New Website: RPI for Division I Women's Soccer

UPSF,

I am sure it is safe to assume you saw the discrepancy in Mississippi's Record between your RPI and the NCAA's. That must've been a good win to jump them from 31-20 in the RPI. Despite the flawed data you have received, fantastic job as always.

aleppiek wrote:UPSF,

I am sure it is safe to assume you saw the discrepancy in Mississippi's Record between your RPI and the NCAA's. That must've been a good win to jump them from 31-20 in the RPI. Despite the flawed data you have received, fantastic job as always.

Yup, in my "audit" process, I found that error (v Arkansas) and some others. I'm in the process of producing a new report.

gnarly wrote:What other criteria does the seeding comm use to pick the seeds? Are they looking at NC rpi besides the adjusted one. seems like last year one of the top RPI teams got dropped from a seed somehow?

The answer to your question depends on how much information you want. Rather than try to answer here, I'll direct you to the RPI website and where to look there. I'll put the link to the website home page below.

Once at the home page, use the navigation bar on the upper left of the page to go to the different pages.

The first page you will want to look at is "NCAA Selection, Seeding, and Bracketing Criteria." This provides a comprehensive review of the criteria.

If you want to know how the NCAA has applied the criteria in actual practice, go next to the "NCAA Actual Tournament Practices" page. This covers how the NCAA applied the criteria in 2007 and 2008. In particular, I suggest you look at both the Selection and the Seeding sections for 2008, since among other things they discuss how the Women's Soccer Committee most likely used the Non-Conference RPI in making some critical at large selection and seeding decisions.

If you want more information on the Non-Conference RPI, go to the "Non-Conference RPI" page.

Finally, if you want even more information about the Non-Conference RPI, including data from which you can form an opinion on whether it's a legitimate criterion, go to the "RPI Compared to Other Systems" page.

All of this probably will be more information than you want, so just stop when you have gotten enough info.

By the way, if you want to print one of the pages, there's a "click on" command for printing at the bottom of each page.

Here's the home page link: http://sites.google.com/site/rpifordivisioniwomenssoccer/

For those following the RPI reports, I've just posted on the RPI website an updated report with corrections for some data errors I had in my system. They don't make much difference for the teams most of us follow.

For those interested, based on what happened in 2007 and 2008 looking at the RPI rankings as of this point in the season:

1. The four #1 seeds are likely to come from the top six teams in the current rankings: Stanford (1), North Carolina (2), Portland (3), Central Florida (4), Florida State (5), and UCLA (6).

2. For the other 12 seeds, they are likely to come from the top 30 in the current rankings, which includes Santa Clara (13), USC (14), Washington State (22), Texas A&M (23), California (24), Arizona State (25), and Oregon State (26).

3. It is likely that the top 32 in the current rankings will make the Tournament bracket. In addition to the teams mentioned above, this includes Washington (31).

4. Teams ranked 33 through 71 should be considered currently "on the bubble." These include Oregon (34), Colorado College (35), Illinois (39), Miami (43), Pepperdine (57), San Diego (61), and St Mary's (67). Loyola Marymount at 73 is just outside this group.

5. If all of the current top 32 make it into the bracket, they represent nine conferences. If the top 32 include the champions of each of those conferences, as seems likely, then the remaining 23 teams are likely to get at large positions in the bracket. Since there are 34 at large positions altogether, this means that the 39 "bubble" teams likely will be competing for 11 positions in the bracket.

There may be outliers who make it into the bracket (#93 as of this point in the season has made it in), or who currently are highly ranked but don't make it into the bracket (#22 as of this point has been left out), but more likely the above "guidelines" will hold.

So UNC loses twice on the weekend and still has a substantial lead over us in RPI? (According to the nc-soccer site.) Florida State overtakes us and UCLA is on our heels and sure to by-pass us if they win their remaining games?

Can we possibly end up with a #2 seed even if we win out? Ugggh!

In terms of the RPI, you may be correct. However, there are other criteria that will help the Pilots if they beat San Diego and Santa Clara. I don't have time to explain right now but will later. D0N'T LOSE HEART!

Massey has published new ratings for teams and for conferences, covering games through October 25. Use the following link: http://www.mratings.com/rate.php?lg=csocw

Thanks for the link.

Very interesting that UP-A&M ranks as the biggest upset among ranked teams. (And PSU over OSU is in the top 10 as well.)

SoreKnees wrote:So UNC loses twice on the weekend and still has a substantial lead over us in RPI? (According to the nc-soccer site.) Florida State overtakes us and UCLA is on our heels and sure to by-pass us if they win their remaining games?

Can we possibly end up with a #2 seed even if we win out? Ugggh!

SoreKnees has spotted that it appears five teams are going to be competing for the four #1 seeds: Stanford, North Carolina, Florida State, Portland, and UCLA. Right now, that is those teams' order in the RPI rankings.

In addition, using the nc-soccer website's "projecting the RPI forward" feature, if UCLA wins out, it may pass the Pilots, just barely, in the RPI rankings, if the Pilots also win out. This is not absolutely certain, however, because it depends on how the two teams' opponents (already played and to be played) do over the balance of the season. Further complicating this is that UNC and FSU either will play each other again in the ACC tournament or that matchup will be avoided because one or both of them has lost an ACC tournament game either by outright loss or by shootout loss (which counts as a tie for RPI purposes). So, even if UNC and FSU win out for the balance of the regular season leading up to the ACC tournament, one or both of them is going to have at least a tie inserted into its record, which counts as half a win and half a loss for RPI purposes. The "projecting the RPI forward" feature can't figure that outcome into its projection.

However, in addition to considering the RPI (and it's cousin, the Non-Conference RPI), the NCAA also has two other primary criteria: (1) head to head results (against teams in competition, in this case, for a #1 seed) and (2) results against common opponents (again, against teams in competition for a #1 seed, in this case).

The Pilots will have no head to head results against the other four teams competing for a #1 seed. However, the Pilots and UNC will have a result against a common opponent, which is Miami, with the Pilots winning and UNC losing. This makes Miami's win yesterday very important, as it could justify a Portland seed ahead of UNC. Combine that with UNC's having beaten UCLA, giving a head to head result almost surely putting UNC ahead of UCLA, and you get a basis for having a Portland-UNC-UCLA order. Combine that with FSU's win over UNC and you get a reasonable seeding order of Stanford (1)-FSU (2)-Portland (3)-UNC (4)-UCLA (5). Whether Florida State holds that position depends on its last pre-ACC tournament games and how it does in the tournament.

By the end of the season, there will be another "results against common opponents" that will come into play, which will be UCLA's and the Pilots' results against San Diego. UCLA and San Diego tied, so if the Pilots defeat San Diego, that provides another basis for a seed ahead of UCLA.

If the Women's Soccer Committee, after considering these factors, still is undecided, then it also will look at both results over the last eight games (game outcomes and strength of opponents) and results against teams already selected for the bracket (but excluding results against teams ranked #76 or worse by the RPI).

I should mention, however, that whereas the above-described criteria are binding on the Committee for purposes of at large selections, they are not binding as to seeds. Still, the Committee considers them and I believe would follow them unless it thinks there's a very good reason to do otherwise.

What this all makes clear is that the next two games are HUGE. The Pilots must win against Santa Clara and they must win against San Diego. Which is exactly what we've thought all along.

There is another result against common opponent with UNC.

UNC beat TAMU, we lost to them, so we are even in that regard.

Geezaldinho wrote:There is another result against common opponent with UNC.

UNC beat TAMU, we lost to them, so we are even in that regard.

I missed that one, so there's no basis there for the Pilots being seeded ahead of UNC. That leaves up in the air the question of what happens in the ACC tournament, so far as UNC and Florida State are concerned. The San Diego game remains the key in relation to UCLA.

GO PILOTS!

I just have posted a new RPI report on the RPI website, including games through October 25. Use the following link to go to the home page and, from there, go to the RPI Reports page.

