2010 RPI

I'll use this thread for RPI info for the 2010 season, plus any other RPI-related info I think those of you who are RPI-interested would like to see. Here's my first piece for this season, related to the RPI and game locations -- home, away, neutral. For those of you who frequent BigSoccer, this is posted there too.

The RPI, as used for Division I women's soccer, is computed without regard to home field advantage. Here is some information related to that, for those who are interested in the RPI and its inner workings. This information is based on all of the regular season results (including end-of-season conference tournaments, but excluding the NCAA Tournament) for the 2007, 2008, and 2009 seasons.

As some know, I've developed a system for measuring how well teams' end-of-season RPI ratings correlate with the actual results of the games that the teams played during the season. It's a way of testing how well the RPI actually does its job of measuring how well teams performed. I also can use the system, however, to focus in on some particular variables, one of which is game location -- home, away, or neutral.

Without going into a lot of detail here (that will come in a lengthy paper to be written after the 2010 season), I can tell you that the data show that home teams on average do better than their RPIs say the should and away teams do more poorly.

One of the challenges is to be able to measure how much a home game helps a team and how much an away game hurts. My system is able to do that and, subject to review after the 2010 season, here is what it shows:

When a team plays at home, it performs on average as though it had an RPI that is 0.0100 higher than its NCAA-measured RPI; and when a team plays away, it performs as though it had an RPI that is 0.0100 lower than its NCAA-measured RPI. So, if Team A and Team B have identical RPIs, as measured by the NCAA, and if they play at Team A's home field, the game results on average will be as though Team A's RPI rating was 0.0200 higher than Team B's. If one is thinking about teams in the 30 to 60 ranking spots under the RPI, this 0.0200 difference is as though the teams were ranked 20 positions apart.

The problem arises when teams play imbalanced numbers of home and away games. This problem affects the RPI (Adjusted) that the NCAA uses, but affects even more the Non-Conference RPI that the NCAA also uses. Here is why it affects the Non-Conference RPI more:

The NCAA does a separate computation for each conference, to come up with the NCRPI. In computing the NCRPI for Conference A's teams, the NCAA disregards all of Conference A's intra-conference games. It includes in the computation only Conference A's inter-conference games, together with all games of all other teams whether intra- or inter-conference for those other teams.

Suppose Conference A has a home-away imbalance, such that it plays more home games than away games. What happens, when the NCAA deletes all of Conference A's intra-conference games is that it magnifies this home away imbalance. This is because, for intra-conference games, there is a home-away balance if the conference has an odd number of teams and one game off a home-away balance if the conference has an even number of teams. To illustrate, suppose Conference A over the 2007-2009 period played a total of 230 away games, 90 neutral site games, and 330 home games. Looking only at the away and home games, this would mean that of the non-neutral site games, Conference A played 58.9% at home and 41.1% away. Now, for purposes of the NCRPI, let's delete Conference A's intra-conference games. The new numbers are 65 away, 48 neutral, and 165 home. This would mean that of the non-neutral site games, Conference A played 71.7% at home and 28.3% away. Given that home field advantage results in Conference A's teams performing as though their RPIs were 0.0100 higher and their opponents' were 0.0100 lower than they otherwise would be, Conference A's home field imbalance is going to result in its teams having higher NCRPIs than they would have if there had not been an imbalance.

One question is, Are there home field imbalances that are significant enough to make the above-described characteristic of the NCRPI a problem?

One of the benefits of having complete data for three seasons is that I can identify home-away imbalances using the three seasons' data, which consist of data from over 9,000 games. Using those three years' data, here are the percentages of non-neutral site games that the conferences played at home:

ACC 71.7%
Big Twelve 61.9%
SEC 59.0%
Big Ten 58.9%
Big East 58.0%
Conference USA 55.8%
Pac Ten 55.2%
Mountain West 54.9%
Ivy 54.9%
West Coast 54.7%
Big West 54.3%
Patriot 52.0%
Colonial 51.5%
Atlantic Sun 51.1%
Sun Belt 50.4%
Southern 48.3%
Atlantic Ten 48.1%
WAC 48.1%
Missouri Valley 47.5%
Summit 47.4%
United (now defunct) 46.2%
Big Sky 45.8%
Ohio Valley 44.9%
Southland 44.4%
Horizon 43.5%
America East 42.3%
Mid American 42.1%
Big South 41.8%
Metro Atlantic 41.6%
Great West 40.5%
Independent 39.1%
Northeast 38.1%
Southwestern 35.3%

If I understand this right, this means North Carolina titles look a little shady and bias. Especially when it comes to location. But I didn't need this post to tell me that. Great info. Thanks!

