Page 3 of 3 FirstFirst 123
Results 31 to 39 of 39
  1. #31


    Quote Originally Posted by Shades View Post
    Realtime has Indiana at 12-0 when they’re actually 11-1. I wonder how many other errors they have.
    Interesting. For starters, they also have Grambling State (the team who beat Indiana) wrong, at 2-9 when they're actually 3-8. Obviously RealTime's data source is not clean.

    This data error (wrongly) puts Indiana at a current RPI rank just ahead of Mississippi State and Iowa.

    Even more interesting is how sensitive (and inconsistent) their Gamer algorithm is to this error. Their Indiana detail page predicts an end of season RPI of 4; going with a season of one loss. Yet the game predictions show them predicted to be undefeated both in preseason and Big Ten. So they call Indiana winning the Big Ten.

    So parts of their algorithm know about the Grambling loss, other parts don't. And it totally throws off their Gamer prediction, which is quite sensitive to the data error. Garbage in, garbage out.
    Last edited by CutDownTheNet; 12-24-2018 at 08:38 AM.

  2. #32


    Quote Originally Posted by Shades View Post
    Realtime has Indiana at 12-0 when they’re actually 11-1. I wonder how many other errors they have.
    Regarding RealTimeRPI (in)accuracy, I did find some other blog that said it's well known that RealTime often has data inaccuracies of various types. I've kept an eye on its Indiana stats, and after over a week it finally corrected its bad Indiana W/L record to 12-1, reflecting its Grambling State loss and its overtime win over Illinois. Adding the missing loss and the win drops its RPI ranking 6 slots to 16, while Minnesota's win over Badgers raises its RPI rank 4 slots to number 23. Indiana still has the highest B1G RPI since it has the most difficult SoS in the B1G.

    It predicts end of season RPIs of 10 for Minnesota and 7 for Indiana. It also still predicts Indiana will be undefeated in the Big Ten. Gamer still predicts Minnesota to have a 5 loss season. It thinks 4 of those 5 losses will be by only 2-4 points. But it thinks the Gophers will lose @Indiana by 21 points.

    By its behavior, it's clear that Gamer quite literally uses (its possibly erroneous) raw RPI as the ranking estimator from which to predict wins and losses.

    And RPI is mostly a measure of SoS and only partly a measure of team basketball quality. So Gamer has Indiana cleaning house in the Big Ten merely because Indiana has the strongest schedule.

    This is a good example why RPI is only partially useful for ranking basketball teams, especially this early in the season.

  3. #33


    Here’s some good news. Incarnate Word finally got their first win and are now 1-0 in conference.
    Now for the bad news. Coppin St is still winless at 14-0.

  4. #34


    Quote Originally Posted by Shades View Post
    Here’s some good news. Incarnate Word finally got their first win and are now 1-0 in conference.
    Now for the bad news. Coppin St is still winless at 14-0.
    Poor Coppin State - everybody's favorite cupcake opponent. Gamer projects they will win 6 of their 17 league games though. Bad news for the Gopher RPI is that they predict Coppin State to have an end of season RPI rank of 304 out of 351.

    In looking at their record, I noticed another RealTimeRPI data error. They have Minnesota down as currently having a 2-1 record in the Big Ten, although they do have the correct overall record. Gamer now has us predicted for 7 losses, including @Michigan State, @Northwestern, @Indiana, @Purdue, @Maryland and @Rutgers. But an end of season RPI rank of 25.

    Aided by their first win, Incarnate Word is now projected by Gamer to have an end of season RPI rank of 285. Both of these cupcake teams are dragging down the Gopher SoS, so I guess we should cheer them on.

    I was curious who had the #1 Strength of Schedule currently. Turns out it's Southern University with an SoS of 0.7149. In addition, Southern has record of 7 losses and no wins (a sparse schedule, apparently). That's good enough to give it a current RPI of 0.5362 for an RPI rank of 116. That RPI ranking puts it ahead of Penn State (7-7), Ohio State (5-7), and Wisconsin (10-5).

    Right next to Southern in the #2 SoS slot is Notre Dame with an SoS of 0.6878. That currently gives ND the #1 RPI of 0.7372. Also, ND's #2 SoS is enough to flip it's RPI relative to UConn, putting ND at RPI #1 and UConn RPI at #2 in spite of the fact that UConn beat Notre Dame for its only loss. Remember, RPI is 50% correlated to SoS (opponents won/loss record), and 25% correlated to opponents SoS, but only 25% correlated to the given team's won/loss record. And lowly UConn only has the 8th ranked SoS (0.6280), which makes RPI call it the second-best team in spite of it's undefeated record against a set of good teams, giving it an SoS only 0.0596 lower than ND's SoS.