The new report is in the form of an Excel spreadsheet attachment at the bottom of the RPI Reports page. As an additional attachment, I have posted a Non-Conference RPI Report. The page, towards the bottom, also includes average RPI's by conference and by region.

For those who are trying to "predict the bracket," here are the guidelines I have developed based on what happened in 2007 and 2008. These most definitely are guidelines, and there can be exceptions:

The four #1 seeds likely will come from teams ranked 1-6

All seeds likely will come from teams ranked 1-30

Teams ranked 1-32 are likely to be in the bracket

Teams ranked 33-67 should be considered "bubble teams"

Here's the link: http://sites.google.com/site/rpifordivisioniwomenssoccer/Home

For those of you who are interested, I've developed what I think is a pretty nifty tool for trying to predict the NCAA Tournament bracket. To access it, go to the RPI website home page and, from there, to the "Predicting the Bracket" page. Once there, I suggest you review the information on that page, particularly the information under the heading "Bonus Tool," and then download the tool itself. The tool is in the form of an Excel spreadsheet attachment titled "Bracket Formation Data" at the bottom of the "Predicting the Bracket" page. The tool contains all the information you will need to apply the NCAA decision-making criteria to the top 70 RPI teams as of October 25.

As a reminder:

The primary criteria are:

Adjusted RPI (including Non-Conference RPI);

Head-to-head results of teams competing for a seed or at large selection; and

Results against common opponents of teams competing for a seed or at large selection.

The secondary criteria are:

Results over the last eight games (results and strength of opponents); and

Results against teams already selected for the bracket but excluding automatic qualifiers with an RPI rank of 76 or worse.

Once again, here's the link for the RPI website: http://sites.google.com/site/rpifordivisioniwomenssoccer/

Kansas and UAB have very low NCRPI (in the 20's)but at the end of the bubble teams on RPI. Any idea if how much weight that will get them to get in the tourney? Seems like they have done poorly in the last half of the season. Uof O played well a few years ago but didn't get in even with wins against the S. Cal teams the last weekend of league play.

Thanks for the new tool. Will be fun to watch the last few results impact the mix.

gnarly wrote:Kansas and UAB have very low NCRPI (in the 20's)but at the end of the bubble teams on RPI. Any idea if how much weight that will get them to get in the tourney? Seems like they have done poorly in the last half of the season. Uof O played well a few years ago but didn't get in even with wins against the S. Cal teams the last weekend of league play.

My best guess is that if the Women's Soccer Committee sees two (or several) teams as close after reviewing the RPI, head-to-head results, and results against common opponents, then it may use the NCRPI as a basis for decision-making. If you go to the "NCAA Actual Tournament Practices" page on the RPI website, there's an analysis of the decisions the Committee made in 2007 and 2008. If you look at the part of the page on 2008 at large decisions, you'll see why I've concluded that the Committee may have used the NCRPI as a basis for advancing some teams ahead of others in the at large selection process. I have to warn you, though, that the analysis is a little tedious. But, if you're interested in this kind of stuff, I think the analysis shows pretty well how the Committee used the NCRPI.

My unofficial detailed RPI report for games through 11/1 now is posted on the RPI website: http://sites.google.com/site/rpiford...er/rpi-reports

Those interested can compare it to the NCAA's official interim RPI report. I'm hoping the Geez has posted the link to the official report. The two are quite close although not identical, since I'm only guessing the bonus and penalty amounts that go into the adjustment process. Also, there are a couple of data inconsistencies, which I lay out at the above link.

The link you have here doesn't work for me, though the one you have on BS does. Here it is:

http://sites.google.com/site/rpifordivisioniwomenssoccer/rpi-reports

(scroll to the bottom)

I gave the link on the Polls thread for the NCAA version. here that is:

http://sites.google.com/site/rpifordivisioniwomenssoccer/rpi-reports


The top 20 are very close between UPSF and NCAA,

except that FSU and UCLA , and also BC, and ND are muddled (4-5,6-7) there is only .0001 between FSU and UCLA, and .0005 between BC and ND.

