Final Four Thoughts

Without further ado, here are a few net efficiency margin-related thoughts on the Final Four teams. Much like the preferred "slash line" in baseball (batting average/on-base percentage/slugging percentage), NEM here will be shown as offensive/defensive/total NEM. (OK, so it's not exactly like the slash line, since you can derive the third from the other two, but work with me.) Remember, larger offensive NEM (more points scored) and smaller defensive NEM (fewer points allowed) is better.

Syracuse D

Through the first four games of the tournament, the vaunted Syracuse zone has vanquished opponents new and old. Their tournament offense has been at roughly their regular season level, but the defense has been Orangetastic.

Regular season and Big East Tournament (35 games): 12.1/-12.8/24.9

NCAA tournament (4 games): 9.8/-38.0/47.8

That -38 defensive NEM means they're allowing 0.38 points per possession fewer than an average Division 1 team would allow against the same opponents. For a typical 66-possession game (the average number of possessions per D1 game), that translates to 25 points.

Michigan Man

Mitch McGary started a few games in February, but was inserted into the Wolverines' starting lineup for good at the start of the NCAA tournament. In four games, McGary has averaged 30 minutes per game, pulled down 8 defensive rebounds per game, and scored a career-high 25 points against Kansas.

He's been described an "X-factor" and "nitty-gritty guy" whose impact doesn't always show up in the box score, but can we measure how the team as a whole has improved?

Remember it's only a four-game sample, so we have to be careful drawing too many conclusions.

Regular season and B1G tournament (32 games): 19.1/-8.0/27.2

NCAA tournament (4 games): 24.8/-20.7/45.5.

The offense is playing about 5 points per 100 possessions better, but the defense is much improved. For comparison, prohibitive favorite Louisville is sporting a 27.4/-19.6/47.0 NEM line for their four tournament games. If Michigan's improved play isn't a fluke, they could give the Cardinals something to worry about if both make it to Monday night.

KenPom recently noted that McGary is putting up some DeJuanian rebounding numbers, but he's also been producing points. McGary has made 33 of his 45 field goal attempts in the tournament (all inside the arc). He hasn't gotten to the line much (10 attempts in 4 games) or converted once there (he made 4 of those 10), so at least he has something to work on.

Louisville (and Michigan) prep?

Much was made before Sunday's Midwest Regional final about Coach K's 11-1 record in Elite 8 games, as if there was any predictive value for quick turnarounds in a 12-game sample that you couldn't already get by knowing he is a Hall of Fame coach with the most all-time D1 wins. (I digress.)

However, the numbers (again, an admittedly small sample) through the first two weeks show that Louisville and Michigan have each excelled in the second game of each pair. It could be superior preparation for games on short rest, or it could (more likely) be statistical noise. In any event, it's an interesting nugget and something to watch should both reach Monday.

First game of each week

Louisville: 22.0/-12.6/34.6

Michigan: 22.0/-8.0/30.0

Second game of each week

Louisville 32.9/-26.6/59.5

Michigan 27.6/-33.5/61.1

Wichita State

The Shockers beat the 1 and 2 seeds in their region on the way to Atlanta. Though none of the other three teams managed that, they didn't face the most difficult path down Peachtree Road.

Seeds defeated on the road to Atlanta:

  • Michigan 13 + 5 + 1 + 3 = 22
  • Wichita State 8 + 1 + 13 + 2 = 24
  • Syracuse 13 + 12 + 1 + 3 = 29
  • Louisville 16 + 8 + 12 + 2 = 38

They have also been the most consistent of the four teams remaining in the tournament, but is that enough?

Average and standard deviation NEM during 2013 NCAA tournament:

TeamAverageStandard
Deviation
Wichita State40.77.3
Michigan45.515.6
Louisville47.012.9
Syracuse47.817.3

A smaller standard deviation indicates more consistent performance, which is usually good. The problem for Wichita State is that with their lower average, it also implies a lower ceiling if they remain consistent.

What are the odds?

Nate Silver, fresh off his 50 for 50 performance in forecasting the 2012 presidential election, is in his third year of publishing NCAA tournament probabilities. Likewise, we're in our third year of using those to calculate the probabilities of interesting results starting with the Sweet 16.

