Changes That You Think Would Improve NCAA Tournament Seeding Process

[sagegrouse]

I question why the eye test should be a factor at all. At the end of the season, it should be just your record.

In the absence of meaningful inter-conference competition after the early-season games, the "records" appear to be highly misleading, based on NCAA-T results and the so-called eye tests. In other words, garbage in, garbage out.

The ACC has been getting screwed in recent NCAA selections. Take 2023, for example, where Duke was the media favorite to win the region, and we were upset in that troublesome game against Tenn. WE WERE A #5 SEED, and Tenn was a 4.

Anyway, the ACC has been getting half the bids and much worse seedings than the other three P4 conferences. And, despite the poorer seedings, the ACC's record is far better the least 3 years.

 

[JasonEvans]

I question why the eye test should be a factor at all. At the end of the season, it should be just your record.

Ummm, but what if -- for example -- you achieve a 30-3 record against a lousy schedule that only includes 2 games (both of which were losses) against top 30 teams and someone else finishes 28-5 against a top 10 schedule that includes 9 victories over teams ranked in the top 30. How could anyone say that the 30-3 team is better than the 28-5 team?

 

[JasonEvans]

As many others have said, "power conference matchup week" in early/mid February seems like both a money-making marketing tool and a great way to see how teams have progressed since the start of conference play.

We are just talking about a pair of games, not at all dissimilar to what Duke and Illinois are doing this February. You could even create some delicious matchups by seeding the conferences and having the matchups determined by seeding. I would not automatically put #1 versus #1, instead making it somewhat varied -- some thing like:

ACC #1 hosting SEC #4 and then plays at Big Ten #3
Big Twelve #2 hosts Big East #1 and then plays at ACC #6
SEC #3 hosts Big East #5 and then at ACC #2
Big Twelve #8 hosts Big Ten #6 and then at SEC #11
ACC #4 hosts Big Twelve #5 and plays at Big East #2
and so on....

 

[LeadingEdge]

Ummm, but what if -- for example -- you achieve a 30-3 record against a lousy schedule that only includes 2 games (both of which were losses) against top 30 teams and someone else finishes 28-5 against a top 10 schedule that includes 9 victories over teams ranked in the top 30. How could anyone say that the 30-3 team is better than the 28-5 team?

No one would say that, including me. I didn’t imply that the better record gets in or a better seed. You look at the quality of wins and losses in addition to the total of each.

I agree with requiring teams to play a couple or so nonconference games after the start of February to get a fresh gauge on relative conference strength.

But even if that does not occur, each ACC team plays approximately 10 nonconference games, which means 150 conference games during the last season. That’s a pretty large sample size. It is more than the number of games ACC teams play in the postseason. When coupled that with various advanced metrics services, the best explanation of the data is that the ACC has been a week conference that has been unusually successful in the tournament in recent times. See NCSU last season.

 

[rsvman]

Ummm, but what if -- for example -- you achieve a 30-3 record against a lousy schedule that only includes 2 games (both of which were losses) against top 30 teams and someone else finishes 28-5 against a top 10 schedule that includes 9 victories over teams ranked in the top 30. How could anyone say that the 30-3 team is better than the 28-5 team?

A decent algorithm could so be that problem. Or, as I mentioned earlier, present as many stats about the teams as you want to a committee, but have them be blinded to the team names.

 

[sagegrouse]

No one would say that, including me. I didn’t imply that the better record gets in or a better seed. You look at the quality of wins and losses in addition to the total of each.

I agree with requiring teams to play a couple or so nonconference games after the start of February to get a fresh gauge on relative conference strength.

But even if that does not occur, each ACC team plays approximately 10 nonconference games, which means 150 conference games during the last season. That’s a pretty large sample size. It is more than the number of games ACC teams play in the postseason. When coupled that with various advanced metrics services, the best explanation of the data is that the ACC has been a week conference that has been unusually successful in the tournament in recent times. See NCSU last season.

This is old territory. The 10 games at the beginning of the season is an excellent measure of team abilities in the first five WEEKS of the season. If there are no sunsequent inter-conference games, then there is no more information to rank the conferences. Most people think this is an inadequate measure for selecting teams based on capability over the full season or at the end of the season. I think it's a joke and an embarrassment.

 

[JasonEvans]

No one would say that, including me. I didn’t imply that the better record gets in or a better seed.

I mean, you wrote the following without any further elaboration -- "At the end of the season, it should be just your record."

Sorry if I took that to mean you were advocating for win percentage. I misunderstood what you meant.

