-1
(.200) (.143) (.350) (.167) (.202)
THE THREE AMIGOS: Notice the balance between the CCHA, WCHA and
HEA: they are all .500 against each other. The ECAC has not fared
as well, going winless against the West and 6 games under .500
against HEA. Also, the ECAC is only managing a 65% winning percentage
against the Independents while the other 3 leagues are at 80% or
higher.
HOLIDAY SWEEP: The CCHA's mark of 4-0 versus the ECAC is significant,
especially when you consider that tied-for-last Ohio State beat
first-place Clarkson in Syracuse. Third-place Harvard's pair of
3-1 losses at the GLI don't help the ECAC cause, either.
A WEAK POINT: 84% of HEA's out-of-conference games have come against
the ECAC and Independents. Since both groups of teams, in general,
are rated low this week by TCHCR, HEA schedules are coming out
"weak." UMass-Lowell's is the best at 19th.
WE WANT MORE: One number that jumps out at me is the *SINGLE GAME*
that has been played between the WCHA and ECAC. There will not be
any more games between those two leagues this season!! More games
between conferences means more data for comparing teams, which
means more reliable ratings from TCHCR. (TCHCR compares every team
with every other team to come up with its ratings.)
CHAIN GANG: Even though the WCHA and ECAC have only played one game
against each other, TCHCR uses the connected schedule graph to find
other ways of comparing teams from the two leagues. For example,
besides directly using the Yale-Wisconsin result, TCHCR simultaneously
looks at ALL other paths of games connecting the two schools. Yale
has played Harvard, who has played Michigan, who has played Minnesota,
who has played Wisconsin. Another chain is
Yale-->Colgate-->Kent-->Maine-->Wisconsin. Hopefully, you can see
that your opponent's performance affects your rating when TCHCR
traces a chain through that opponent.
And now, to answer some viewer mail:
>From the Hockey East office:
>>Do you factor for the number of games played?
Yes. When comparing game outcome measures (GOM) to calculate ratings,
TCHCR uses the AVERAGE GOM between the two teams. For example,
let's say you play a team twice at home and beat them 5-3 one night
while tieing them (score is irrelevant) the second. You GOMs would
be +10+2-1=+11 for the win and -1 for the tie. Your average GOM
would then be (+11-1)/2 = +5. This +5 is what is used to calculate
the ratings.
A side note about schedule strength and games played: There is
nothing inherent in TCHCR that rewards teams who play more games.
However, a team that schedules more games can benefit IF THEY
SCHEDULE GOOD OPPONENTS. Likewise, a team can hurt itself if they
use these extra games to play "cream puffs." For example, Maine
moves up to #5 this week if I don't count its 15-5 thrashing of
Army in the Dexter. The Black Bears' schedule jumps up 2.5 points
(to 27th) and their rating increases by 1.5 points. This is a
simplification, but the win over Army is worth 1.0 points of
"performance" while it costs Maine 2.5 points in schedule, leading
to a 1.5 total rating point loss. The moral: it is not a good idea
for the #1 team to play the #45 team.
>Also from the Hockey East office:
>>I really wonder about...the GOM (Game Outcome Measure)...You
refer to it as "how much one team beat up on another."
The GOM is my measure for how well a team performed in a game. A
team gets 10 points for a win, plus UP TO 5 additional points for
margin of victory. If you outscore your opponent by 6 or more goals,
you get the full +5 margin of victory bonus. If you score 4 or 5
more goals, you get +4. Fewer than 4 goal margins earn a bonus
equal to that margin. Thus, if you "run up the score" and win by
8 goals instead of only 4, you will only increase your GOM for that
game by 1 (ignoring the defensive bonus for now). That 1 GOM point
represents only 6% of the maximum GOM you can get, and is merely
10% as important as the fact that you won. I can't completely ignore
the 4 extra goals for several reasons, one being that perhaps the
winning team was not even trying to run up the score, but the losing
team is that much more incompetent.
And as the Maine example above indicates, how much you win by is
not as important as who you play.
===END of TCHCR notes and mailbag for 1/4/92.
|