47.8% probability of a #1 seed winning the championship

19 champs have been #1 seeds - most since seeding began in 1979

6.7% probability of three #1 seeds in the Final Four

Three or more #1 seeds have made it four times since seeding began

14.5% probability of no #1 seeds in the Final Four

A #1-free Final Four has happened three times since seeding began (last in 2011)

19.9% probability of a Louisville vs. Florida final

The most likely championship game pairing

15.2% probability of #2 seed winning the championship

Six champs have been #2 seeds - second most since seeding began

25.3% probability of #3 seed winning the championship

The 2011 champion (UConn) was a #3 seed. Florida (21.3%) accounts for most of this chance.

6.8% probability of a #12 seed or lower in the Final Four

The lowest team to make a Final Four is an #11 seed (George Mason 2006, VCU 2011).

3.6% probability of a first time national champion

Miami, Wichita State, and Florida Gulf Coast are the only schools left wihtout a championship. The other 13 schools have a combined 25 championships.

7.1% probability of all-B1G Final Four championship game matchup (Corrected)

One Big Ten team has won the national championship in the last 23 seasons (2000 Michigan State)

3.6% probability of both remaining ACC teams making the Final Four

More likely, there is an 65.7% chance neither will

0.4% probability of all-B1G Final Four

The 1985 Final Four saw three teams from the Big East.

5.7% probability of a rematch of the 2007 Florida/Ohio State championship game

100% probability of a Sweet 16 rematch of the 1987 Indiana/Syracuse championship game (just making sure you're paying attention)

Final thoughts

The final update is live. Here are some final thoughts:

  • It looks like Middle Tennessee is going to squeak by Virginia in the final selection model. The final scores were 25.6 (out of 100) for MTSU and 22.7 for Virginia. I wouldn't be surprised with either. Virginia's profile is so wacky, there is really no good comparison for them. They have some good wins, but some really, really bad losses.
  • Kentucky was the other selection wildcard, with their post-Noel performance falling off quite a bit. Despite the selection model's confidence down to the endm, I don't see them getting in.
  • The CTD seeding model bumped Indiana down to a 2, but I'm not sure I see that happening. I'm not sure whether it would be Duke or Gonzaga that would be bumped down to the 2 line, but I feel safe with any four out of Louisville (overall #1), Kansas, Duke, Gonzaga, or Indiana. Despite the protests of some, I don't see Miami rising to the top line.
  • The seeding model might also have New Mexico (2) and St. Louis (3) a bit high, but it's been fooled in the past by some non-power conference resumes. I'd say the next line down for each is probably the right guess.

For posterity, I'll go with the 1 seeds:

Louisville - Indianapolis

Kansas - Dallas

Indiana - Washington, D.C.

Duke - Los Angeles

That would likely then put Gonzaga in L.A. as the 2 seed, so it may not matter all that much whether Duke or Gonzaga gets the formality of the 1 seed.

Let's roll the dice and start playing some games.

BubbleGrid for March 16

See this to learn what the BubbleGrid all about.

Much of the remaining bubble debate has been over Middle Tennessee State, which has a gaudy 28-5 record but very little meat on the resume. Their power numbers are decent though not spectacular (31 Pomeroy, 53 Sagarin, 13 LMRC), suggesting they have at least taken care of business against the vast majority of their schedule that is outside the RPI top 100.

The Crashing the Dance selection model is built using the selection committee's picks from the past 13 seasons. It assigns a score from 0 to 100 to each team - 0 means that team's profile doesn't look like that of an at-large team and 100 means that it does. (A score in between indicates a mix of good and bad attributes.) The teams are then ranked by their score to find the 37 at-large teams.

Keep in mind that no bubble team exists in a vacuum; you have to compare their profile against the other teams fighting for at-large bids. That said, looking to the past helps us get an idea of how the committee generally looks at teams with this type of profile.

The two things that stand out to me in their profile are a lack of RPI top 50 wins (though their best win is just outside that arbitrary threshold) and 12 of their 28 wins coming outside the RPI top 200.