 

[LeadingEdge]

This is old territory. The 10 games at the beginning of the season is an excellent measure of team abilities in the first five WEEKS of the season. If there are no sunsequent inter-conference games, then there is no more information to rank the conferences. Most people think this is an inadequate measure for selecting teams based on capability over the full season or at the end of the season. I think it's a joke and an embarrassment.

I respectfully submit that math says that the relative strength of conferences as determined by about 150 non-con games in November and December is a pretty good indicator of relative conference strength at the end of the regular season. One team might improve against the also improving field for various reasons, such as players returning from injury or a freshman-heavy team gaining experience (such as a freshman-dominated Duke team).

But the odds that a whole conference significantly improves in comparison to other conferences is nearly mathematically impossible - the odds are very long on that due the number of teams and players. There is no reason to believe that 15 ACC teams should improve at a faster rate of ascent than about that many teams in another power conference.

As I said above, I favor requiring a couple of non-cupcake non-cons after Feb 1 to get fresh inter-conference data. But, if that does not happen, weighing the number of bids a conference gets by conference strength remains as a mathematically sound decision and preferable to human subjective judgment (or bias) as to relative conference strength. If your individual team has improved much more than average, that should show up in you conference play.

 

[sagegrouse]

I respectfully submit that math says that the relative strength of conferences as determined by about 150 non-con games in November and December is a pretty good indicator of relative conference strength at the end of the regular season. One team might improve against the also improving field for various reasons, such as players returning from injury or a freshman-heavy team gaining experience (such as a freshman-dominated Duke team).

But the odds that a whole conference significantly improves in comparison to other conferences is nearly mathematically impossible - the odds are very long on that due the number of teams and players. There is no reason to believe that 15 ACC teams should improve at a faster rate of ascent than about that many teams in another power conference.

As I said above, I favor requiring a couple of non-cupcake non-cons after Feb 1 to get fresh inter-conference data. But, if that does not happen, weighing the number of bids a conference gets by conference strength remains as a mathematically sound decision and preferable to human subjective judgment (or bias) as to relative conference strength. If your individual team has improved much more than average, that should show up in you conference play.

Unless, like the Big 12, they are gaming the system like crazy. Also, higher turnover of rosters in the ACC; different coaching styles, like, ahem emphasizing conference games. Anyway, the NCAA tournament results over the past three years strongly suggest under-seeding and, almost certainly, under-representation in the NCAAT for the ACC.

Anyway, the first month of the season should not be the only way to rank, or align, conferences. And it currently is.

This has been heavily discussed here for ten years. See various threads titled "The NCAA TSC is on a Fool's Errand."

 

[FosterCoachK82]

This is old territory. The 10 games at the beginning of the season is an excellent measure of team abilities in the first five WEEKS of the season. If there are no sunsequent inter-conference games, then there is no more information to rank the conferences. Most people think this is an inadequate measure for selecting teams based on capability over the full season or at the end of the season.

I respectfully submit that math says that the relative strength of conferences as determined by about 150 non-con games in November and December is a pretty good indicator of relative conference strength at the end of the regular season. One team might improve against the also improving field for various reasons, such as players returning from injury or a freshman-heavy team gaining experience (such as a freshman-dominated Duke team).

But the odds that a whole conference significantly improves in comparison to other conferences is nearly mathematically impossible - the odds are very long on that due the number of teams and players. There is no reason to believe that 15 ACC teams should improve at a faster rate of ascent than about that many teams in another power conference.

As I said above, I favor requiring a couple of non-cupcake non-cons after Feb 1 to get fresh inter-conference data. But, if that does not happen, weighing the number of bids a conference gets by conference strength remains as a mathematically sound decision and preferable to human subjective judgment (or bias) as to relative conference strength. If your individual team has improved much more than average, that should show up in you conference play.

The odds that the ACC would overperform by as much as they have in recent tournaments is also near impossible unless the ACC teams were actually better than their seedings. Sage has posted the results somewhere in another thread and the odds of them occurring by chance are ridiculously low. The ACC has definitely been systematically underseeded.

 

[uh_no]

The odds that the ACC would overperform by as much as they have in recent tournaments is also near impossible unless the ACC teams were actually better than their seedings. Sage has posted the results somewhere in another thread and the odds of them occurring by chance are ridiculously low. The ACC has definitely been systematically underseeded.

one could say the same about the big east. it's exceptionally unlikely UConn would be blowing out the tournament like they have while struggling in games vs the big east unless the big east were not being ranked appropriately in the tournament.

then again, other top big east teams dropped the ball, as have some top acc teams. I'm the end the tournament is a bit of a crapshoot, and the acc (and big east) should do better ooc of they want more respect.

i didn't necessarily agree with the system as is, but it's no secret that it is what it is, and the acc has not executed by subjective and objective regular season metrics that people care about.