I don't know about UNC's titles looking shady. It does say that ACC teams may be getting seeding benefits that result in their having more NCAA Tournament home games than they deserve. I haven't run new RPI ratings for the 2007-2009 seasons with home-away adjustments that will result in "corrected" ratings, but I'll do that eventually. Then, we'll get to see whether this NCRPI problem may be causing teams to be mis-seeded.

My guess is that UNC won't be the issue, but that other ACC teams' seeds will be in question.

On the other hand, over the period for which NCAA Tournament brackets are available, UNC never has played an away game -- its only non-home games have been at the "neutral" College Cup site.

Those non-home games as an example Cary, NC., are only approx. 30 miles, from what I figured. Seems like home field advantage.

Very interesting. We've made the case before that the RPI is far from perfect.

Thanks for this research. Reading abut RPI is another sign that the season is almost here.

EPL started yesterday, with my lads thwarted by a GK playing superbly.

Next weekend it's the Pilots.

UPSoccerFanatic wrote:On the other hand, over the period for which NCAA Tournament brackets are available, UNC never has played an away game -- its only non-home games have been at the "neutral" College Cup site.

I have a vague memory of a year when UNC was seeded 5th or 6th overall and had an away playoff game. In Nebraska around 1997, maybe?

I think my data go back farther than 1997. Maybe I missed a year when UNC wasn't at home.

I've now computed, for 2007, what teams' RPIs and Non-Conference RPIs would have been if they had been adjusted (in the amount I think is appropriate) for home-away imbalances. The adjustments do not result in revolutionary changes in teams' ratings, but the changes are significant enough to suggest (very strongly) that the NCAA needs to fix this problem. Based on 2007, RPIs that take into account game sites could result in one or two differences in the teams that receive at large positions in the NCAA Tournament and also could affect seeding, though not often which teams get the four #1 seeds. I'll know more after I've computed adjusted ratings for the 2008 and 2009 seasons and even more after the 2010 season is completed.

GO PILOTS!

eProf wrote:

UPSoccerFanatic wrote:On the other hand, over the period for which NCAA Tournament brackets are available, UNC never has played an away game -- its only non-home games have been at the "neutral" College Cup site.

I have a vague memory of a year when UNC was seeded 5th or 6th overall and had an away playoff game. In Nebraska around 1997, maybe?

UNC, unlike (ahem) some sites has records going back that far. They show 27 wins in 1997-1998, so that couldn't have been it. they were 17-0-1 during the regular season. they played 28 that year, which may be a record.

They also show 25 wins for 1992-1994, 1995 and 1996, so I suspect they weren't a 5-6 seed then, either.

They won it all in 1999.

They lost to ND in 1995 in Chapel Hill in the semis

They lost to Florida with Abby Wamback in the finals in Greensboro in 1998. since they slept in their own beds, I'm calling that a home venue.

And I think that other than their loss to George Mason in 1985 in Fairfax Virginia in the final, they lost no tournament games before the 1990's, regardless of venue.


So, they lost a game in the tournament on the road before 2000. -- The final against George Mason in 1985.

I just have posted RPI and Non-Conference RPI reports on the RPI for Division I Women's Soccer website, covering all games through Sunday, September 12. I also have posted conference and region (the regions I have identified as the "pools" within which teams play the predominant number of their games) average RPIs. The posts are in the form of Excel spreadsheet attachments.

Here is a link to get to the website's RPI reports page: https://sites.google.com/site/rpifordivisioniwomenssoccer/rpi-reports

For those interested in what teams' current unadjusted RPIs are from day to day, I recommend using GoCourage's NC-Soccer webpages. Here is a link to those pages: http://www.nc-soccer.com/wsoccer/2010/

The reports on the RPI for Division I Women's Soccer website are for those of you interested in the Adjusted RPI (I use my "best guess" adjustments, since the actual adjustments are secret), in the Non-Conference RPI, and in having your own week-to-week spreadsheets so that you can study how teams' RPIs evolve from week to week over the course of the season. In addition, the website contains a lot of detailed information about the RPI, about the NCAA Division I Women's Soccer Tournament, and about the process the NCAA Women's Soccer Committee goes through in selecting at large teams to participate in the Tournament and in seeding teams.

