Since some people continue to insist that Quinnipiac will (or should)
get an at-large bid based on their RPI and PWR, I decided to modify my
pairwise comparison script to calculate the RPI, record vs teams under
consideration and record in the last 16 games with games between MAAC
teams omitted.
Here are the the results for the six MAAC teams, with each team's
rank and ratings by the normal procedure in the first row and the
modified one in the second. (RPI already leaves out games involving a
team from its opponents' opponents' winning percentage; I just
modified this to also leave out games vs other MAAC teams when both
teams in question were from the MAAC. Record in the last 16 games I
just replaced with total non-conference record, which is less than 16
games in each case.)
Team RPI PF-PA PCT OPPCT OPOPP vsTUC Last16
9 Quinnipiac 0.554 36-10 0.783 0.427 0.446 0.625 0.750
41 Quinnipiac 0.422 36-10 0.783 0.203 0.315 N/A 0.667
18 Connecticut 0.521 30-16 0.652 0.451 0.447 0.300 0.625
44 Connecticut 0.413 30-16 0.652 0.276 0.307 0.000 0.333
27 Holy Cross 0.493 28-20 0.583 0.444 0.443 0.444 0.562
46 Holy Cross 0.403 28-20 0.583 0.306 0.303 0.000 0.125
42 Canisius 0.440 18-28 0.391 0.474 0.443 0.200 0.531
48 Canisius 0.356 18-28 0.391 0.350 0.295 0.000 0.000
47 Iona 0.419 18-26 0.409 0.417 0.445 0.125 0.375
50 Iona 0.294 18-26 0.409 0.212 0.300 N/A 0.500
51 Fairfield 0.284 0-44 0.000 0.437 0.441 0.000 0.000
51 Fairfield 0.204 0-44 0.000 0.325 0.274 N/A 0.000
All six D1 MAAC teams take a big hit in opponents' opponents' winning
percentage, and a corresponding drop in RPI. Note that even with this
modified RPI, Quinnipiac only drops one place in the pairwise
rankings, and that is due to the loss of pairwise comparisons with
Princeton and Mankato, who are much lower ranked. However, if we
exclude MAAC teams from Q's record vs TUC and in the last 16, their
pairwise comparisons go in the toilet, and the only non-MAAC team with
which they win a comparison is Niagara. (UConn and Holy Cross win
*no* comparisons out of conference in that case.)
Just to make sure I wasn't tapping into some other effect which had
nothing to do with the MAAC's closed schedule, I tried the same thing
using the ECAC instead of the MAAC, and found the following:
Team RPI PF-PA PCT OPPCT OPOPP vsTUC Last16
6 Clarkson 0.583 43-21 0.672 0.545 0.498 0.447 0.875
5 Clarkson 0.601 43-21 0.672 0.577 0.516 0.278 0.350
10 St Lawrence 0.550 43-25 0.632 0.507 0.502 0.500 0.688
8 St Lawrence 0.567 43-25 0.632 0.536 0.519 0.167 0.364
14 RPI 0.533 42-22 0.656 0.458 0.498 0.346 0.656
12 RPI 0.547 42-22 0.656 0.479 0.517 0.000 0.667
15 Princeton 0.533 37-21 0.638 0.468 0.501 0.636 0.500
13 Princeton 0.544 37-21 0.638 0.485 0.521 1.000 0.714
17 Colgate 0.521 38-26 0.594 0.479 0.495 0.393 0.531
16 Colgate 0.530 38-26 0.594 0.489 0.515 0.333 0.625
24 Yale 0.503 30-28 0.517 0.495 0.495 0.179 0.625
18 Yale 0.526 30-28 0.517 0.536 0.513 0.000 0.286
29 Cornell 0.488 27-31 0.466 0.503 0.493 0.265 0.344
24 Cornell 0.505 27-31 0.466 0.531 0.513 0.250 0.500
30 Vermont 0.485 28-34 0.452 0.503 0.501 0.344 0.281
25 Vermont 0.503 28-34 0.452 0.535 0.520 0.500 0.625
31 Harvard 0.480 24-30 0.444 0.499 0.501 0.214 0.594
29 Harvard 0.494 24-30 0.444 0.520 0.523 0.500 0.600
36 Brown 0.466 24-34 0.414 0.494 0.495 0.265 0.500
33 Brown 0.478 24-34 0.414 0.512 0.515 0.600 0.643
40 Dartmouth 0.455 20-34 0.370 0.504 0.493 0.300 0.375
35 Dartmouth 0.469 20-34 0.370 0.524 0.516 0.667 0.750
49 Union 0.364 9-55 0.141 0.483 0.492 0.147 0.094
49 Union 0.387 9-55 0.141 0.522 0.509 0.375 0.278
All of the ECAC teams actually *improve* their RPI if ECAC vs ECAC
games are left out of opponents' opponents' winning percentage. (Not
sure why this is.) Their records vs TUC and in the last 16 vary a lot
in both directions, as one would expect from a smaller sample size.
The effect on pairwise comparisons is to drop Clarkson and SLU a lot
and boost Princeton a lot and RPI and Colgate a little, with the net
result that three ECAC teams instead of two end up in the top twelve:
Team PWR RPI Comparisons Won
5 Clarkson 21 .583 ________ CCDUQnBCSLMiOSNtNMRPPnCgCtMkPvNiMLFSYaBGHC
11 Clarkson 13 .601 __________CC____Pn__ RP__OSNMCgPv__SLMLNi__Mk__YaHC
10 St Lawrence 16 .550 __________________ MiOSNtNMRPPnCgCtMkPvNiMLFSYaBGHC
19 St Lawrence 7 .567 ________________Pn____RP__OS__Cg____ ______MkFSYa__
15 RPI 12 .533 ______________________OS____ PnCgCtMkPvNiMLFSYaBGHC
12 RPI 13 .547 ________________Pn____ NtOSNMCgPvCt____NiBGMkFSYaHC
16 Princeton 11 .533 ______________________OS______ CgCtMkPvNiMLFSYaBGHC
9 Princeton 15 .544 __________CC____ Mi____NtOSNMCgPvCt__MLNiBGMkFSYaHC
17 Colgate 8 .521 ________________________________ CtMkPv__MLFSYaBGHC
16 Colgate 10 .530 __________________________OS__ PvCt__MLNiBGMkFSYaHC
24 Yale 3 .503 ________________________________________Ni____ BGHC
25 Yale 3 .526 ________________________________________Ni__Mk__ HC
(the comparisons flipped are a little harder to make out because the
script orders the opponents based on their PWR, which is different in
the two cases)
This is not to say that games among ECAC teams should be excluded, of
course. The ECAC's non-conference schedule provides enough of a
baseline for the additional information about games between ECAC teams
to be meaningful. The MAAC, on the other hand, provides the committee
with very little independent information about how to judge MAAC
*intra*conference games, and removing the effects of those games
within the conference makes the high overall ratings of Quinnipiac and
the other MAAC leaders crumble into dust. The committee noted in the
NCAA News last summer "that it reserves the right to evaluate each
team based on the relative strength of their respective conference"
for just this reason, and I see no sensible reason why they will not
use this language to exclude Quinnipiac on this basis. (For a
discussion of how the criteria might be modified to make this sort of
ad hoc exclusion unneccessary, see the my next post.)
John Whelan, Cornell '91
[log in to unmask]
http://www.amurgsval.org/joe/
HOCKEY-L is for discussion of college ice hockey; send information to
[log in to unmask], The College Hockey Information List.
|