 

[Nugget]

To see what a non-Commmittee, metrics-only, field might be, I sketched out what last year’s field would have been based on an average of:

(1) Torvik’s adjusted efficiency ratings (his “predictive” metric, which includes adjustments for schedule strength, game location and, as noted above, builds into the model factors that (i) weight games played the last 2-3 months of the season more heavily and (ii) turn off, or at least reduce, individual games once they reach garbarge time to not unduly weight blowouts); and

(2) Torvik’s Wins Above Bubble (his "results" metric, comparing your record to what an average bubble team would do vs. your schedule).

The full field constructed this way is below, but the main differences with the Committee’s actual field are:

· Teams who would have made it: Indiana St., St. John’s, Pitt

· Teams who would have been left out: Florida Atlantic, Boise St., Texas A&M (the Aggies would have been way out – the 9th team out)

· Teams who would have been seeded 2 lines or more higher: Auburn (up to 2 seed from 4), Florida (up to 5 seed from 7), Nebraska (up to 6 from 8), Nevada (up to 7 from 10), Colorado (up to 7 from 1st Four), New Mexico (up to 7 from 11)

· Teams who would have been seeded 2 lines or more lower: Kentucky (5 seed down from 3), Clemson (9 seed down from 6), Miss St. (10 seed down from 8)

2024 field/seed order based on avg of Torvik efficiency ratings and Wins Above Bubble + 32 AQs

1 seeds: U.Conn-AQ, Houston, Purdue, N. Carolina
2 seeds: Iowa St.-AQ, Auburn-AQ, Marquette, Tennessee
3 seeds: Arizona, Illinois-AQ, Creighton, Duke
4 seeds: Kansas, Alabama, Baylor, Gonzaga
5 seeds: Kentucky, Wisconsin, BYU, Florida
6 seeds: St. Mary’s-AQ, Texas Tech, Nebraska, San Diego St.
7 seeds: Colorado, Nevada, New Mexico-AQ, S. Carolina
8 seeds: Dayton, Indiana St., Texas, Washington St.
9 seeds: St. John’s, Clemson, Utah St., Michigan St.
10 seeds: Northwestern, Miss. St., TCU v. Virginia, Drake-AQ
11 seeds: Colo. St. v. Pitt, Oregon-AQ, Grand Canyon-AQ, James Madison-AQ
12 seeds: NC St.-AQ, McNeese St.-AQ, Samford-AQ, Duquesne-AQ
13 seeds: UAB-AQ, Vermont-AQ, Yale-AQ, Charleston-AQ

Out: Florida Atlantic, Boise St., Oklahoma, Ohio St., Villanova, Providence, Wake Forest, Seton Hall, Texas A&M

NCAA 2024 Committee Seeds/Published Rating of the 68 teams
1 seeds: U.Conn-AQ, Houston, Purdue, N. Carolina
2 seeds: Tennessee, Arizona, Marquette, Iowa St.-AQ
3 seeds: Baylor, Creighton, Kentucky, Illinois-AQ
4 seeds: Duke, Kansas, Auburn-AQ, Alabama
5 seeds: San Diego St., Wisconsin, St. Mary’s-AQ, Gonzaga [flipped w/ BYU]
6 seeds: BYU [rated #17], Clemson, Texas Tech, S. Carolina
7 seeds: Florida, Wash. St., Texas, Dayton
8 seeds: Nebraska, Utah St., Florida Atlantic, Miss St.
9 seeds: Mich St., Texas A&M, TCU, Northwestern
10 seeds: Nevada, Boise St. v. Colo., Drake-AQ, Virginia v. Colo. St.
11 seeds: N. Mexico-AQ, Oregon-AQ, NC St.-AQ, Duquesne-AQ
12 seeds: Grand Canyon-AQ, James Madison-AQ, McNeese St.-AQ, UAB-AQ
13 seeds: Vermont-AQ, Yale-AQ, Samford-AQ, Charleston-AQ

1st 4 out: Oklahoma, Seton Hall, Indiana St., Pitt

 

[jam]

ESPN could move the ACC-SEC Challenge to early/mid February.

 

[azzefkram]

You need a better than .500 record in your conference to get an at-large bid. It should be etched in stone. I'm fairly certain SkyB has a stonecutter with some time on his hands at Casa de SkyB.