I could find only 5 similar teams in the 13 years of data used to build the CTD selection model that were either at-large (AL) selections or were automatic qualifiers (AQ) but would have been selected at-large (based on their seed):

SeasonTeamTypeSeedW-Lvs. t50vs. 201+
2003UNC WilmingtonAQ1124-60-213-0
2008DavidsonAQ1025-60-312-0
2009VCUAQ1124-90-111-3
2012VCUAQ1228-60-111-0
2012IonaAL1425-70-212-2

There are only two teams with at least 28 wins and no top 50 wins - the above 2012 VCU team and 2011 Belmont. Both won their conference tournament, so an at-large comparison isn't exact, but VCU was just ahead of the last at-large team (Iona) on the seed list. Belmont was a 13 seed, which was behind the final at-large teams (all of the First Four at-large games were 12 seeds). We can reasonably conclude that neither would have been selected if they didn't win their conference tournament.

Last year's Iona may be a good comparison. Here are the nitty gritty details of that team and this year's MTSU:

 RankConfN/CSparkline1-251-501-1001-200101+201+
SchoolSeasonScoreW-LSag.PollRPIRPI+/-RPIWW+/-W+/-+/- 
Iona20120.025-7534078121800-252820-412-2
MTSU20130.828-553313042181600-31-21127-212-0

The profiles are remarkably similar. The CTD selection model didn't think either profiles were at-large quality; Iona's selection score assigned by the model was 0 (out of 100). Middle Tennessee's conference performance is slightly better, but the biggest difference is Iona's 5-3 record against the RPI top 100 compared to 1-3 for Middle Tennessee.

Middle Tennessee does have a home win against fellow bubble dweller Ole Miss, though I tend to put less emphasis on one game out of more than 30 than others do. I don't have a strong opinion on the high-major vs. mid-major vs. low-major debate. The differences between teams around the cut line are by nature slight, and good arguments can be made for and against all of them.

  • Hover over a value for more information
  • Shaded teams are predicted to be in the field
  • Teams with a * are the projected automatic qualifier (AQ) for their conference. Bold shaded teams have clinched the AQ berth
  • Teams are listed in order of their selection confidence by the CTD model. In other words, the last shaded non-AQ team is the last team in, and the first non-shaded team is the first team out.

Data is through games of Friday, March 15.

1-251-501-1001-200RoadGood road winsBad home losses
TeamScoreW-LWW+/-W+/-+/-+/-
Temple> 99.923-924110498-4#51 Villanova
#56 Massachusetts
#70 Charlotte
#115 Canisius
#137 St. Bonaventure
#229 Duquesne
Illinois> 99.921-1246-37-345-6#6 Gonzaga
#33 Minnesota
#168 Northwestern
Oregon> 99.925-833081116-5#21 Nevada Las Vegas
#23 UCLA
#90 Washington
Colorado> 99.921-112409085-7#48 Oregon
#78 Stanford
#180 Oregon St.
Wichita St.> 99.926-803183117-5#27 VCU
#73 Air Force
#74 Indiana St.
#106 Evansville
Iowa St.> 99.922-1124-48-144-9#68 Baylor
Cincinnati> 99.922-1115-49-245-5#43 Pittsburgh
Belmont *> 99.924-601-16498-5#72 Eastern Kentucky
#78 Stanford
Oklahoma> 99.920-1123-49055-7#68 Baylor
Villanova> 99.919-1345-38-327-7#46 Connecticut#274 Columbia
Boise St.> 99.919-1024-38055-8#26 Creighton
#76 Wyoming
St. Mary's> 99.926-601-240119-2#69 Brigham Young
California98.920-112506-456-5#14 Arizona
#48 Oregon
#66 Denver
Mississippi90.824-802-18276-6#59 Tennessee
#84 Louisiana St.
La Salle85.621-912-36-278-6#27 VCU#199 Central Conn.
Kentucky78.321-1112-27-214-8#52 Mississippi
#97 Texas A&M
Virginia73.221-112418453-8#34 Wisconsin
#63 Maryland
#140 Delaware
Tennessee17.920-121309-104-7#97 Texas A&M#139 Georgia
Southern Mississippi15.323-800-45-199-7#100 East Carolina
Middle Tennessee128-500-31-21111-3
Massachusetts0.721-1002-49158-4#44 La Salle
#85 Xavier
#146 George Washington
Maryland0.722-1123-14-553-7
Alabama< 0.120-1100-48144-8#112 Dayton
#138 Mercer
#177 Tulane
Iowa< 0.121-1204-55-423-8
Baylor< 0.117-1422-95-6-44-7#54 Kentucky#135 Coll. of Charleston
#168 Northwestern
Akron *< 0.124-601-14177-4#65 Ohio#134 Kent St.
Bucknell *< 0.127-501032512-4#99 Loyola MD#107 Lehigh
Stephen F. Austin *< 0.123-301120611-3#40 Oklahoma
Louisiana Tech< 0.125-60113-1710-5
Valparaiso *< 0.125-700-22-1710-5#67 Detroit#225 Loyola Chicago
New Mexico St. *< 0.122-1001-24-206-8

BubbleGrid for March 14

See this to learn what the BubbleGrid all about.

The CTD selection model is telling us right now that all of the teams it currently forecasts in the field have at-large worthy profiles (their selection "score" is close to 100) by historical standards. The next team out (Tennessee) also has a score close to 100, but there is a steep drop after that down to Mississppi (2.9).

The best way to interpret that you could swap Tennessee for just about any team above them on this list and not get any complaints from me. After the last few days, you might be tempted to swap out Boise State or St. Mary's, but I'd caution you not to overreact to 1 game out of 30 in a team's profile.

The committee doesn't overreact:

[Bracketologists] react literally on a day‑to‑day or sometimes half‑to‑half or game‑by‑game basis to what they see. That's who they are, what they get paid to do. It helps to drive and build interest, makes for great conversation the next day around the water cooler. But we have a different circumstance here. We try to allow all the dust to settle and have all the facts and detail in front of us, evaluate it in totality, not have those emotion‑based ups and downs during the course of the year. We find what they do great. I think it's good for the game. But it's a little different than how we go about it.

Wait - people get paid to do this?

  • Hover over a value for more information
  • Shaded teams are predicted to be in the field
  • Teams with a * are the projected automatic qualifier (AQ) for their conference. Bold shaded teams have clinched the AQ berth
  • Teams are listed in order of their selection confidence by the CTD model. In other words, the last shaded non-AQ team is the last team in, and the first non-shaded team is the first team out.

Data is through games of Wednesday, March 13.

1-251-501-1001-200RoadGood road winsBad home losses
TeamScoreW-LWW+/-W+/-+/-+/-
Colorado> 99.921-1023-19195-7#54 Oregon
#73 Stanford
#180 Oregon St.
Oklahoma> 99.920-1023-39165-7#63 Baylor
Cincinnati> 99.922-1014-49-155-5#38 Pittsburgh
Wichita St.> 99.926-803183117-5#27 VCU
#68 Air Force
#76 Indiana St.
#106 Evansville
Villanova> 99.919-1245-28-237-7#45 Connecticut#272 Columbia
Belmont *> 99.924-601-16498-5#73 Stanford
#74 Eastern Kentucky
Virginia> 99.921-102428563-8#42 Wisconsin
#83 Maryland
#140 Delaware
Boise St.> 99.919-1024-38055-8#26 Creighton
#71 Wyoming
St. Mary's> 99.926-601-240119-2#70 Brigham Young
Iowa St.> 99.921-1023-47-144-8#63 Baylor
La Salle> 99.921-812-16-188-6#27 VCU#200 Central Connecticut
Kentucky99.221-1012-27-224-7#56 Mississippi
#96 Texas A&M
California9920-1013-26-466-5#13 Arizona
#54 Oregon
#57 Denver
Oregon9823-823-27096-5#21 Nevada Las Vegas
#28 UCLA
#88 Washington
Tennessee97.319-111409014-7#96 Texas A&M#124 Georgia
Mississippi2.923-801-37166-6#55 Tennessee
#90 Louisiana St.
Middle Tennessee< 0.128-500-21-21111-3
Baylor< 0.117-1323-75-5-34-7#50 Kentucky#138 Coll. of Charleston
#165 Northwestern
Southern Mississippi< 0.121-800-54-279-7#95 East Carolina
Iowa< 0.120-1114-45-322-8
Denver< 0.121-801-33-528-6#84 Northern Iowa
Alabama< 0.119-1101-37034-8#109 Dayton
#136 Mercer
#178 Tulane
Massachusetts< 0.119-1001-58038-4#40 La Salle
#78 Xavier
#143 George Washington
Louisiana Tech *< 0.125-50113-1710-5
Xavier< 0.117-133525-314-7#126 Vanderbilt
#246 Wofford
Akron *< 0.123-601-14167-4#75 Ohio#141 Kent St.
Bucknell *< 0.127-501032512-4#100 Loyola MD#107 Lehigh
Stephen F. Austin *< 0.122-301120611-3#35 Oklahoma
Stanford< 0.118-1402-76-605-6#49 California
#85 Arizona St.
#115 Southern California
Arkansas< 0.119-1214-17-3-11-9
New Mexico St.< 0.120-1002-14-2-16-8
Valparaiso *< 0.125-700-22-1710-5#66 Detroit#226 Loyola Chicago

