Announcement

Collapse
No announcement yet.

How to improve the Pairwise

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • Shirtless Guy
    replied
    Re: How to improve the Pairwise

    Originally posted by LordofBrewtown View Post
    My suggestions: Scrap QWB completely (it's already part of RPI). Count nuetral games half home, and half road (or just count full for both if that's easier).
    QWB really isn't in the RPI calc with it being 25/21/51. Here are the most valuable teams to beat for this season with adjusted RPI rank in paranthesis (before adding QWB):
    (1) Quinnipiac
    (8) Harvard
    (10) Boston University
    (4) Providence
    (6) Denver
    (14) Minnesota-Duluth
    (7) Michigan
    (2) St. Cloud State
    (5) Boston College
    (12) Notre Dame
    (13) Northeastern
    (17) Minnesota
    (3) North Dakota
    (16) Cornell
    (18) Robert Morris
    (23) Minnesota State
    (11) Yale
    (9) Mass.-Lowell
    (15) Michigan Tech
    (22) Dartmouth

    Missing 19, 20, 21

    I really do like the idea of giving QWB to road teams getting a positive results against the top 20 home RPI teams...

    Leave a comment:


  • LordofBrewtown
    replied
    Re: How to improve the Pairwise

    My suggestions: Scrap QWB completely (it's already part of RPI). Count nuetral games half home, and half road (or just count full for both if that's easier).

    Leave a comment:


  • Shirtless Guy
    replied
    Re: How to improve the Pairwise

    Originally posted by LordofBrewtown View Post
    I'd love to see the results if you split the RPI into the 2 componenets as discussed below...eagerly awaiting as I'm far too lazy to do this myself
    For all who like splitting RPI into home/road do we keep the weighted game values? How do we deal with neutral sites? I would assume we keep the game weights and count neutral site games as half home half away games?

    EDIT: How to deal with QWB? Are we removing that completely? Giving QWB in Road RPI to the victories vs the top 20 teams in Home RPI and vice versa?
    Last edited by Shirtless Guy; 03-24-2016, 10:09 AM.

    Leave a comment:


  • LordofBrewtown
    replied
    Re: How to improve the Pairwise

    Originally posted by Shirtless Guy View Post

    Other thoughts? I'd like to gather a few ideas and incorporate them into the last 3 years of the current system and see how it would have changed the field.
    I'd love to see the results if you split the RPI into the 2 componenets as discussed below...eagerly awaiting as I'm far too lazy to do this myself

    Leave a comment:


  • Khryx
    replied
    Re: How to improve the Pairwise

    First, I would like to see LordofBrewtown's proposal (home and road rpi calculations) or at least the rpi portion. I think that would be interesting. Because I don't know, does the current RPI calculation exclude the team in question in both the Oper and OOper?

    Second, I agree with mookie and dxmnkd316. The need to encourage scheduling tougher OOC games and especially road games is important. Like dxmnkd31 said, I'm not sure that the QWB is causing it but I think the system rewarding it is important.

    Leave a comment:


  • dxmnkd316
    replied
    Re: How to improve the Pairwise

    I tend to think mookie is right. Though I don't think it's the QWB that's doing it. They are starting to get harder negotiations from teams. They're demanding return trips because they can now. Minnesota isn't in the same bargaining position they used to be in.

    Leave a comment:


  • mookie1995
    replied
    Re: How to improve the Pairwise

    Originally posted by LordofBrewtown View Post
    Since that is problematic/probably won't fly, I propose separating the RPI into a separate Home RPI and Road RPI (essentially you have 120 teams instead of 60). That should help account for scheduling quality road games, and help account for various levels of home stadium advantage..
    mookie would be on board for a split like this. much like democrats and tax incentives () i think the ncaa has to wield its stick to push inclusive scheduling.

    mookie is quite satisfied with bu for in our boys ooc games we visiting union, ylae, quinni, and harvard. for neutral games met nu, bc, & then cornell at msg. and hosted michigan 2x, denver, bucky, and bentley [ok, last 2 teams blew, but...].

    other top programs are finally getting out of their home rink as well. even the goofs visited notre dame this year so the days of them playing all their ooc games at home appears to be waning. but it is because of this bonus. if this didn't exist the goofs would continue to invite uconn, merrimack and canisius to the mariucci for the holiday dodge tourney year after year.