I've posted RPI (unadjusted and "best guess" adjusted), non-conference RPI, conference average RPI, and region average RPI reports covering games through Sunday, September 26 on the RPI website. They are in Excel spreadsheets at the bottom of this page: https://sites.google.com/site/rpifordivisioniwomenssoccer/rpi-reports

Just as a heads up as to how RPI is computed in different sports, The NCAA women's basketball committee just announced it will compute RPI the way Men's Basketball does.

Now a road win will count as 1.4 wins and Home wins will count as .6 wins, to reflect home court advantages and to encourage teams to play Games on the road.

There are no hidden penalties or bonuses as the RPI is computed.

http://www.ncaa.org/wps/wcm/connect/public/ncaa/resources/latest+news/2010+news+stories/september+latest+news/division+i+womens+basketball+announces+rpi+revision

Geezaldinho wrote:Just as a heads up as to how RPI is computed in different sports, The NCAA women's basketball committee just announced it will compute RPI the way Men's Basketball does.

Now a road win will count as 1.4 wins and Home wins will count as .6 wins, to reflect home court advantages and to encourage teams to play Games on the road.

There are no hidden penalties or bonuses as the RPI is computed.

http://www.ncaa.org/wps/wcm/connect/public/ncaa/resources/latest+news/2010+news+stories/september+latest+news/division+i+womens+basketball+announces+rpi+revision

Thanks, Geez. I note that the NCAA's announcement does not explain how the NCAA came up with the 0.6/1.4 ratio that men's basketball uses and that the women now will be using. I've searched but never been able to find an explanation. As a matter of interest, I believe they use the ratio only in computing Element 1 of the RPI, which is a team's winning percentage. They do not use it in computing Element 2 (opponents' average winning percentage) or Element 3 (opponents' opponents' average winning percentage). I don't see how it makes sense to use the ratio for one element and not for the others.

For women's soccer, my work demonstrates that for Division I women's soccer, a home team performs as though its rating were 0.0100 higher than as computed using the RPI formula; and an away team performs as though its rating were 0.0100 lower. After the 2010 season, I'll be proposing to the Women's Soccer Committee that they factor game locations into the RPI, in a different manner than they do for basketball, based on those numbers. The method I will propose will result in some changes to teams' ratings, although not massive ones.

UPSoccerFanatic wrote:

Geezaldinho wrote:Just as a heads up as to how RPI is computed in different sports, The NCAA women's basketball committee just announced it will compute RPI the way Men's Basketball does.

Now a road win will count as 1.4 wins and Home wins will count as .6 wins, to reflect home court advantages and to encourage teams to play Games on the road.

There are no hidden penalties or bonuses as the RPI is computed.

http://www.ncaa.org/wps/wcm/connect/public/ncaa/resources/latest+news/2010+news+stories/september+latest+news/division+i+womens+basketball+announces+rpi+revision

Thanks, Geez. I note that the NCAA's announcement does not explain how the NCAA came up with the 0.6/1.4 ratio that men's basketball uses and that the women now will be using. I've searched but never been able to find an explanation. As a matter of interest, I believe they use the ratio only in computing Element 1 of the RPI, which is a team's winning percentage. They do not use it in computing Element 2 (opponents' average winning percentage) or Element 3 (opponents' opponents' average winning percentage). I don't see how it makes sense to use the ratio for one element and not for the others.

For women's soccer, my work demonstrates that for Division I women's soccer, a home team performs as though its rating were 0.0100 higher than as computed using the RPI formula; and an away team performs as though its rating were 0.0100 lower. After the 2010 season, I'll be proposing to the Women's Soccer Committee that they factor game locations into the RPI, in a different manner than they do for basketball, based on those numbers. The method I will propose will result in some changes to teams' ratings, although not massive ones.

There was a pretty exhaustive explanation given when the Men's RPI was changed a few years ago. RPI in Men's basketball is one of the most discussed issues over the last 10 years or so because there is so much money involved.

