2021 RPI and NCAA Tournament Bracket

Discussion in 'Women's College' started by cpthomas, Aug 28, 2021.

  1. cpthomas

    cpthomas BigSoccer Supporter

    Portland Thorns
    United States
    Jan 10, 2008
    Portland, Oregon
    Nat'l Team:
    United States
    It is a new season and I will be providing weekly RPI and NCAA Tournament bracket information as we go through the season.

    I have posted three new pieces at the RPI and Bracketology blog. They are:

    2021 Season: Background for Upcoming Reports If you are going to follow my reports, then it is almost a must for you first to go through this backgrounder. It explains the method I use to produce a simulation of the entire season, its limitations, and things you can do to see how your team will end up if it has different results than what the simulation says.

    2021 Season: Pre-Season Simulated RPI Ratings and NCAA Tournament At Large Selections and Seeds This shows simulated ratings and Tournament at large selections and seeds based on simulated results for all games during the season.

    2021 Season: 8.22.21 Simulated RPI Ratings and NCAA Tournament At Large Selections and Seeds This shows simulated ratings and Tournament at large selections and seeds based on the actual results of games played through August 22 and simulated results of games not yet played.

    With the August 22 simulation, things already have evolved from the pre-season simulation.
     
  2. Val1

    Val1 Member+

    Arsenal
    Mar 12, 2004
    MD's Eastern Shore
    Club:
    Arsenal FC
  3. espola

    espola Member+

    Feb 12, 2006
    I find the idea of a pre-season simulation amusing.
     
  4. cpthomas

    cpthomas BigSoccer Supporter

    Portland Thorns
    United States
    Jan 10, 2008
    Portland, Oregon
    Nat'l Team:
    United States
    That is a good way to look at a full season pre-season simulation, although if you look at the top bunch of teams many (but not all) of them still will be there at the end of the season and some that are not there now will be there.
     
    Soccerguy1022 repped this.
  5. Soccerguy1022

    Soccerguy1022 Member

    Manchester City
    United States
    Nov 28, 2018
    Will you be posting week to week simulations?
     
  6. cpthomas

    cpthomas BigSoccer Supporter

    Portland Thorns
    United States
    Jan 10, 2008
    Portland, Oregon
    Nat'l Team:
    United States
    Yes. Next one to come out later today.
     
    Soccerguy1022 repped this.
  7. espola

    espola Member+

    Feb 12, 2006
    That's what I would have said based on no arithmetic whatsoever.
     
  8. cpthomas

    cpthomas BigSoccer Supporter

    Portland Thorns
    United States
    Jan 10, 2008
    Portland, Oregon
    Nat'l Team:
    United States
    True, but ... the arithmetic will do better than you would do with no arithmetic unless you simply get lucky. If you read the initial Background post, which I am guessing you have not, you would get a good picture of the limitations of the arithmetic but also how to use the simulation and the related information to see how a team may do if you do not agree with the arithmetic for that team. You also would see that as the season progresses, the arithmetic will become progressively more reliable.
     
  9. cpthomas

    cpthomas BigSoccer Supporter

    Portland Thorns
    United States
    Jan 10, 2008
    Portland, Oregon
    Nat'l Team:
    United States
  10. cpthomas

    cpthomas BigSoccer Supporter

    Portland Thorns
    United States
    Jan 10, 2008
    Portland, Oregon
    Nat'l Team:
    United States
    McSkillz repped this.
  11. Tom81

    Tom81 Member+

    Jan 25, 2008
    So apparently FSU will lose to UNC, UVA and Duke.
    It's possible; meanwhile South Carolina, Xavier and Michigan are ranked higher.
    Hmmmmm!
     
  12. Nooneimportant

    Leeds United
    Jan 12, 2021
    Don’t read much into these. They don’t mean anything and they are a fun application of math to stimulate discussion, but they are based on bulk and it will get a team or 2 horribly wrong right now. For example, Yale is a 0-4 and anyone who has watched them know they are not a good team, but somehow his system says they will make the tourney. I would love to know, however, how this simulates Yale to go 9-1-3 the rest of season with Hofstra, Central CT, Brown, Harvard, and Princeton still on the schedule just to name a few. Math is nice and I was a math major, but never doubt that sometimes your eyes are your best judge in specific cases. Although this specific case is so completely ridiculous it does make you doubt the system.
     
    L'orange repped this.
  13. cpthomas

    cpthomas BigSoccer Supporter

    Portland Thorns
    United States
    Jan 10, 2008
    Portland, Oregon
    Nat'l Team:
    United States
    The system and its limitations are explained in detail in the blog piece that is recommended reading for those interested in the simulation. A major problem this year is that the results from the 2020-21 school year are not usable. If you were to look at the Yale trend through 2019, you would see why the system puts Yale where it is right now and also would understand how it projects the Yale results for the remainder of the season.

    Since the system is based strictly on rating trends, it cannot take into account roster changes, player injuries, or anything else of a detailed nature. That, of course, is its major weakness. It thus is just one way of looking at what might happen, additional to other systems that are driven by more season-specific data or the opinions of those who have a lot of knowledge about the details of rosters and so on.