I hadn't checked the size of the differences, but that confirms what I thought, which is that my differences from the NCAA probably are due to my not having exactly correct bonus amounts (and, farther down in the rankings, penalty amounts). It's also conceivable that the NCAA's error in the Jacksonville U v Mercer game result (tie rather than win/loss) is responsible for the difference.

After the season's over, maybe I'll take a run at coming up with better penalty amounts.

Anyone think the teams currently with the top 16 RPI that won't be in the list of seeds released next week. The following teams have high NC RPI: ND-27, FL 26, PSU 35, OSU 20, and WSU 29. Teams higher than 16 with low NCRPI are UGA 9, MD 10, Uconn 6. I don't think IND and Vill which have low NC RPI will be considered since RPI is too high. PSU is one school that I believe was unseeded before. Of course we need to wait for final results this weekend, but I expect the NCRPI won't change much since everyone is playing conference games this time of year.

I think the Non-Conference RPI has less weight in the Committee members' minds than the RPI. It should have less weight, as it's less reliable as a representation of teams' accomplishments during the season. Nevertheless, you're right that it is the only explanation for Penn State not getting a seed last year.

Based on what happened in 2007 and 2008, it's likely that the 16 seeds will come from the top 19 teams in the current RPI ratings. There is the possibility of outliers getting seeds (#28 seems to be the outside limit), but that would be atypical. The four #1 seeds should come from the top 6.

There was a posting on NCAA.com today from the chair of the selection committee that talked about the process they use for selection.

http://www.ncaa.com/blog/2009d1womenssoccer/2009/11/a-message-from-the-chair.html

When the committee does get behind closed doors in Indianapolis, the selection and seeding of teams for the championship will be driven by three primary criteria. The first is the team's RPI - which consists of their winning percentage and their strength of schedule, the second is their head-to-head results against other teams being considered for selection, and third is a comparison with other teams under consideration of their results against common opponents. The hardest call for the committee is in comparing teams with very similar profiles, but no head-to-head results and no common opponents. That is when we often have to turn to our secondary selection/seeding criteria which includes results against teams already selected for the tournament and late season performance (last 8 games).

It seems to confirm the view that they start with RPI and then look at the other criteria to see if there are reasons to override that ordering for selection and seeding.

In order to "bump" up, it appears the team much show it was superior in other criteria. What isn't clear to me is when secondary critera can override the RPI and when it can't. with often tenuous linkage between regions, it seems those comparisons often aren't valid. Even in the same regions that might be true.

One of the secondary criteria, for example, is results for the last 8 games. I submit that test is a lot easier if you play in the SWAC than in the ACC...

Geezaldinho wrote:There was a posting on NCAA.com today from the chair of the selection committee that talked about the process they use for selection. .... What isn't clear to me is when secondary critera can override the RPI and when it can't. with often tenuous linkage between regions, it seems those comparisons often aren't valid. Even in the same regions that might be true.

One of the secondary criteria, for example, is results for the last 8 games. I submit that test is a lot easier if you play in the SWAC than in the ACC...

Great catch, Geez. That's the kind of declaration I love to have in my files and to be able to quote on the RPI website in the material on how the process works.

What I interpret the criteria as saying, and what Bradshaw is saying, which is consistent with what my credible source has said, is that the Committee first tries to make a decision based on RPI, head-to-head, and teams' results against common opponents. But, suppose they can't make a decision based on those criteria. For example, suppose two teams' RPIs (and NCRPIs, which they use but talk about only in Appendix I of the Championship Manual) are very similar which is what I take Bradshaw to mean when he says the teams have "very similar profiles," they have no head to head game, and their results against common opponents are the same. At that point, they have no basis in the primary criteria for making a decision and have to go to the secondary criteria. Since the RPIs are essentially the same, the secondary criteria aren't over-riding the RPI, they're providing a needed other basis for making a decision.