The BCS conferences by agreement of their association, permitted BCS schools to schedule no more than 4 games per year with non BCS schools, and from a definition of what is a BCS school in the last NCAA revenues and expenses report, that is still the case.

This resulted in BCS schools being able to dictate the terms of playing venues to the midmajors. The result was that the BCS schools could set their schedules so they played no away dates in their non-conference schedule , or dictated 2 for1 or 3 for 1 home a-away terms for multiple games. As evidence that this practice is still in effect, I'll cite UP"s contract with Kentucky which specifies we will play two games there for their one game here, and an earlier 2 for 1 deal we had with Duke.

In Basketball, the home team wins 60% of the time regardless of relative strength, according to statistics compiled, so in order to encourage the BCS schools to play away, they changed their RPI to reflect this fact by weighting the home/away games by that ratio.

It didn't turn out quite as planned. It turned out that coaches were more interested in protecting their own win/loss records and the mid majors started getting higher RPI's than the BCS statisticians anticipated, and a "disproportionate" number of mid majors got in the tourney. Then they started winning those games. Money started flowing away from the BCS conferences.

What to do? Well, the selection seeding committee (60% BCS) decided that RPI would count only a small amount to selection and seeding criteria, and Polls and "the computer",) i.e. Pomeroy, Sagarin, Massey, and other systems that don't reflect the home-away bias as well) would count more. It will be interesting if the Women's tourney subsequently comes out with a policy downgrading the RPI altogether like the men have.

In Men's basketball, and to a much lesser extent with Women's BB, these decisions are worth real money, and it is no accident that conferences like the ACC and Big East get 10 times the tournament revenue the WCC gets.

I vaguely remember reading about what you've described. What I haven't seen, though, is why they picked the 1.4/.6 ratio; and why they don't include the ratio in computing Elements 2 and 3.

A lot of this is from memory, but I recall that there was more thought to inducing schools to play away games that to make a direct correlation in the RPI. The though was that you couldn't hold a team accountable for an opponent's schedule, just what the team itself did. They thought the 1.4 - .6 ratio would do it.


A more accurate apportionment, of course, would be 1.5 for away wins (60/40) and .667 (40/60) for home wins. But it is close.

But I think you are right that it doesn't accurately reflect the supposed 60-40 home wins ratio. , and that is part of the problem the BCS shools had with the plan. There was still more incentive for coaches to watch their own wins record than the RPI when they go to get that job at the next level up. That coupled with the 4 non-BCS limit for BCS teams meant that mid-majors gained a larger RPI advantage than anticipated.

None of this has to make rational sense. There is politics and money involved.

The Geez suggested elsewhere that the differences among the top conferences are smaller this year than last. I thought he might be wrong (I guess I should know better), but indeed he was correct. The Diff (Difference) number in the data below is the difference between the listed conference and the conference with the highest average ARPI for that year:

2009

1. Pac 10 .6414 (average adjusted RPI)
2. ACC .6350; Diff .0064 (Pac 10 ARPI - ACC ARPI)
3. SEC .6016; Diff .0398 (Pac 10 ARPI - SEC ARPI)
4. WCC .5999; Diff .0415
5. Big 10 .5846; Diff .0568
6. Big East .5681; Diff .0733
7. Big 12 .5629; Diff .0785

2010

1. ACC .6120
2. Pac 10 .6100; Diff .0020
3. Big 12 .6020; Diff .0100
4. Big 10 .5826; Diff .0294
5. Big East .5824; Diff .0296
6. SEC .5781; Diff .0339
7. WCC .5764; Diff .0356

What this intuitively seems to suggest, and what the Geez suggests, is that the strongest conferences will not get the RPI boost relative to the strong conferences below them, from conference play, that they got last year. It will be interesting to watch and see if this proves true.

Geez, did I get your thinking right?

UPSoccerFanatic wrote:
What this intuitively seems to suggest, and what the Geez suggests, is that the strongest conferences will not get the RPI boost relative to the strong conferences below them, from conference play, that they got last year. It will be interesting to watch and see if this proves true.

Geez, did I get your thinking right?

yeah, but iI didn't do any elaborate number crunching to see that the ACC isn't going to get the kinds of RPI's either as individual schools or as a conference that they have in the past, even though their win percentage was in the NCAA leading .750+ range. (the first week of conference it dropped 30 points)

All I did is look at the number of teams they scheduled this year that had RPI's over 150, 200, and even 300 as compared to past years. I looked at 2009 and current RPI's to see patterns. The results are kind of stunning. Remember, RPI is supposed to weed out (in?) tournament teams, and wins over teams in the bottom half or third of the NCAA seem irrelevant to that task, unless you lose to such a team.

Clemson, for instance, scheduled 7 teams over 200 in their non conference schedule of 9 teams.

NC state scheduled 3 over 200 and one near that number

Miami scheduled 6 teams over 150

Wake has a team over 250, one over 150 , and a couple in the 100 area.

Duke, who just a couple years ago had aspirations to greatness, sheduled 2 teams well over 200 and a team near that number. their schedule for what those and other teams they scheduled are doing this year are even worse.

Florida state scheduled 2 teams in the 180's and a couple over 150.

Virginia has 2 teamsover 200 and 3 teams over 150 currently on their schedule.

Maryland has 2 teams over 200 and 3 in the 150+ range.

Even UNC has played teams with current RPI's of 131,157,210, and of course they will play Clemson later (166)

It stuck me that only BC and VT were playing anything approaching a rigorous and consistently difficult non-con string of opponents. VT isn't doing all that well with it. BC has 3 100+ teams, so even their schedule isn't all that stunning, other than they played Stanford in non-con.

Not a lot of math was required.

Now, conference RPI is an aggregate of those schedules, so i didn't see how they were going to sustain a big leading conference RPI, which is essentially the second element of the RPI for the conference season.


I also couldn't think of a reason why teams that consider themselves contenders for the quarterfinals or even the tournament would create such schedules, until I thought a bit more about the RPI. It makes a bit of sense if other teams in your conference also do it, especially the bottom dwellers.

Element 1 which is the biggest component of the RPI (I think UPSF figured 50%) is your wins. so if you schedule cupcakes, the column will be good.

Element 2 is a bit more problematic. UPSF has figured out it is about 40%(?) of the RPI. If you schedule cupcakes, they will give you a bad element 2 in your RPI. But if the rest of your conference mates also schedule cupcakes, this is offset by the fact that THEIR component of your element 2 will be better than their real strength, because they won a lot of games by beating, well, the same cupcakes you beat. (Francis Marion and Tennessee-Martin must be real test pieces. everyone wants to play them)

Element 3 turns out to be worth only about 10% of your RPI, (wins of opponents of opponents
) you can take a hit there and it will be swamped by the other 2 elements.



A couple years ago, UPSF suggested a strategy for West coast teams was to schedule East coast teams because the weakness of the East isn't reflected in the RPI.

I think it may be worth investigating if scheduling cupcakes for the weaker teams in the conference isn't also a viable strategy for your conference. More teams will get in the tournament, in good years.


This line of thought was precipitated by a poster on BS bragging about ACC wins percentage, and a thought that few of those teams made the trip out West, and it got me thinking- always a dangerous thing.
I did a quick check a week or two ago of what the ACC winning percentage would be if if you discounted all wins against 200+ RPI teams. (sorry, Clemson, you get to keep your losses to cupcakes) The conference win percentage would be 120+ points lower.

There is another strategy I can see, and that is scheduling all teams in the RPI range of 30-90, with maybe 1 marquis game. Notre Dame is doing real well with that right now.


I've been wearing my tinfoil hat all week.

Geez, I think your analysis is "right on." I've been thinking all season, that a top team wants the other teams in its conference to schedule nothing but cupcakes.

UPSoccerFanatic wrote:Geez, I think your analysis is "right on." I've been thinking all season, that a top team wants the other teams in its conference to schedule nothing but cupcakes.

But what is the motivation for lesser teams to go along?

It also makes sense for the ones low in the pecking order. Clemson might not have a Win all season if they didn't schedule their 200+ teams. They even lost to one of those ( oh , dear) there was never much issue about them getting in the tournament since their ACC season is more than half their schedule and you need a .500 record. They were just looking for wins.


Miami may be in much the same boat,

and Duke's results from last year would indicate they would be on the 50% bubble again for wins. Remember that last year they only made .500 and became tournament eligible because they beat a team so bad the game was called after 70 minutes in a sort of mercy killing.
The reason given was injury to too many players, but the timing that it was exactly when the game became legal is a bit fishy. (there is no truth to the report that fans were screaming "for the love of God, end it!!")



