Well I guess I should have been clearer - what I heard is that Dayton chose not to submit a bid. I was under the impression it was not a mistake. Now this is like telephone where someone told someone who told me a rather uninterested party who may not have asked the right questions. They did just host the A-10 tournament so maybe that busted their budget (and this is pure conjecture on my part.)
A cautionary note: DO NOT rely on the NCAA's RPI Report issued on the NCAA.com website this afternoon. It contains errors. How they occurred is under review. I understand that this report is NOT consistent with what was provided to the Women's Soccer Committee for purposes of their NCAA Tournament bracket formation process.
Forgive my ignorance but what happens if the seeded team loses in the first round? Will the game (s) still be played at that field or will it move to the next highest seed field?
In theory, the sites for the second/third round games have not been set yet. For practical purposes, it's pretty clear they will be played at the site of the highest seeded team, assuming it has bid for the games. If the top seed in the group has lost in the first round, then the games will go to the site of the other seed in the group. If both seeds have lost, then the games in all likelihood will go to the site of the team in the group with the best RPI. That is consistent with past practice and with how the NCAA sited the games between non-seeds this year.
Using the Massey ratings (converted to the old Albyn Jones rating scale) here's the expected win probabilities for the 1st round. I'm using the figure of 60 rating points for homefield advantage. favored team / rating / (H) = home / rating diff / expected win prob. / underdog / rating rating difference adjusted for homefield advantage Code: Louisville 1679 (H) 19 0.526 Dayton... 1720 Kentucky 1683 (H) 26 0.543 Washington St 1717 Kansas... 1671 (H) 59 0.594 Georgia... 1672 UC Irvine 1757 (H) 74 0.619 San Diego 1743 West Virginia 1774 (H) 93 0.651 Virginia Tech 1741 Miami...... 1693 (H) 128 0.704 Alabama... 1625 Tennessee 1738 (H) 129 0.704 Ohio St... 1669 South Carolina 1750 (H) 138 0.718 Texas...... 1672 Pepperdine 1854 (H) 161 0.752 Long Beach St 1753 Santa Clara 1813 (H) 162 0.752 California 1711 Oregon St 1755 (H) 165 0.758 Portland... 1650 Auburn... 1732 (H) 182 0.777 Utah St... 1610 Illinois... 1834 (H) 198 0.794 Notre Dame 1696 Maryland 1764 (H) 199 0.794 LaSalle... 1625 Texas A&M 1834 (H) 201 0.8 LSU...... 1693 Boston U 1691 (H) 229 0.826 Harvard... 1522 North Carolina 1859 (H) 238 0.836 William & Mary 1681 Milwaukee 1789 (H) 242 0.841 Illinois St 1607 UCF...... 1661 (H) 284 0.874 FIU...... 1437 Florida... 1761 (H) 288 0.878 Fl Gulf Coast 1533 Marquette 1810 (H) 307 0.892 Toledo... 1563 UCLA...... 1914 (H) 315 0.899 New Mexico 1659 Baylor... 1763 (H) 398 0.939 Texas St... 1425 Florida St 1883 (H) 416 0.947 Samford... 1527 Boston College 1778 (H) 420 0.948 Marist... 1418 Penn St... 1894 (H) 480 0.965 Army...... 1474 Virginia... 1877 (H) 498 0.969 Long Island 1439 Memphis... 1880 (H) 520 0.974 UT Martin 1420 Wake Forest 1901 (H) 578 0.981 Oakland... 1383 Duke...... 1947 (H) 630 0.985 Radford... 1377 Oklahoma St 1900 (H) 846 0.996 Ark-Pine Bluff 1114 Stanford 2095 (H) 899 0.996 Montana... 1256
Has a corresponding 2 and 3 seed (or 1 and 4) ever both lost in the first round before? Would the contingency plan be apparent or would the NCAA have to go through a bidding process for hosts all over again?