    Soon, I will be changing over the result simulation process to being based on this year’s RPI ratings (i.e., once they become semi-realistic) rather than my pre-season simulated ratings. It that point, there should be a correction for teams where the current simulation seems to have gotten their ratings horribly wrong. And, over the course of the remaining weeks of the season, the simulation will become more and more accurate.

    So far as where the current simulation fits Florida State, it is not going to get teams in exactly the correct positions at this stage, under any circumstance. It does, however, have Florida State in more or less the right ball park, which is all that reasonably could be expected. Florida State is an interesting case, as its history is a little more varied than some of the other teams at the top of the ratings, which is why its simulated position is not quite as good as a few others.
     
    BigBear and L'orange repped this.
  14. FreshPow

    FreshPow New Member

    United States
    Jun 12, 2017
    Nat'l Team:
    United States
    Does anyone know when NCAA starts releasing their actual RPI? I always love comparing cpthomas work with theirs...lots of fun.
     
  15. cpthomas

    cpthomas BigSoccer Supporter

    Portland Thorns
    United States
    Jan 10, 2008
    Portland, Oregon
    Nat'l Team:
    United States
    Normally, the NCAA has started releasing RPIs at the end of the fifth week of the season. That would be a release roughly September 20 this year.
     
    socalsoccer23 and West Ender repped this.
  16. espola

    espola Member+

    Feb 12, 2006
    You use last year's results?
     
  17. cpthomas

    cpthomas BigSoccer Supporter

    Portland Thorns
    United States
    Jan 10, 2008
    Portland, Oregon
    Nat'l Team:
    United States
    No, as explained in the introductory piece to all of the simulation results reports, the simulation does not use the 2020-21 results. Teams simply did not play enough games (on average), did not play enough non-conference games, and did not play enough out-of-region games to allow for legitimate rankings of all teams within a single system. So, the simulation this year is based on historic ranking trends through 2019.

    The lost year, of course, is a problem but I decided it would be less of a problem than including it would have been.
     
  18. cpthomas

    cpthomas BigSoccer Supporter

    Portland Thorns
    United States
    Jan 10, 2008
    Portland, Oregon
    Nat'l Team:
    United States
    I have posted 2021 Season: 9.12.21 Simulated RPI Ratings and NCAA Tournament At Large Selections and Seeds. It shows simulated ratings and Tournament at large selections and seeds based on the actual results of games played through Labor Day and simulated results of games not yet played.

    Next week, I will base the simulated results of games not yet played on then-current team actual RPI ratings, so the overall results and NCAA Tournament bracket may be fairly different. I suspect it still will be a good distance from where things ultimately will end up.
     
    socalsoccer23 repped this.
  19. uncchamps2012

    uncchamps2012 Member

    Jul 9, 2011
    Thanks for posting all this. Very Interesting. And no complaints about predicting the Heels to go undefeated :)
     
  20. Carolina92

    Carolina92 Member

    Sep 26, 2008
    What does the current actual RPI look like? Or do you not post that?

    Would be curious if you could also measure a given team's possible upside or downside given their schedule and historic results winning games they statistically shouldn't? Like if you know team A has typically had 2 unlikely wins (or losses) in a given season then their potential high end ranking could be X (or low end) as opposed to just one single simulated ranking. Or if that's something that's already basically covered by your historical data, what about including rankings if a team (from this current point in the season) ran the table on the rest of their schedule or lost the remainder of their games. Just to get a sense of what a team's range of potential actually is. Not sure if any of this makes sense from a statistical perspective.

    Also, from a simulated results perspective are you including games that are forfeited for COVID reasons even though those games don't actually measure a team's strength?
     
  21. cpthomas

    cpthomas BigSoccer Supporter

    Portland Thorns
    United States
    Jan 10, 2008
    Portland, Oregon
    Nat'l Team:
    United States
    1. Regarding current RPI ratings and ranks, I do not publish them yet and neither does the NCAA. You can get them, however, using this link to the College Women’s Soccer Schedule website presented by All White Kit (under the guidance of Chris Henderson). By clicking twice on the Adjusted RPI column header, you can put the teams in rank order starting with #1. Also, next week the NCAA will start publishing RPI rankings (and hopefully, RPI ratings).

    Since it appears you are not familiar with the linked website, take a stroll around it by clicking on the Information header at the top left of the page, which will give you links to other website pages. It has a lot of good and current information.

    2. I do not do the kind of analysis you refer to regarding whether specific teams do better or more poorly than the statistics indicate they should (although it would be possible to do it). On the other hand, it is possible to use the data in the Excel workbook to which my blog includes a link, to estimate where a team would fit within the ratings and rankings if it were to have a different end-of-season record than what my simulation indicates. You may be able to figure out how to do that by reviewing the introductory post that precedes the weekly simulation reports. This would allow you to do your own best case and worst case estimation of the range within which a team might end up.