The ones in the middle are a bit harder to figure out. At which point does it become advantageous and at which point does it not? It might be at a lower threshold than we imagine. it could be a way to maximize tournament bids.
Or it could be that this year each team thought the others in the conference would carry the conference RPI and they could cheat on their brethren a bit and ride the conference RPI without doing the share of work.

Game Theory folks would have a field day with the possibilities. (Read about the prisoner's dilemma and other games.)




Sophie wants to know if you think this is a good look for her...

UPSF

in your comparisons of conferences above, is that just non-conference for both years or does it include all games?

It strikes me that if it is the latter, the differences will narrow yet more as composite records will trend to .500 from now on for all conferences this year.

They are already closer than even I imagined. the range is smaller across 7 conferences this year than over the top 3 last year.

Geezaldinho wrote:UPSF

in your comparisons of conferences above, is that just non-conference for both years or does it include all games?

It strikes me that if it is the latter, the differences will narrow yet more as composite records will trend to .500 from now on for all conferences this year.

They are already closer than even I imagined. the range is smaller across 7 conferences this year than over the top 3 last year.

The conference numbers include all games (not just non-conference) through the last weekend of September of each year. Since the 2009 and 2010 seasons began at the same time, this is an "apples to apples" comparison.

I too noticed what you noticed. At first, I was bummed because the WCC fell from 4th at this stage in 2009 to 7th in 2010. I then noticed, however, that in 2010 the WCC at #7 actually is closer to the top than it was in 2009 at #4 and in fact is even closer than the 2009 #3 conference was.

I wish I had the 2006 and 2007 numbers, but I don't. It would be interesting to see if there's evidence of a trend towards equality among the top conferences.

A good result for the Pilots today: Wisconsin beats Minnesota 1-0. Wisconsin goes from 4-3-3 to 5-3-3. Not spectacular, but good for the Pilots. Wisconsin appears to be coming on after a slow start. Minnesota goes from 9-1-1 to 9-2-1. Although Element 3 of the RPI doesn't count for a lot (effective weight of 10% of the RPI), this opponents' opponents' record is a good addition to the Pilots' strength of schedule. This is the kind of win we really like the Pilots' opponents to achieve.

If this doesn't make sense, simply refer to the "throw the computer" emoticon.

I'm something of a numbers geek, and aside from enjoying the games, I enjoy watching how the RPIs change over on nc-soccer.

I have lots of questions that relate to Element Two, as it doesn't feel like the right representation for the goal of ranking teams.

In pursuing this, I wondered what would be the theoretical ideal schedule to play?
To determine this, I downloaded the raw game data for 2010 from nc-soccer, then implemented the RPI calculations as helpfully written by UPSoccerFanatic.
I then created some imaginary teams that play opponents based on RPI. So the first team played teams ranked 1 through 20. The next team played teams 11 through 30.

I think the results are quite interesting, such as beating all teams from 71-90 would position the imaginary team to have an RPI (0.697) very close to Portland's projected "win out the season" RPI of 0.696.
I also hadn't realized that each team's element3 is influenced by the team's own record.