I have received two fantastic documents related to the NCAA Tournament bracket process that are intended to be available for public viewing. They are NCAA-generated decison-criteria-related documents that I understand contain information provided to the Women's Soccer Committee during its bracket formation process. I don't know whether these were the only information provided, but one of the documents appears extremely complete. I haven't had a chance to fully analyze the documents yet, but I'm certain that getting a full understanding of them will give great insights into the Committee's decision-making process. The pieces are in pdf files. Unfortunately, the most important piece is in too large a file to be accommodated by BigSoccer's attachment function. So, if anyone wants these, send me a PM with an email address and I will email them to you. For those who really, really want to know how the process works, the more important piece, in particular, is a "must see" document. The less significant piece (to me) is called "Nitty Gritty" and contains a breakdown of basic information about each team's record (all 322 of the teams). The really significant piece is called "Team Sheets." This is an absolute gold mine for those really interested in understanding the process. There is a one-page sheet for each team (all 322 of them) chock full of information, including information that can be used in evaluating the applicable criteria. With the release of this information and, I believe, more information to come, it is looking very much like the NCAA is opening up its process greatly.
Thanks to everybody posting all the team rankings. A question: Is there an explicit formula for an expected win percentage, based on the teams rankings? I.e. kolabear listed the following table, but how would I determine the expected win percent for, say, 176 pt differential. kolabear listed some specific numbers in a post just above, so it seems there is some formula! 100 pt differential: .667 expected win pct (2 to 1 win ratio) 200 pt differential: .800 expected win pct (4 to 1 win ratio) 300 pt differential: .889 expected win pct (8 to 1 win ratio) 400 pt differential: .941 expected win pct (16 to 1 win ratio)
Kolabear is trying to mimic the Albyn Jones rating system, now defunct. Without knowledge as to the conversion methods or formulas from Massey, it's a private system. But the original jones system Was straightforward log function base 2 with the ranking difference/100 representing the exponent of the odds:1 With a scientific calculator it's easy to figure if it has a function. Just use y=2 and x = (ranking difference)/100. The win percentage will be Result/(Result + 1) So a 400 pt differential would be x=(400/100) = 4 2(EXP)4 = 16:1 win odds 16/17 = .941. Win percentage Your example of 176 would be 2(exp) 1.76 = 3.387:1 odds 3.387/4.387 = .772 win percentage It didn't figure ties, however, and a different function figured those probabilities, but i never quite figured it out and I haven't found anybody who has figured that out and published it. I lost interest in figuring it out when Jones stopped publishing current data.
Ummm, I'm trying to remember... Give me a few minutes! **** if R is the rating differential, you start by dividing it by 100 to get n. Then the expected win pct = (2 to the power n) / ((2 to the power n) + 1) So if R (the rating differential) is 100, you divide it by 100 to get n=1 Then the expected win pct = (2 to the power 1) / ((2 to the power 1) + 1 = 2 / (2+1) = 2/3 = .667 *** yeah, yeah, cliveworshipper has it, I think, and he knows how to paste the exponents properly as a superscript! Why am I such an old-cow Luddite on these things?! And okay I said I would re-copy this chart of conversions: 0 0.5 10 0.517 20 0.535 30 0.552 40 0.569 50 0.586 60 0.602 70 0.619 80 0.635 90 0.651 100 0.667 110 0.682 120 0.697 130 0.711 140 0.725 150 0.739 160 0.752 170 0.765 180 0.777 190 0.789 200 0.8 210 0.811 220 0.821 230 0.831 240 0.841 250 0.85 260 0.858 270 0.867 280 0.874 290 0.882 300 0.889 310 0.896 320 0.902 330 0.908 340 0.913 350 0.919 360 0.924 370 0.929 380 0.933 390 0.937 400 0.941 410 0.945 420 0.948 430 0.952 440 0.955 450 0.958 460 0.96 470 0.963 480 0.965 490 0.968 500 0.97 550 0.978 600 0.985 650 0.989 700 0.992 750 0.995 800 0.996 Some of you will notice that I did some rounding off in the expected win percentages for the 1st round games.
Can you imagine the NCAA trying to explain its new rating system in the terms Cliveworshipper and Kolabear are using? And we'll all say, "Oh, yeah, that sounds like the right system to me."
Well, they have done such a fine job explaining the significance of the records of opponents of opponents, why bonuses are secret, and why home field has no bearing on RPI
Actually, the significance of records of opponents of opponents isn't that difficult (and, I believe Massey uses it too, except that he carries it out to a theoretical infinite number of extensions). Of course, the bonuses and penalties being secret is indefensible, but at least we know there are bonuses and penalties and that we don't know what they are -- well, actually we do know what they are, but no thanks to the NCAA (stay tuned for more information on this topic, coming out shortly). And, we know that home field advantage should have a bearing and that with the RPI it doesn't. In other words, with the RPI at least we know what we don't know. But with Massey and Jones, what are in their systems that we don't know? We don't know. But maybe ignorance is bliss?
Thanks! That formula is exactly what I was looking for. (BTW, I don't think you can write superscripts/subscripts in BBCode. You can in HTML, but it is a pain.) Resaying this: Odds = 2^(difference in ratings / 100) I.e., 'odds' are Odds:1 P(win) = Odds / (Odds + 1) This part doesn't seem arcane to me (I like math though), it is how you come up with the ratings, that is the arcane part!
I agree, partly. RPI is less of a "black box" than Massey's ratings, Pablo, and Elo -based systems. * But it's also because some of us really want to understand the nitty-gritty of how it all works. Driving a car doesn't require understanding the physics of internal combustion. And chess ratings (or its like) have been popularly used and accepted for decades, not only by chessplayers but, I'm led to understand, by many people playing all kinds of different board games and online games. The NCAA and BCS even use similar methods for basketball and football. As Pablo's creator has said (Pablo is a rating system for volleyball), the NCAA uses other ranking systems (non-RPI ones) in sports where a lot of people really care about the tournament. Yes. I agree it's not really the arcane part at all. I'll be talking a bit more about these things in the Massey/ Elo thread from time to time during the playoffs. * Speaking of getting into the nitty-gritty of designing a rating system, the creator of Pablo shared this quote: "What part of an inverse tangent function approaching an asymptote did you not understand?"
There's always that...(!) Just as an aside, someone here mentioned something about one of Illinois State's (not Illinois) top goal scorers. Massey's ratings ranks Illinois State as the 10th best offense in Division I. Wow.
Yeah, they've scored 51 goals as a team, and Rachel Tejada has 20 of them herself. That's an interesting matchup they have with Milwaukee and Sarah Hagan. The Redbirds beat the Panthers 3-1 a couple of weeks ago. I picked Milwaukee to win the rematch at home, but it should be a great game.
Here's a little bit of information that shows how mistakes can be made that affect the RPI: In the Atlantic Ten tournament semi-finals, Dayton beat Richmond. Dayton was the tournament host. Richmond, in the RPI rankings, ended up as a Top 41-80 team. Because Richmond was in that group and Dayton beat them, Dayton had bonus points added to its Unadjusted RPI as part of converting to the Adjusted RPI. The way the bonus points are set up, for a win against a Top 41-80 team, you get the least points if the game is at home, you get an additional 0.0002 points if the game is at a neutral site, and you get a further 0.0002 points if the game is away. So, the difference for the bonus between a home win and an away win is 0.0004 points. When the game result got entered into the NCAA's system (I don't know who entered it although ordinarily the home team is supposed to do that), rather than getting correctly entered as a home game for Dayton, it got entered incorrectly as being at Richmond. No one caught this, so the rating for Dayton provided to the Women's Soccer Committee for the bracket formation decisions over-rated Dayton by 0.0004. 0.0004 is not a huge amount, from an RPI perspective, and in this case it did not change Dayton's ranking position, only their rating. On the other hand, in a different situation that amount could have affected their ranking position. Since Dayton was a potential seed candidate (although near the outside edge although better-rated than Tennessee), this could have had some relevance to the seeding discussion if it in fact had changed Dayton's ranking position. A similar problem, if it occurred in the area of rankings where the "bubble" teams reside, would be more likely to change rankings since the teams' ratings are more compressed there. It could mean the difference between a team being a "bubble" team and the team not getting considered at all. This provides an interesting little illustration of how important it is that correct data get into the system.
Sorry to plug a blog while on a blog, but since we are all fans.....check out the work that Chris Henderson has done at All White Kit. Detailed breakdowns of each of the first round matches. Amazing.