As always, feel free to send questions & comments to me directly, either
to this address ([log in to unmask]) or [log in to unmask]
Keith
========================================================================
These notes deal with an important concept in The College Hockey
Computer Rating: the CONNECTION. Remember, TCHCR uses a connected
schedule graph as the basis for its computations, so understanding
connections is pretty important.
When two teams play against each other for the first time, we connect
the two teams together in our schedule graph. We assign a direction and
value, depending on who won and by how much. If they play a second
time, we do not add another connection, we merely modify the direction,
value and weight of the existing connection.
The more connections there are, the better (more reliable) the rating.
For a team, the number of connections is really the number of different
opponents it has faced. In the following, I will report these
"connection" numbers in terms of number of opponents (in hopes of
reducing the amount of "TCHCR jargon" used).
This table shows the number of games played and the number of opponents
faced as of 1/25/92 for each team. Notice that in general, the eastern
teams have played fewer games but have seen more opponents than their
western counterparts.
Team GP #Opp
Alabama-Huntsville 15 7
Air Force 16 8
Alaska-Anchorage 23 11
Alaska-Fairbanks 21 9
Army 12 9
Boston College 23 15
Bowling Green 22 9
Brown 18 15
Boston University 21 13
Clarkson 18 17
Colgate 18 16
Colorado College 24 11
Cornell 16 14
Dartmouth 16 14
Denver 26 9
Ferris State 23 10
Harvard 14 13
Illinois-Chicago 22 7
Kent 17 9
Lake Superior 22 8
Minnesota-Duluth 24 9
Merrimack 23 16
Miami 24 9
Michigan 25 11
Minnesota 26 12
UMass-Lowell 21 14
Maine 20 13
Michigan State 23 11
Michigan Tech 26 10
North Dakota 26 10
Notre Dame 15 8
Northeastern 21 13
New Hampshire 22 15
Northern Michigan 26 11
Ohio State 24 11
Princeton 15 13
Providence 24 15
RPI 19 16
St Cloud 20 8
St Lawrence 18 15
Union 14 13
Vermont 19 16
Wisconsin 24 10
Western Michigan 22 6
Yale 14 13
Totals 461 261
Average per team 20.5 11.6
The totals of 461 & 261 may need some explaining. So far, 461 games
have been played, and 200 of those have been re-matches between 2 teams
that had already played. Thus, 261 games represent "connections."
The average team has played 20.5 games against 11.6 opponents. That's
about 9 re-matches.
Of course, all this changes as the season progresses. The first part of
the table below will let you compare 1/18, 1/25 and the end of this
season. Notice that the rest of the way, we have mainly re-matches.
Total Per Team Total Per Team
Season Teams Games Max Min Avg Opps Max Min Avg
1/18 45 423 24 12 18.8 254 17 6 11.3
1/25 45 461 26 12 20.5 261 17 6 11.6
91-92@ 45 690 38 18 30.7 279 18 8 12.4
90-91 44 711 38 21 32.3 289 19 9 13.1
89-90 44 691 39 16 31.4 283 19 7 12.9
88-89 43 670 40 12 31.2 306 21 5 14.2
88-89* 42 658 40 12 31.3 301 21 7 14.3
87-88 45 666 38 8 29.6 302 21 3 13.4
87-88* 41 630 38 13 30.7 283 20 6 13.8
86-87 43 646 39 6 30.0 267 17 2 12.4
86-87* 39 621 39 22 31.8 253 17 7 13.0
@ - Does not include Independent Tourney at end of season
* - Adjusted totals after eliminating teams which played
fewer than 14 Division I games that season
But the burning question is: because of NCAA cutbacks from 38 to 34
games, is the 91-92 season significantly different from past years?
Compared with 90-91, there is an average of 1.6 fewer games being
played by each team (from 32.3 to 30.7), but only 0.7 fewer opponents
being played (from 13.1 to 12.4).
This season's numbers are actually pretty close to those of 89-90. We
have added 1 more team, but are only playing 3 more games (the
independent tourney will give us 4 more). Still, the average number of
opponents is only down from 12.9 to 12.4. I suspected a larger drop
(perhaps 2.0) because most western teams went from 6 non-conference
games to only 2. I think we only lost 0.5 opponents because teams
scheduled fewer Division III and Canadian schools, plus more teams went
to Alaska to "steal" extra games. The independents have better Division
I schedules this season, too.
Comparing 91-92 with earlier years is more difficult because schedules
were different back then. Starting in 89-90 (I think), the NCAA said
you had to play 20 Division I games to be eligible for the post-season
tournament. Before then, you had teams like Merrimack playing in a
Division III conference and managing about 10 Division I games a year.
Or even worse, Kent in 86-87, who played 6 games against only 2
opponents.
(The "Per Team" Max and Min columns should give you an idea of the
range of games played and opponents played throughout the years.)
Therefore, I have 2 lines in the table above for each of the
years from 86-87 to 88-89. The first line includes everyone, but
in the second I got rid of the "stragglers." The starred line
should be a better comparison for this season.
Look at the high number of opponents for 88-89. I believe that was
still back when the WCHA and HEA had their interlocking schedule. Boy,
those were the days when college hockey was not regional!
The 12.4 for this year matches the "raw" average opponents for 86-87,
so again, we are not too far off.
So, my conclusions are that although 91-92 is not as well connected as
the past seasons, it is not too bad, either.
=====================END TCHCR NOTES FOR 1/25==========================
|