    As a point of reference, Minnesota currently has an SoS of 0.5156. That's 0.1722 lower than ND's SoS. Or about 3X bigger delta than between ND and UConn. But Minnesota has an RPI rank of #25 right now. So 3X bigger SoS delta gets you a 24X bigger offset in RPI ranking number. And technically speaking, Minnesota has a better won/loss record (13-1) at the moment than Notre Dame (12-1).

    Moral of the story: If the Gophers had only refrained from playing lowly Coppin State and lowly Incarnate Word, there's a pretty good chance that right now the RPI ranking order would be:
    #1: Minnesota Golden Gophers
    #2: Notre Dame Fighting Irish
    #3: UConn Huskies

    That's what we get for being Minnesota Nice. "Send us your poor, your downtrodden, your crappy basketball teams - we'll play them anyway, even if it does kick us from 1st place down to 25th place (in RPIville)."

    Fortunately, the purpose of basketball is to win.

    The purpose of RPI is to inform us whether we played a sufficient number of winning basketball teams (but without getting beat by too many of them). RPI is like a game of chance, really, which is played at scheduling time (with the dealt cards revealed at game time). Like a poker game with many tables containing poker players of increasing skill. You get to pick what set of tables you play at (or in our case, the previous coach gets to pick). The payout ratio is higher at the better tables. On the other hand, if you play at a table over your skill level, you might get beat and maybe lose your stake. So RPI is a partial measure of basketball skill, but mostly a measure of scheduling skill.

    The purpose of the new NCAA men's NET metric is to give us all a good laugh about how ludicrous of an idea it was to lock a committee of mathematically/statistically incompetent basketball coaches in a room, and expect them to come up with something better than the already ridiculous RPI metric.
    Last edited by CutDownTheNet; 01-04-2019 at 03:39 AM.

  5. #35


    Through Sunday's (1-6-19) games with SOS and B1G record:

    12 Indiana 70 (3-0)
    15 Iowa 17 (1-1)
    21 Rutgers 37 (2-1)
    28 Michigan State 93 (1-2)
    38 Maryland 210 (2-1)
    71 Purdue 72 (1-2)
    72 Northwestern 77 (2-1)
    73 Minnesota 214 (1-2)
    79 Nebraska 79 (2-1)
    83 Michigan 159 (1-2)
    84 Penn State 14 (1-3)
    122 Illinois 118 (1-2)
    145 Wisconsin 240 (1-3)
    161 Ohio State 19 (0-3)

  6. #36


    With the Illinois loss, we got the RPI double whammy. Played a team in the lower half of the B1G, thus impacting SoS, plus lost. This dropped us from RPI #25 to #73. Gamer predicts us at a 7 loss season (only 5 more at this point), but its predictions are still a bit funky. Its predicts we beat Iowa, for instance, which may be hard to achieve.

  7. #37


    NCAA RPI through Sunday's games with SOS and conference record:

    14 Iowa 13 (2-2)
    16 Indiana 83 (4-1)
    20 Rutgers 47 (5-0)
    21 Michigan State 92 (2-3)
    25 Maryland 152 (4-1)
    46 Purdue 62 (4-1)
    64 Michigan 92 (2-3)
    80 Penn State 16 (2-3)
    83 Minnesota 206 (1-3)
    92 Northwestern 56 (2-3)
    95 Nebraska 24 (2-3)
    129 Ohio State 17 (1-3)
    148 Illinois 103 (1-4)
    153 Wisconsin 165 (1-4)

  8. #38


    Edit: I found a typo, plus RPI was updated, before the Iowa game. So updates are provided in-line.

    Gamer predicts Gophers at an end of season RPI of 64, with an updated B1G record prediction of 8 wins and 10 losses. I've learned to have a healthy disbelief of Gamer, but to the extent that its prediction might at least be in the ball park and if the NCAA selection committee sees little on our resume besides a Syracuse win and a 64 RPI, then we are indeed likely to be just off the bubble.

    Gamer still predicts us beating Iowa tonight (go Gophers, make that prediction true!).

    But (via its bias toward guessing teams lose road games) it thinks we lose all B1G road games. It thinks we lose @Wisconsin and @Nebraska by one point.

    I think we win more road games than that, so how well we do so, plus how well we protect the Barn, pretty much determines our NCAA fate (unless we also do really well in the B1G tourney).
    Last edited by CutDownTheNet; 01-14-2019 at 04:46 PM.

  9. #39


    The drop reaches triple digits. Through Sunday's games with SOS and conference record.

    12 Iowa 13 (5-2)
    16 Rutgers 35 (7-0)
    27 Maryland 140 (5-2)
    28 Michigan State 53 (3-4)
    29 Indiana 78 (4-3)
    34 Purdue 31 (5-2)
    65 Michigan 91 (3-4)
    71 Nebraska 25 (4-3)
    80 Northwestern 63 (4-3)
    99 Ohio State 26 (3-4)
    113 Penn State 22 (2-5)
    116 Minnesota 210 (2-5)
    161 Wisconsin 142 (1-6)
    164 Illinois 92 (1-6)

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts