This will be long, and for some of you tedious, so get yourself a beverage, kick back, and read on.
For those who are new to my bracket analyses or have forgotten how I do them, here's a brief explanation of my process:
The Women's Soccer Committee must use certain factors in making NCAA Tournament at large selections. The Committee no doubt also uses those factors in seeding teams, but for seeding the Committee is not required to use them and also can look at other factors. The Committee members individually decide how much weight to assign to the factors.
For each factor, and for each factor paired with each other factor, I've created a program for measuring how a team has performed over the course of the season in relation to that factor or factor pair. For factor pairs, each factor in the pair has a 50% effective weight. There are 92 factors or factor pairs.
I have the game data for the last 11 seasons, as well as the Committee's bracket decisions for those seasons. My program puts these together and, for each factor or factor pair, for each type of decision the Committee must make, comes up with a "yes" and a "no" standard. A "yes" standard means that if a team meets that standard, it always (over the last 11 years) has gotten a particular favorable decision from the Committee. For example, it has gotten a "yes, you get a #1 seed" decision. A "no" standard means that a team meeting that standard always has gotten an unfavorable decision, as in "no, you don't get a #1 seed." There are standards for #1, #2, #3, and #4 seeds and for at large selections.
Some of the factors may not have a "yes" or a "no" standard for some of the decisions the Committee must make. And, my program applies some of the factors to only the Top 60 ARPI teams. The Top 60, for practical purposes, are the only teams, since the most poorly ranked team to get an at large selection over the last 11 years was #57.
When all the games are done on the last day of the season, my program analyzes how each team performed in relation to all of the factor standards, with a focus on the Top 60 teams in the ARPI rankings. The program generates a summary of how many "yes" and "no" standards a team meets for each Committee decision -- on the seed pods and on the at large selections. It also generates a table I can use to see which "yes" and "no" standards each team met.
From these numbers, I can see what decisions the Committee will make if it follows the same pattern it's followed over the last 11 years. Or, in some cases, it will show where the Committee is seeing a team with a profile it hasn't seen over the last 11 years such that the team meets some "yes" and some "no" standards for the same decision. In those cases, the Committee must decide what to do with the team; and that decision will give insight into the Committee's thinking about the weight it assigns to the different factors.
For a more complete explanation of this process and for tables showing all of the factors and their standards, see these three webpages at the RPI for Division I Women's Soccer website: Predicting the Bracket: At Large Selections, Predicting the Bracket: Seeding, and Predicting the Bracket: Track Your Team. On the Track Your Team webpage, you also may want to look at the 2018 Website Factor Workbook 11.5.2018 that is an attachment at the bottom of that page.
The following review of the Committee's #1 seeds will put meat on the bones of the above explanation:
The Committee's four #1 seeds are Stanford, North Carolina, Florida State, and Georgetown. Here's how they and other ARPI highly ranked teams fared in relation to the #1 seed factor standards. I'm leaving out #6 Santa Clara, which was not close to being in the running for a #1 seed.
Stanford: Met 30 "yes" standards and no "no" standards. That's so one-sided I'm not going to discuss Stanford further. They were a clear #1 seed.
North Carolina: Met 3 "yes" standards and no "no" standards. This says they were a #1 seed. For illustration of how my process works, I'll go through the "yes" standards they met:
Yes Standard 24: This is a paired factor standard using a team's ARPI Rating and its Top 60 Common Opponents Rank. For each Top 60 team, my program does a complex computation that compares each Top 60 team to each other Top 60 team based on the results they had against opponents they had in common. It then assigns each team a Top 60 Common Opponents Score and a Top 60 Common Opponents Rank based on that Score. For this paired factor, the standard for a "yes" #1 seed decision is a paired factor score of >=4.2287. North Carolina's ARPI Rating was 0.6697 and its Top 60 Common Opponents Rank was #2, together producing a score of 4.2504, which resulted in a #1 seed "yes."
Yes Standard 35: This is a paired factor standard using a team's ARPI Rank and its Top 60 Common Opponents Rank. The standard for a "yes" #1 seed decision is a score of <=7.0. North Carolina's ARPI Rank was #3 and its Top 60 Common Opponents Rank was #2, together producing a score of 5.0, which resulted in a #1 seed "yes."
Yes Standard 45: This is a paired factor standard using a team's Adjusted Non-Conference RPI Rating and its Top 60 Common Opponents Rank. The standard for a "yes" #1 seed decision is a score of >=2.5762. North Carolina's ANCRPI Rating was 0.6591 and its Top 60 Common Opponents Rank was #2, together producing a score of 2.6090, which resulted in a #1 seed "yes."
It's worth noting that none of the individual factors, prior to pairing, resulted in a #1 seed "yes." It is the paired factors that produced the three "yes" results. This illustrates, I think, how Committee members most likely think: It's great that your ARPI Rank is #3, but that alone isn't good enough to get you a #1 seed. The same is true of your Top 60 Common Opponents Rank of #2. But when I look at the two of them together, that's enough to say you get a #1 seed. And, I'll add that the Committee's decision patterns show that as a team's performance on one half of this factor pair gets poorer, its performance on the other half must get better in order to maintain a #1 seed "yes." This interrelationship is true for all of the factor pairs.
So, North Carolina was a pretty easy "yes" #1 seed decision.
After Stanford and North Carolina, however, it gets much more interesting.
Florida State. Met 6 "yes" standards, but also 7 "no" standards for a #1 seed. In other words, it met some standards that say "yes" you get a #1 seed but also some that say "no" you don't get a #1 seed. Since the standards are entirely consistent with Committee decisions over the last 11 years, what this means is that Florida State presented the Committee with a profile it hasn't seen over those years.
Here are the "yes" standards Florida State met:
Yes Standard 26: This is a paired factor standard using a team's ARPI Rank and its Adjusted Non-Conference RPI. The standard for a "yes" #1 seed decision is a score of >=2.7525. Florida State's ARPI Rank was #2 and its ANCRPI was 0.7042, together producing a score of 2.7536, which resulted in a #1 seed "yes."
Yes Standard 28: This is a paired factor standard using a team's ARPI Rank and its Top 50 Results Score. The Top 50 Results Score looks at each game a team played against an opponent ranked in the ARPI Top 50. For each positive result the team has (a win or a tie counts as a positive result), the team receives an amount of points that depends on the rank of the opponent, the game site, and whether the result was a win or a tie. The points system is very heavily weighted towards positive results against highly ranked opponents. For this particular paired factor, the standard for a "yes" #1 seed decision is a score of >=63525.3. Florida State's ARPI Rank was #2 and its Top 50 Results Score was 29062, together producing a score of 64062, which resulted in a #1 seed "yes."
Yes Standard 29: This is a paired factor standard using a team's ARPI Rank and its Top 50 Results Rank. The standard for a "yes" #1 seed decision is a score of <=6.0. Florida State's ARPI Rank was #2 and its Top 50 Results Rank was #3, together producing a score of 5.0, which resulted in a #1 seed "yes."
Yes Standard 31: This is a paired factor standard using a team's ARPI Rank and its Conference ARPI. The standard for a "yes" #1 seed decision is a score of >=3.6970. Florida State's ARPI Rank was #2 and its Conference ARPI was 0.5858, together producing a score of 3.7806, which resulted in a #1 seed "yes."
Yes Standard 34: This is a paired factor standard using a team's ARPI Rank and its Top 60 Common Opponents Score. The standard for a "yes" #1 seed decision is a score of >=15.800. Florida State's ARPI Rank was #2 and its Top 60 Common Opponents Score was 5.57, together producing a score of 16.5714, which resulted in a #1 seed "yes."
Yes Standard 35: This is a paired factor standard using a team's ARPI Rank and its Top 60 Common Opponents Rank. The standard for a "yes" #1 seed decision is a score of <=7.0. Florida State's ARPI Rank was #2 and its Top 60 Common Opponents Rank was 5, together producing a score of 7.0, which resulted in a #1 seed "yes."
Here are the "no" standards Florida State met:
No Standard 7: This is a factor (not paired) using a team's Conference Standing. For Conference Standing, my system uses the average of a team's Conference Regular Season Standing and its Conference Tournament Standing. Florida State's Regular Season Standing was #7 and its Conference Tournament Standing was #4. The average of these is #4. For the Conference Standing factor, the standard for a "no" #1 seed decision is a score of >=4. Florida State's Conference Standing score was 4, which resulted in a #1 seed "no."
No Standard 13: This is a factor (not paired) using a team's poor results. The Committee does not have an overt poor results factor that it must use. It does, however, have a Last Eight Games Results factor. It is not clear what this factor is intended to cover, but my best guess is that it is looking for poor results over the last eight games. For computer programming reasons, my system looks at poor results over the entire season rather than over the last eight games, using a weighted scoring system that increases penalties for poor results as the opponent's ARPI rank gets poorer. I call this the Last Eight Games Results factor but it really is a surrogate for that factor. For the Last Eight Games Results factor, the standard for a "no" #1 seed decision is a score of <=-4. Florida State's Last Eight Games Results score was -4, which resulted in a #1 seed "no." Its -4 was for its loss to #135 Miami FL.
No Standard 49: This is a paired factor standard using a team's ANCRPI Rank and its Conference Standing. The standard for a "no" #1 seed decision is a score of >=85.0. Florida State's ANCRPI Rank was #2 and its Conference Standing was 4, together producing a score of 91.0, which resulted in a #1 seed "no."
No Standard 71: This is a paired factor standard using a team's Conference Standing and its Conference ARPI. The standard for a "no" #1 seed decision is a score of <=3.2727. Florida State's Conference Standing was 4 and its Conference ARPI was 0.5858, together producing a score of 3.2377, which resulted in a #1 seed "no."
No Standard 72: This is a paired factor standard using a team's Conference Standing and its Conference Rank. The standard for a "no" #1 seed decision is a score of >=11.9. Florida State's Conference Standing was 4 and its Conference Rank was #3, together producing a score of 14.20, which resulted in a #1 seed "no."
No Standard 76: This is a paired factor standard using a team's Conference Standing and its Last Eight Games Results (poor results). The standard for a "no" #1 seed decision is a score of <=6.5789. Florida State's Conference Standing was 4 and its Last Eight Games Results score was -4, together producing a score of 5.0000, which resulted in a #1 seed "no."
No Standard 88: This is a paired factor standard using a team's Head to Head Score and its Last Eight Games Results. For each Top 60 team, my program does a complex computation that determines the result of each game a Top 60 team has with another Top 60 team. Based on the results of all of the team's games against Top 60 opponents, it then assigns the team a Top 60 Head to Head Score and a Top 60 Head to Head Rank based on that Score. For No Standard 88, the standard for a "no" #1 seed decision is a score of <=3.0000. Florida State's Head to Head score was 0.80 and its Last Eight Games Results score was -4, together producing a score of 2.4000, which resulted in a #1 seed "no."
The Committee, in giving Florida State a #1 seed, decided that Florida State's positives, as indicated by the "yes" #1 seed standards it met, outweighed its negatives, as indicated by the "no" #1 seed standards. Looking at all of the "no" #1 seed standards Florida State met, they each consisted entirely or in part of its Conference Standing or its poor result against Miami. Effectively, the Committee said that even though a team in the #3 ARPI conference comes in #7 in its conference regular season standings and #1 in its conference tournament, for an average of #4, and even though it has a loss to the #135 ARPI team, it can get a #1 seed. Thus, at least for #1 seeds, conference regular season standing is not that important; and a quite poor result is not that important. (It does not surprise me that the Committee considered Florida State's poor result against Miami as not that important. Over the years, I do not recall seeing any evidence that the Committee singles out individual poor results as a basis for its decisions. Rather, it appears to feel that the RPI by itself sufficiently addresses poor results.)
The Committee's decision to look past Florida State's #7 regular season conference standing in the ARPI #3 conference, very possibly, could have set the table for decisions the Committee was going to have to make later on at large selections.
Georgetown. Met 1 "yes" standard and 9 "no" standards. Georgetown too presented the Committee with a profile it hasn't seen over the last 11 years.
Here is the "yes" standard Georgetown met:
Yes Standard 57: This is a paired factor standard using a team's Top 50 Results Score and its Conference Standing. The standard for a "yes" #1 seed decision is a score of >=99763. Georgetown's Top 50 Results Score was 22929 (ranked #7) and its Conference Standing was 1, together producing a score of 99929, which resulted in a #1 seed "yes." Since it is in a conference with a conference tournament, its #1 Conference Standing means it was #1 in both the conference regular season and the conference tournament.
Here are the "no" standards Georgetown met:
No Standard 20: This is a paired factor standard using a team's ARPI Rating and its Conference ARPI. The standard for a "no" #1 seed decision is a score of <=1.2285. Georgetown's ARPI Rating was 0.6691 and its Conference ARPI was 0.5437, together producing a score of 1.2128, which resulted in a #1 seed "no."
No Standard 21: This is a paired factor standard using a team's ARPI Rating and its Conference Rank. The standard for a "no" #1 seed decision is a score of <=3.8550. Georgetown's ARPI Rating was 0.6691 and its Conference Rank was 6, together producing a score of 3.8469, which resulted in a #1 seed "no."
No Standard 41: This is a paired factor standard using a team's ANCRPI Rating and its Conference ARPI. The standard for a "no" #1 seed decision is a score of <=1.6684. Georgetown's ANCRPI Rating was 0.6881 and its Conference ARPI was 0.5437, together producing a score of 1.6667, which resulted in a #1 seed "no."
No Standard 50: This is a paired factor standard using a team's ANCRPI Rank and its Conference ARPI. The standard for a "no" #1 seed decision is a score of <=3.2560. Georgetown's ANCRPI Rank was 5 and its Conference ARPI was 0.5437, together producing a score of 3.2446, which resulted in a #1 seed "no."
No Standard 79: This is a paired factor standard using a team's Conference ARPI and its Top 60 Common Opponents Score. The standard for a "no" #1 seed decision is a score of <=72.9024. Georgetown's Conference ARPI was 0.5437 and its Top 60 Common Opponents Score was 3.65, together producing a score of 69.4356, which resulted in a #1 seed "no."
No Standard 80: This is a paired factor standard using a team's Conference ARPI and its Top 60 Common Opponents Rank. The standard for a "no" #1 seed decision is a score of <=3.3206. Georgetown's Conference ARPI was 0.5437 and its Top 60 Common Opponents Rank was 10, together producing a score of 3.1446, which resulted in a #1 seed "no."
No Standard 81: This is a paired factor standard using a team's Conference ARPI and its Last Eight Games Results (poor results). The standard for a "no" #1 seed decision is a score of <=98.4001. Georgetown's Conference ARPI was 0.5437 and its Last Eight Games Results score was -1, together producing a score of 97.4965, which resulted in a #1 seed "no."
No Standard 83: This is a paired factor standard using a team's Conference Rank and its Top 60 Common Opponents Score. The standard for a "no" #1 seed decision is a score of <=8.8421. Georgetown's Conference Rank was 6 and its Top 60 Common Opponents Score was 3.65, together producing a score of 7.3167, which resulted in a #1 seed "no."
No Standard 84: This is a paired factor standard using a team's Conference Rank and its Top 60 Common Opponents Rank. The standard for a "no" #1 seed decision is a score of >=20.1. Georgetown's Conference Rank was 6 and its Top 60 Common Opponents Rank was 10, together producing a score of 22.6, which resulted in a #1 seed "no."
What stands out about Georgetown's "no" standards is that they all involved either its conference's ARPI or its conference's rank of #6. The Committee, in giving Georgetown a #1 seed, apparently decided that it was not going to "punish" Georgetown because of the conference it's in. Given that Georgetown won both the conference regular season competition and the conference tournament, one can argue that there's no reason to see Georgetown as limited in any way by its conference's rank, it may be far better than its conference affiliation indicates. This appears to be an argument the Committee as a whole bought, leaving Georgetown as a double conference champion in the #6 conference with excellent results against Top 50 opponents sufficient to justify a #1 seed.
Baylor. Met no "yes" standards and no "no" standards.
Baylor was a possible #1 seed, with no negatives. The fact that the Committee did not give it a #1 seed reinforces the likely correctness of the above explanation about Florida State and Georgetown.
Baylor. Met no "yes" standards and no "no" standards.
Baylor was a possible #1 seed, with no negatives. The fact that the Committee did not give it a #1 seed reinforces the likely correctness of the above explanation about Florida State and Georgetown.
Southern California. Met no "yes" standards and 4 "no" standards.
Here are the "no" standards Southern California met:
No Standard 2: This is a single factor standard using a team's ARPI Rank. The standard for a "no" #1 seed decision is a rank of >=8. Southern California's ARPI Rank was 8, which resulted in a #1 seed "no."
No Standard 21: This is a paired factor standard using a team's ARPI Rating and its Conference Rank. The standard for a "no" #1 seed decision is a score of <=3.8550. Southern California's ARPI Rating was 0.6485 and its Conference Rank was 4, together producing a score of 3.8167, which resulted in a #1 seed "no."
No Standard 36: This is a paired factor standard using a team's ARPI Rank and its Last Eight Games Results. The standard for a "no" #1 seed decision is a score of <=4.0000. Southern California's ARPI Rank was 8 and its Last Eight Games Results score was -1, together producing a score of 3.1250, which resulted in a #1 seed "no."
No Standard 72: This is a paired factor standard using a team's Conference Standing and its Conference Rank. The standard for a "no" #1 seed decision is a score of >=11.9. Southern California's Conference Standing was 3 and its Conference Rank was 4, together producing a score of 12.40, which resulted in a #1 seed "no."
Fundamentally, Southern California was out of the picture for a #1 seed due to its #8 ARPI Rank, which was at a level that has not received a #1 seed over the last 11 years. In addition, it was out due to its #3 position in the #4 ARPI conference.
UCLA. Met no "yes" standards and 6 "no" standards.
Here are the "no" standards UCLA met:
No Standard 17: This is a paired factor standard using a team's ARPI Rating and its Top 50 Results Score. The standard for a "no" #1 seed decision is a score of <=259875. UCLA's ARPI Rating was 0.6485 and its Top 50 Results Score was 4512 (rank #24), together producing a score of 259381, which resulted in a #1 seed "no."
No Standard 18: This is a paired factor standard using a team's ARPI Rating and its Top 50 Results Rank. The standard for a "no" #1 seed decision is a score of <=3.6994. UCLA's ARPI Rating was 0.6485 and its Top 50 Results Rank was 24, together producing a score of 3.6734, which resulted in a #1 seed "no."
No Standard 21: This is a paired factor standard using a team's ARPI Rating and its Conference Rank. The standard for a "no" #1 seed decision is a score of <=3.8550. UCLA's ARPI Rating was 0.6485 and its Conference Rank was 4, together producing a score of 3.8169, which resulted in a #1 seed "no."
No Standard 28: This is a paired factor standard using a team's ARPI Rank and its Top 50 Results Score. The standard for a "no" #1 seed decision is a score of <=14574.8. UCLA's ARPI Rank was 7 and its Top 50 Results Score was 4512, together producing a score of 14512.0, which resulted in a #1 seed "no."
No Standard 39: This is a paired factor standard using a team's ANCRPI Rating and its Top 50 Results Rank. The standard for a "no" #1 seed decision is a score of <=2.0844. UCLA's ANCRPI Rating was 0.6364 (rank #21) and its Top 50 Results Rank was 24, together producing a score of 2.0781, which resulted in a #1 seed "no."
No Standard 42: This is a paired factor standard using a team's ANCRPI Rating and its Conference Rank. The standard for a "no" #1 seed decision is a score of <=2.2408. UCLA's ANCRPI Rating was 0.6364 and its Conference Rank was 4, together producing a score of 2.2228, which resulted in a #1 seed "no."
The most obvious deficiencies for UCLA were its Top 50 Results Rank of #24 and its ANCRPI Rating that produced an ANCRPI rank of #21.
Key "Decisions." The following appear to have been key Committee decisions:
* A team's regular season poor standing position in a highly ranked conference does not get a lot of weight.
* One poor result for a team does not get a lot of weight.
* If a team wins both its conference regular season and its conference tournament, the ranking level of its conference will not get as much weight as it otherwise might get.
No comments:
Post a Comment