    Leave a comment:


  • LordofBrewtown
    replied
    Re: How to improve the Pairwise

    I am all for adding a 4th, or 5th criteria as long as they are the right ones.
    I am against using a team's record in the last X number of games - games in the first part of the season should count as much as the end.
    I am against any consideration for missing players/injuries. I know, it sucks…but it's a team sport and every game played adds to your resume.
    I am also against anything that introduces an arbitrary cutoff or cliff - so no record against Teams under consideration. That introduces double jeopardy anyway (since it's already in the RPI component).
    For the same reason - get rid of the QWB - it is already in the RPI (double jeopardy), and maybe we can replace in a new 4th or 5th comparison.

    I think the best criteria to add is something that accounts for Home/Road games. Since conference games are evenly home/road, my first inclination is to just use non-conference road wins as the criteria. The problem is the discrepancy in number of road games by conference/unequal opportunity. I am against using the win pct for those games, as I think its more impressive for someone to go 3-4-1 non-conf on the road than 1-1.

    Since that is problematic/probably won't fly, I propose separating the RPI into a separate Home RPI and Road RPI (essentially you have 120 teams instead of 60). That should help account for scheduling quality road games, and help account for various levels of home stadium advantage. The only dilemma with this is how to account for neutral site games: not at all, or count in both the home and road RPI (but for half?). I would count in both the home and road RPI. I'd let the NCAA assign any questionable games pre-season/tournament (e.g if The Big10 tourny happened to be in Milwaukee, it should be designated a home game for Wisconsin and away game for Michigan if playing each other).

    Might be a little tougher to go back and look at prior years, as you'd have to track/assign home and away teams; but, I'd love to see this comparison.

    I would also alter the RPI formula to put more emphasis on winning (50-20-30), and am for changing the tie-breaker to H2H, followed by Common Opponents, then overall RPI.

    So the 4 comparisons would be: Home RPI, Away RPI, H2H, Common Opponents, (Overall RPI only used as 3rd tie breaker if needed, or potentially added as a 5th comparison)

    So, there's 1 option if you have time on your hands.

    Leave a comment:


  • mookie1995
    replied
    Re: How to improve the Pairwise

    Rank-rpi unadjusted//Team//Rank*krach
    1 Quinnipiac 1
    2 St. Cloud State 2
    3 North Dakota 3
    4 Providence 4
    5 Boston College 5
    6 Denver 6
    7 Michigan 7
    9 Massachusetts-Lowell 8
    8 Harvard 9
    11 Yale 10
    10 Boston University 11
    12 Notre Dame 12
    14 Minnesota-Duluth 13
    13 Northeastern 14
    16 Cornell 15
    19 Nebraska-Omaha 16
    15 Michigan Tech 17
    21 St. Lawrence 18
    17 Minnesota 19
    22 Dartmouth 20
    27 Miami 21
    20 Penn State 22
    23 Minnesota State 23
    25 Rensselaer 24
    24 Clarkson 25
    26 Bowling Green 26
    30 Union 27
    18 Robert Morris 28
    29 Ferris State 29
    35 Vermont 30
    31 Ohio State 31
    32 Bemidji State 32
    38 Merrimack 33
    28 Air Force 34
    34 Northern Michigan 35
    43 Western Michigan 36
    33 Holy Cross 37
    39 New Hampshire 38
    42 Wisconsin 39
    47 Michigan State 40
    36 RIT 41
    41 Colgate 42
    37 Mercyhurst 43
    44 Lake Superior 44
    45 Connecticut 45
    50 Brown 46
    49 Massachusetts 47
    53 Maine 48
    40 Army 49
    51 Alaska-Anchorage 50
    52 Alaska 51
    55 Colorado College 52
    46 Bentley 53
    56 Alabama-Huntsville 54
    48 Sacred Heart 55
    54 Canisius 56
    57 Princeton 57
    58 Niagara 58
    59 Arizona State 59
    60 American International 60

    Leave a comment:


  • mookie1995
    replied
    Re: How to improve the Pairwise

    this looks to have that quality bonus in, so bu is higher in rpi due to their wins @ qu and hu, plus home wins vs mich & du at home ooc

    Leave a comment:


  • dxmnkd316
    replied
    Re: How to improve the Pairwise

    No, KRACH-PWR. Where you use KRACH instead of RPI in the PWR calculation.