2013: A Bracketing Adventure (Episode I)

Last year around this time, I went through the exercise of bracketing the first four seed lines (using the CTD seed list current at the time). While I think most specific predictions of sites are basically crapshoots beyond the first few seed lines, it can be useful for understanding how the committee does this. We can also spot some potential issues the committee may face during bracketing, though it depends on how similar the actual seed list is to ours.

Speaking of the actual seed list:

See last year's post for details, but we first need to identify which pod and regional sites are prohibited for particular teams.

From the bracketing principles:

A team will not be permitted to play in any facility in which it has played more than three games during its season, not including conference post-season tournaments.
A host institution’s team shall not be permitted to play at the site where the institution is hosting. However, the team may play on the same days when the institution is hosting.

Pod site prohibitions (not comprehensive)

Auburn Hills - Oakland

Austin - Texas

Dayton - Dayton

Kansas City - none (MVC hosting; Kansas played 3 games at the Sprint Center)

Lexington - Kentucky

Philadelphia - Temple

San Jose - none (WCC hosting)

Salt Lake City - Utah

Region site prohibitions (not comprehensive)

Washington, D.C. - Georgetown

North Texas - none (Big 12 hosting)

Indianapolis - IUPUI

Los Angeles - Pepperdine

The committee will then place the four “number 1 seed” teams seeded 1 through 4 in each of the four regions, then determine the Final Four semifinals pairings, making best effort to pair the top No. 1 seed’s region against the fourth No. 1 seed’s region and the second No. 1 seed’s region against the third No. 1 seed’s region.

The main principle to follow during bracketing (across all seed lines, not just the first line) is: Teams will remain in or as close to their areas of natural interest as possible. A team moved out of its natural area will be placed in the next closest region to the extent possible. If two teams from the same natural region are in contention for the same bracket position, the team ranked higher in the seed list shall remain in its natural region.

So, with that in mind, let's start bracketing. Keep in mind these are based on today's 1-68 seed list as determined by the Crashing the Dance seeding model, and is not necessarily how I think the seed list will end up on Sunday.

By the way, like others you'll find our distance chart useful here.

1. Duke - East (Washington, D.C.)

This is pretty straightforward. Durham is easily closer to D.C. than to any other regional site. Duke is the only school in the first 10 of our seed list with D.C. as its closest regional site, so even if they slide down a bit on the overall seed list, D.C. may be a good bet.

2. Indiana - Midwest (Indianapolis)

Even more of a no brainer than Duke to Washington. I don't think any of the other potential Big 10 champs could jump up to the top seed line, but there are other potential No. 1 seeds that could be sent to Indy. Don't book your hotels quite yet.

3. Louisville - South (North Texas)

The Cardinals are the first team sent outside their natural region, but whoever is the third No. 1 seed likely won't have to travel too far because...

4. Gonzaga - West (Los Angeles)

With Gonzaga a good candidate for one of the No. 1 seeds, it's likely that for the first time since 2008 (UCLA) that the top seed in the West will actually come from the West.