Team{id=10, name='Played 1 -> 20', rpi=Rpi{elementOne=1.000, elementTwo=0.807, elementThree=0.633, total=0.812}}
Team{id=20, name='Played 11 -> 30', rpi=Rpi{elementOne=1.000, elementTwo=0.762, elementThree=0.610, total=0.784}}
Team{id=30, name='Played 21 -> 40', rpi=Rpi{elementOne=1.000, elementTwo=0.716, elementThree=0.608, total=0.760}}
Team{id=40, name='Played 31 -> 50', rpi=Rpi{elementOne=1.000, elementTwo=0.715, elementThree=0.593, total=0.756}}
Team{id=50, name='Played 41 -> 60', rpi=Rpi{elementOne=1.000, elementTwo=0.690, elementThree=0.583, total=0.741}}
Team{id=60, name='Played 51 -> 70', rpi=Rpi{elementOne=1.000, elementTwo=0.615, elementThree=0.593, total=0.706}}
Team{id=70, name='Played 61 -> 80', rpi=Rpi{elementOne=1.000, elementTwo=0.624, elementThree=0.577, total=0.706}}
Team{id=80, name='Played 71 -> 90', rpi=Rpi{elementOne=1.000, elementTwo=0.613, elementThree=0.564, total=0.697}}
Team{id=90, name='Played 81 -> 100', rpi=Rpi{elementOne=1.000, elementTwo=0.534, elementThree=0.576, total=0.661}}
Team{id=100, name='Played 91 -> 110', rpi=Rpi{elementOne=1.000, elementTwo=0.543, elementThree=0.563, total=0.662}}
Team{id=110, name='Played 101 -> 120', rpi=Rpi{elementOne=1.000, elementTwo=0.568, elementThree=0.545, total=0.670}}
Team{id=120, name='Played 111 -> 130', rpi=Rpi{elementOne=1.000, elementTwo=0.531, elementThree=0.549, total=0.653}}
Team{id=130, name='Played 121 -> 140', rpi=Rpi{elementOne=1.000, elementTwo=0.547, elementThree=0.532, total=0.656}}
Team{id=140, name='Played 131 -> 150', rpi=Rpi{elementOne=1.000, elementTwo=0.555, elementThree=0.521, total=0.658}}
Team{id=150, name='Played 141 -> 160', rpi=Rpi{elementOne=1.000, elementTwo=0.496, elementThree=0.541, total=0.633}}
Team{id=160, name='Played 151 -> 170', rpi=Rpi{elementOne=1.000, elementTwo=0.491, elementThree=0.532, total=0.629}}
Team{id=170, name='Played 161 -> 180', rpi=Rpi{elementOne=1.000, elementTwo=0.517, elementThree=0.509, total=0.636}}
Team{id=180, name='Played 171 -> 190', rpi=Rpi{elementOne=1.000, elementTwo=0.519, elementThree=0.500, total=0.635}}
Team{id=190, name='Played 181 -> 200', rpi=Rpi{elementOne=1.000, elementTwo=0.487, elementThree=0.506, total=0.620}}
Team{id=200, name='Played 191 -> 210', rpi=Rpi{elementOne=1.000, elementTwo=0.458, elementThree=0.506, total=0.605}}
Team{id=210, name='Played 201 -> 220', rpi=Rpi{elementOne=1.000, elementTwo=0.418, elementThree=0.512, total=0.587}}
Team{id=220, name='Played 211 -> 230', rpi=Rpi{elementOne=1.000, elementTwo=0.406, elementThree=0.507, total=0.580}}
Team{id=230, name='Played 221 -> 240', rpi=Rpi{elementOne=1.000, elementTwo=0.414, elementThree=0.492, total=0.580}}
Team{id=240, name='Played 231 -> 250', rpi=Rpi{elementOne=1.000, elementTwo=0.386, elementThree=0.493, total=0.566}}
Team{id=250, name='Played 241 -> 260', rpi=Rpi{elementOne=1.000, elementTwo=0.394, elementThree=0.480, total=0.567}}

Here's something pretty cool.

A day after UPSF tells us about cracking the adjusted RPI adjustments, nc-soccer has added the feature to his website with a complete breakdown and live running update.

http://www.nc-soccer.com/wsoccer/2010/index_arpi
(it's a menu item at the top of the other pages)

The way these two guys are keeping us updated is just terrific for the rest of us soccer geeks.

gmx, very interesting. But, when it comes to Women's Soccer crunch time, I don't think playing only 71-90 teams would work. This is because the RPI isn't the only consideration.

The primary factors the Women's Soccer Committee considers are the RPI (which includes a number of different items that are RPI-related), head-to-head results of teams under consideration, and results of teams under consideration against common opponents.

If the Committee can't make a decision based on the primary factors, then it considers each of two secondary factors: (1) results over the last 8 games (both record and strength of opponents) and (2) results against teams already selected for the NCAA bracket that are ranked 75 or better by the RPI (both record and strength of opponents).

From my studies, I think the Committee gets to the secondary criteria pretty regularly. Effectively, this means that the Committee is looking at how a team did in games against highly ranked teams. Last year, the Committee chair's explanation for why the Pilots did not get a #1 seed was that the four teams that got a #1 seed had better results against top 25 teams. My work indicated that this also was a factor that affected at large selections.

So, at least when looking at #1 seeds, a high RPI based on results against top 71-90 teams won't do it. There also must be excellent results against highly ranked teams.

But still, your work on how the RPI functions is very helpful.