    Also, in two weeks I will be adding additional information that uses teams’ then current RPI ranks to show the teams that are within the historic range, as of that date, of those that have gotten NCAA Tournament at large selections and seeds -- and also, the historic range of teams that currently are in the running but that might not get at large selections or seeds. This essentially will give you a pretty good picture of which teams still have reasonable hopes of getting into and getting seeded in the Tournament and which ones still have reasonable fears of not getting in or getting seeded.
     
    Carolina92 and L'orange repped this.
  22. Soccerhunter

    Soccerhunter Member+

    Sep 12, 2009
    At the risk of jumping the gun, CP, (at 7 or 8 games into the season we all know that the RPI values will change mightily) I've tabulated the conferences with teams represented as of this date in Chris Henderson's current RPI. It's going to be really interesting to see how the current conference strengths pan out (and for that matter how the individual teams do) with regard to your projections at this point or with a current sample of a real RPI. I personally am betting that with a few outliers your current projections are going to be pretty close.

    This is the juncture where the larger conferences have finished their non conference schedule and will embark on their internecine contests. Certainly the ACC's and Big 10 current status in the top 50 will drop significantly but how so the others?

    As of 9/14/21 All White Kit adjusted RPI, the Top 10, 25, and 50 teams by conference:

    Top 10 Teams by Conference

    3 ACC
    1 SEC
    1 IVY
    1 Big 12
    1 Big10
    1 PAC 12
    1 Big East
    1 American Athletic

    Top 25 Teams by Conference

    6 ACC
    3 SEC
    3 PAC 12
    3 American Athletic
    2 Big 12
    2 Big 10
    1 Big East
    1 Ivy
    1 West Coast
    1 Western Athletic
    1 Conference USA
    1 Atlantic 10

    Top 50 Teams by Conference

    10 ACC
    8 Big 10
    7 SEC
    4 PAC 12
    4 Big 12
    3 American Athletic
    3 Big East
    3 West Coast
    2 Conference USA
    2 Ivy
    1 Atlantic 10
    1 Western Athletic
    1 Colonial
    1 Horizon
     
  23. cpthomas

    cpthomas BigSoccer Supporter

    Portland Thorns
    United States
    Jan 10, 2008
    Portland, Oregon
    Nat'l Team:
    United States
    Good info, SoccerHunter. As we move out of the non-conference part of the season, there are a couple of things to bear in mind regarding the RPI:

    1. The conference part of the season tends to sort out the teams within a conference in relation to each other. This is true in terms of conference standings but also is true in relation to the RPI. Where it affects the RPI relative standings within a conference is through team winning percentages, which in effective weight by the end of the season comprise 50 percent of the RPI rating. It only slightly affects the strength of schedule portion of the RPI, since the teams in a conference are playing roughly (but not exactly) the same strength of schedule during the conference portion of the season.

    2. The conference winning percentage during the non-conference part of the season is very important. The winning percentage of opponents, in effective weight by the end of the season, comprises 40 percent of the RPI rating. And, the winning percentage of conference opponents during the conference part of the season is roughly 50 percent for every conference. (During conference competition, for every conference opponent win, there is a matching conference opponent loss.) This means, as an example, that the ACC teams will benefit greatly from the really good non-conference winning percentage its teams have had this year. I suspect this will cause ACC teams, overall, to rise in the rankings as compared to teams from other conferences whose conference opponents will contribute less to their winning percentages of opponents.

    When we get to NCAA Tournament at large selections and seeds, the other thing to bear in mind is that results against Top 50 opponents join the RPI as being very important, especially for at large selections. That factor, based on my studies, is very highly skewed towards good results (wins or ties) against very highly ranked opponents. For mid-level teams from a strong conference (ones that do not have good results against highly ranked in-conference teams), this means their NCAA Tournament prospects will depend greatly on their good results against highly ranked opponents during the non-conference part of the season. Thus, as an example, this means that a team like Pittsburgh, if it finishes with no good results against highly ranked conference opponents, will have to be hoping that Oklahoma State almost runs the table in the Big 12 and thus moves high in the rankings, since its win over Oklahoma State looks to be its only good non-conference result with the potential to be against a highly ranked opponent.
     
  24. Nooneimportant

    Leeds United
    Jan 12, 2021
    I feel like Wake Forest would be another one in this category. Not sure who has an easier non-conference schedule, but at least Pitt has a non-con win over OK st. Wake doesn’t play FSU, UNC, or Clemson so I have to imagine the games against UVA, Duke, and even Pitt will be huge for them. Depending on those results they could be an interesting case study to see how the committee treats them.
     
    cpthomas repped this.
  25. Enzo the Prince

    Sep 9, 2007
    Club:
    CA River Plate
    If they beat UVA or Duke then they're golden. They'll have enough total wins, and they'll get a big RPI boost from such a win. But we have no evidence they're capable of a win like that. I would expect them to lose all three of those games. I think the big make-or-break ones will be Louisville, Va Tech, and NC State, assuming they can beat Syracuse and Miami. Win those two plus one or two of Lvill/Va Tech/NCSU and it MAY be enough for the committee.
     

Share This Page