    Leave a comment:


  • mookie1995
    replied
    Re: How to improve the Pairwise

    KRACH
    Rank Team Rating RRWP Win % Rk W-L-T Win % Win Ratio SOS Rk SOS
    1 Quinnipiac 668.6 .8564 1 29-3-7 0.8333 5.000 13 141.8
    2 St. Cloud State 572.9 .8372 3 31-8-1 0.7875 3.706 8 161.1
    3 North Dakota 537.9 .8287 2 30-6-4 0.8000 4.000 15 140.7
    4 Providence 470.4 .8098 4 27-6-4 0.7838 3.625 18 135.5
    5 Boston College 398.7 .7844 5 26-7-5 0.7500 3.000 17 137.5
    6 Denver 389.3 .7806 9 23-9-6 0.6842 2.167 3 183.6
    7 Michigan 296.2 .7334 6 24-7-5 0.7361 2.789 32 109.7
    8 Massachusetts-Lowell 291.8 .7306 7 24-9-5 0.6974 2.304 22 129.7
    9 Harvard 275.4 .7198 12 19-10-4 0.6364 1.750 9 160.1
    10 Yale 270.2 .7162 10 19-8-4 0.6774 2.100 20 131.9
    11 Boston University 259.3 .7083 14 21-12-5 0.6184 1.621 7 162.1
    12 Notre Dame 255.6 .7055 13 19-10-7 0.6250 1.667 10 155.6
    13 Minnesota-Duluth 239.2 .6924 27 18-15-5 0.5395 1.171 2 205.0
    14 Northeastern 222.1 .6775 15 22-13-5 0.6125 1.581 12 142.2
    15 Cornell 194.0 .6493 20 16-11-7 0.5735 1.345 11 145.5
    16 Nebraska-Omaha 181.4 .6350 31 18-17-1 0.5139 1.057 6 171.9
    17 Michigan Tech 171.3 .6225 8 23-9-5 0.6892 2.217 48 79.07

    RPI
    Rank Team RPI Adj RPI QWB QWB Adj RPI W-L-T Win % Win % Rank SOS SOS Rank
    1 Quinnipiac .5904 .5938* .0070 .6008 29-3-7 .8229 1 .5129 17
    2 St. Cloud State .5902 .5929* .0063 .5991 31-8-1 .7822 4 .5262 6
    3 North Dakota .5876 .5883* .0057 .5939 30-6-4 .8010 2 .5165 15
    4 Providence .5762 .5762 .0057 .5819 27-6-4 .7877 3 .5057 30
    5 Boston College .5712 .5712 .0073 .5785 26-7-5 .7473 5 .5125 18
    6 Denver .5706 .5706 .0055 .5761 23-9-6 .6896 8 .5310 3
    7 Michigan .5669 .5672* .0027 .5699 24-7-5 .7314 6 .5121 20
    8 Massachusetts-Lowell .5509 .5518* .0070 .5587 24-9-5 .6917 7 .5040 32
    9 Boston University .5489 .5489 .0095 .5584 21-12-5 .6257 13 .5233 11
    10 Harvard .5519 .5519 .0021 .5540 19-10-4 .6364 12 .5238 10
    11 Yale .5482 .5485* .0030 .5515 19-8-4 .6741 10 .5063 27
    12 Notre Dame .5462 .5462 .0051 .5513 19-10-7 .6124 16 .5242 9
    13 Northeastern .5428 .5428 .0059 .5487 22-13-5 .6193 14 .5173 13
    14 Minnesota-Duluth .5368 .5368 .0071 .5440 18-15-5 .5405 26 .5356 1
    15 Cornell .5312 .5312 .0054 .5366 16-11-7 .5719 20 .5177 12
    16 Michigan Tech .5354 .5354 .0003 .5358 23-9-5 .6892 9 .4842 45

    Leave a comment:


  • LtPowers
    replied
    Re: How to improve the Pairwise

    Originally posted by mookie1995 View Post
    what does krach look like this year?
    http://www.uscho.com/rankings/krach/d-i-men/


    Powers &8^]

    Leave a comment:


  • mookie1995
    replied
    Re: How to improve the Pairwise

    Originally posted by dxmnkd316 View Post
    I'd use a KRACH PWR like the RPI fan page does.
    what does krach look like this year?

    Leave a comment:


  • dxmnkd316
    replied
    Re: How to improve the Pairwise

    I'd use a KRACH PWR like the RPI fan page does.

    Leave a comment:

Working...
X