RPI for 2008

The RPI monster is off and running again. I've just run the program for 2008 games played through September 1. So far as I can figure out, PilotNation doesn't have the capacity to handle attachments, but BigSoccer does. So, for those interested in seeing the RPI after these few games, use the following link and go to my post as of approximately 11:44 am today: http://www.bigsoccer.com/forum/showthread.php?p=15622235#post15622235. A 9/1/2008 RPI report is attached to that post as a pdf file. (I'm cpthomas on that blog. Not very discrete, I guess.)

The 9/1/2008 report is not meaningful for purposes of teams' ranks or for how the RPI will rank teams as of the end of the season. However, I'll be posting these weekly (I hope), and if you follow them weekly, you'll get an idea of how the RPI does and does not work. Plus, I'll be making "educational" comments with each posting that you may find of interest. I'll post a message plus a link on this thread each week as I post the reports on the BigSoccer blog.

I saw it on the BS forum. Nice job. I'll be following his as you go forward.

I just entered an RPI report for games through September 7 as a pdf attachment to an RPI thread on the BigSoccer forum. If you want it, use the following link: http://www.bigsoccer.com/forum/showthread.php?t=650592. The report is attached to the cpthomas post at 12:07 pm today. I've also included in the text of the post a discussion of how the NCAA's data gathering process works and of how errors can creep into it.

Nice job again.

It's fun to watch how the RPI is working. It shows how the RPI works much better than just getting a pile of data at the end.

It's interesting to see how the teams are starting to sort themselves out. I am a little puzzled about how elements 2 and 3 seem to be sorting themselves so early. I wouldn't have though that there wouldn't be enough data to link up enough schools.

It's pretty interesting that it works at all...

I really should spend some time with the details and understand a little better how that works, All I know is I like seeing UP near the top Should have a few more Quality opponents along the way to help our RPI, with this schedule winning the ones that count should put us right there at the top when it matters.

I just noticed this thread. I will look into the attachments and see if this software can allow attachments. . .

-SciFi

I really don't know why ; but, unfortunately it appears that attachments are not an option with this software.

-SciFi

Scifi, I know that my old employer used PHPBB and we had file uploads available, so it is possible, how worthwhile? well that is another matter all together.

Hmmm. . . I'll keep digging then.

-SciFi

I've posted on BigSoccer, as an attachment, the RPI report covering games through 9/14/2008. If you want to see it, use this link: http://www.bigsoccer.com/forum/showthread.php?p=15748743#post15748743. It is an attachment to the cpthomas post dated 9/15 at 10:52 pm.

UPSF

Can't open the file on Big Soccer. Where is UP on your latest report?

You probably can't access the report on BigSoccer because you need to set up an account with a user name and password. I think you can do that for free, so if you want you might try it. The three reports I've done have us ranked as follows:

Through 9/1: 22

Through 9/7: 7

Through 9/14: 15

How could we win both games last weekend and see our RPI position drop? Has to do with the records of Northern Arizona and San Diego State being added into the calculation of our RPI, plus with how the other teams we've played did last weekend. Essentially, the strength of schedule element of our RPI dropped significantly (primarily our opponents' winning percentages against other teams, more-so than our opponents' opponents winning percentages). The drop in the average of our opponents' winning percentages (against teams other than us) outweighed the improvement in our own winning percentage.

The positions of teams will continue to jump around a lot for the next couple of weeks, depending on how teams do and what their opponents' records are. As we get toward the end of the season, the changes will be smaller since each team's RPI will be based on a greater number of games.

As a matter of interest and how the RPI works, although USC won both games last weekend, it did nothing for our RPI. In calculating USC's contribution to our RPI, the formula excludes their game against us. So, for our RPI, USC's winning percentage is 100% (they've beaten everyone else). The week before, when they were 4-0 for our RPI's purposes, their percentage was 100%; and after adding this week's two wins, they still were at 100%. The opponents' winning percentage element of the RPI does not simply add up all opponents' wins, losses, and ties and then compute a winning percentage. Instead, it calculates each opponent's winning percentage and then takes the average of all opponents' winning percentages. Since USC contributed 100% to each calculation, the fact that the second week's 100% represented two more wins is irrelevant.

Just for fun, I took a look at inter-regional game results by region, for games played so far this year. Here are the results. The numbers, in order, represent wins, losses, ties, total inter-regional games, # of teams in region, and # of inter-regional games per team.


Central 63 58 12 133 57 2.33
Great Lakes 74 82 8 164 62 2.65
Mid Atlantic 55 68 19 142 48 2.96
Northeast 47 54 12 113 41 2.76
Southeast 81 73 17 171 61 2.80
West ****56 41 14 111 51 2.18

Sorry I can't do better with the table format, but I can't figure out how to make the table function work.

Since there's no strength of schedule element to these computations, they're of limited worth. What's most interesting to me is the # of inter-regional games per team. The Mid-Atlantic region has the most inter-regional games per team, which makes sense because the Mid-Atlantic region is surrounded by three other regions: Northeast, Southeast, and Great Lakes. The West region has the least, being bordered only by the Central region. Plus, the Central and West regions, being by far the largest, have the greatest travel distances to other regions and therefore would be expected to have less inter-regional games per team. This is one of the reasons it is difficult for computer models to compare West region teams to the teams from other regions.

Still, based on games so far, the West region clearly is the strongest region, right?

Beautiful!

Nice work. It was worth trying to get into a table, but the one on the forum is pretty cumbersome.

The [tr] operator makes rows
The [td] operator makes data cells
the [table border="x"] operator will make a border "x" pixels wide

It's probably easier to take a screen shot of an Excel section. as is on BS

The new RPI report covering games through 9/21/2008 now is posted on BigSoccer. Use the following link: http://www.bigsoccer.com/forum/showthread.php?t=650592. The report is a pdf attachment to the 9/22/2008 3:21 pm post of cpthomas.

This week, in the RPI computations, the Pilots added the win over Florida to their winning percentage. In the strength of schedule part of the computations, the Pilots added Florida's 5-1-1 record, which included Florida's Sunday win over Kansas but excluded Florida's Friday loss to the Pilots (because that game doesn't count in computing Florida's contribution to the Pilots' strength of schedule). Also in the strength of schedule part of the computations, the Pilots added its earlier opponents' weekend results, which were a hearty 9-0-1. The result was an unusually big increase in the Pilots' computed strength of schedule.

So, although we're only just short of half way through the season and things no doubt will change a lot, guess who's #1 in this week's RPI ratings?

GO PILOTS!

Is that cool or what?

Since I don't have a subscription to Big Soccer, I assume you mean the Portland Pilots are number one in the RPI this week? Which would be very good.

FANAtic it looks like it. It seems the teams we played and beat have been doing very well which helps our RPI not to mention that we are winning too.

FANatic, I don't think you need a subscription. I think if you register with a name and password, you will be able to use the link and go to the website and download the RPI Report attachment. (You do have to subscribe to be able to post attachments, I think.) In any event, here are the top 25:

1. Portland
2. Texas A&M
3. Stanford
4. Notre Dame
5. UCLA
6. North Carolina
7. Boston College
8. Florida
9. Texas
10. Illinois
11. Colorado
12. Duke
13. Central Florida
14. California
15. Florida State
16. Rutgers
17. Old Dominion
18. James Madison
19. USC
20. Oklahoma Stte
21. South Carolina
22. Brown
23. Wake Forest
24. Penn State
25. Charleston

Interesting that Rutgers has higher RPI than Penn State. Hope that Pilots can add wins over two (RPI) top 25 teams this weekend.

So Rutgers is #16 and Penn St #24. (Wrote this before I saw Martin's post ahead of mine).

I assume if we beat these two, it is very, very good for our RPI, what with both teams having good records against what I would guess are very good opponents.

Still havin' a hard time looking at that Top 25 with Portland on top. It just doesn't happen very often. Maybe I'll stare at it some more tomorrow!

You don't have to subscribe to RS to post or use the forums. You just have to register. That's free.

In addition to the weekly RPI computations, I've run a computation that deletes all intra-regional games and computes only the average team RPI by region, based strictly on inter-regional games. This has a number of limitations, as some teams (24) have played no inter-regional games and there are not enough inter-regional games to allow the calculation of all teams' RPIs based only in inter-regional games. With those limitations in mind, here are the regions' average team RPIs, using the inter-regional RPIs for those 148 teams for which it is possible to compute them with the data generated so far this year:

Central .4982
Great Lakes .5049
Mid Atlantic .4923
Northeast .4797
Southeast .4940
West .5554

These numbers are pretty consistent with the number from last year; and are supported by Albyn Jones' SoccerRatings and the Massey ratings from last year.

As you can see, these data indicate that the West region is significantly stronger than any of the other regions. This creates a problem, because the RPI, in comparing teams, is not able to take regional strength differences into account and therefore discriminates against teams from stronger regions. It does this because it is weighted in favor of assuming that all the regions are equal in strength; and there are not enough inter-regional games to overcome the effect of this weighting. Part of my "fun" is waging a long term battle to cajole, embarrass, force the NCAA to recognize this defect in the RPI and at least take regional strength differences into account in making "hard" decisions during the Tournament at large selection and seeding processes