Regarding the last 8 games, here is exactly what the criterion says: "Late season performance -- defined as the last eight games including conference tournaments (strength and results)." So, in addition to looking at the teams' records over the last eight games, the Committee also looks at the strength of the teams they played. In terms of strength, I assume they primarily look to the RPIs of those teams.

[quote="UPSoccerFanatic"]

Geezaldinho wrote: But, suppose they can't make a decision based on those criteria. For example, suppose two teams' RPIs (and NCRPIs, which they use but talk about only in Appendix I of the Championship Manual) are very similar which is what I take Bradshaw to mean when he says the teams have "very similar profiles," they have no head to head game, and their results against common opponents are the same. At that point, they have no basis in the primary criteria for making a decision and have to go to the secondary criteria. Since the RPIs are essentially the same, the secondary criteria aren't over-riding the RPI, they're providing a needed other basis for making a decision.

But that's just it-- the RPI's are NEVER exactly the same, are they? at least I don't see it. If the NCAA recognized a margin of error, then sure, but they don't. They treat the RPI as statistically accurate.

So that must mean RPI always trumps secondary criteria.

Last year Notre Dame jumped over the 4 top teams in the RPI based on the other two primary criteria, so we know That's possible.

Do you have an idea what teams jumped over people based on what must have been secondary criteria, and how big a leap in the RPI seeding did that encompass?

I haven't been able to find any cases in which it appeared that the NCAA used the secondary criteria to advance teams. From the detailed review I did of the 2008 tournament bracket, in those cases in which the NCAA "jumped" teams over teams with better RPIs, the secondary criteria wouldn't have justified it. The justification appeared to have been the Non-Conference RPI.

I guess if we get a case in which only the secondary criteria could have justified a "jump," then we'll know how small an RPI difference there must be in order for the Committee to turn to the secondary criteria. I'm guessing it's a pretty small difference. That probably would occur as to at large selections and would be less likely to occur as to seeds and particularly as to the four #1 seeds. This is because the RPI rating differences are greater at the top of the rankings (and at the bottom) than in the middle.

Looking at the the latest RPI update by UPSF I think we can hone in on what the bracket outline might look like. Seems like certain teams host annually.

By region:

Pacific NW. Portland to host with likes of maybe PSU and one of the pac 10 teams. U of O out on w/l record. Leaving either UW or OSU to be another partner. WSU as seed likely to get shipped unless EW wins Big Sky.

N. Cal: Stanford to host with likes of Cal Poly or SJ st. With SCU likely to get seed they could host with partner Cal.

S. Cal: UCLA to host with SD st and SD. USC gets shipped. Maybe A&M way.

You could have host in the Rockies: could be BYU or Col Coll or Den and maybe USC or WSU could be seed that goes there.

Great Lakes area:
We have group of teams not seeded that are in MIL, Wis, Marq that could have seeded team come there. Minn also in area

ND likely to host with lots of partners within 400 m (C. Mich--Mid Am, Pur or IND, Ill st (MVC) or Dayton (Alt 10).

Ohio St likely to host with similar partners

N.E.: BC likely to host. Several local partners BU, NE, Harvard, CCSU, Uconn

FL area: Could have FSU, FL and UCF all seeded. My thought is FSU hosts others get shipped as not enough teams in area to all host. Local partners for FSU could be LSU, Miss, Georgia, or Aub from SEC. and maybe samford or Atl sun representitive.

SE area

UNC likely to host with partner like S. Car, UNC-GR, Charlotte. WF should get seed, but will they bring in Loyola, OVC rep, WV, Georgia or Patriot rep who are all within 400 Mi?

Penn State: could host with WV, MD and some leftovers from Boston/ Conn area.

Texas area: I think texas is where someone like USC could land. You could have A&M host, plus Tex St, and bring in SWAC Grambling or Ark_PB.

VT probably hosts with some of the mid atl region teams.

I'll have update once we get lock on auto bids and final RPI. Not worth it to do a whole bracket now with so many auto bids outstanding.