In this scenario, the Final Four pairings will be East vs. West and Midwest vs. South. (I'm not clear why the "making best effort" part is in there. I can see no other bracket principles that conflict pairing 1 vs. 4 and 2 vs. 3.)

The committee will then place the No. 2 seeds in each region in true seed list order.

This is when several of the bracketing principles can come into play, including:

Each of the first three teams selected from a conference shall be placed in different regions.
Conference teams shall not meet prior to the regional final unless a ninth team is selected from a conference. If the committee is unable to reconcile the bracket after exhausting all reasonable options, it has the flexibility to waive this principle to permit two teams from the same conference to meet each other after the second round.

5. Kansas - South (North Texas)

As the first team from the Big 12, there are no conference restrictions here. As I noted last week:

That was referring to the Jayhawks being on the No. 1 line, but they still get their choice here as the first No. 2 seed. So Jerry's World, it is.

6. New Mexico - East (Washington, D.C.)

7. Georgetown - Midwest (Indianapolis)

8. Miami - West (Los Angeles)

The Lobos are actually marginally closer to Arlington than to L.A., but with Kansas already there we'd next like to put them in California.

However, the next two teams cause conflicts, so we have to end up sending them all the way to D.C.

(Aside: For what it's worth, I think the seeding model is slightly overrating the Lobos at #6 overall. If New Mexico is on the 3 seed line, they may avoid being shuffled for other conflicts and could stay west of the Mississippi.)

As host of the East regional, Georgetown is prohibited from playing in D.C. The next closest site is Indianapolis, and since with the first Big East team (Louisville is still in the Big East, right?) already slotted to North Texas, we have no conflicts.

With Duke already in D.C., Miami can't go there because of conference restrictions. Since Georgetown is higher on the our imaginary seed list than Miami, the Hurricanes unfortunately have to head west.

This assumes that the committee puts a higher preference on keeping a team on their natural seed line near the top of the bracket than keeping them closer to home.

The committee will then place the No. 3 seeds in each region in true seed list order.

Additional bracketing principles can come into play on the third line, including:

No more than one team from a conference may be seeded in the same grouping of four in line Nos. 1-4 and 13-16 in a region, unless a conference has four or more teams seeded in line Nos. 1-4. In lines No. 5-12, two teams from the same conference may be placed in the same group of four

9. Michigan State - East (Washington, D.C.)

10. Michigan - South (North Texas)

The next two are essentially the same from a bracketing perspective - same conference, roughly the same distance from each regional site. Indiana is already in the Midwest, so neither get their first choice. The eventual relative order these two could come down to the B1G tourney, but for now we'll send the Spartans to Washington and Michigan to to North Texas.

11. Florida - Midwest (Indianapolis)

The Gators will be the first team from the SEC, so they don't have to worry about conference conflicts of their own creation. They may still fall prey to other conference conflicts or relative balance between regions, but for now we'll put them in the Midwest.

12. Ohio State - West (Los Angeles)

A fourth B1G team in the top 12. The first three have to be in different regions, but there is more flexibility with the fourth. However, putting them in Indy doesn't really help with balance, and it sends a team above them on the seed list to an unfavorable location. So, California it is for the Buckeyes.

13. Marquette - East (Washington, D.C.)

14. St. Louis - Midwest (Indianapolis)

15. Arizona - West (Los Angeles)

16. Syracuse - South (North Texas)

We're going to keep these four together because the initial slotting is fairly straightforward and the interesting stuff doesn't come up until afterward. Marquette is closer to Indy, but we already have Georgetown there. St. Louis and Arizona get their preferred site, and Syracuse heads south. That does set up a potential Big East Sweet 16 matchup, but that's allowed as long as there are four Big East teams in lines 1 through 4. It also slightly helps the regional balance. Speaking of that…

After the top four seed lines have been assigned, determine the relative strengths of the regions by adding the "true seed" numbers in each region to determine if any severe numerical imbalance exists. Generally, no more than five points should separate the lowest and highest total.

Uh oh. Here's how the numbers add up:

East: 29

South: 34

Midwest: 34

West: 39

So much for Duke always getting the easy bracket, right? We do have a bit of an imbalance, but our hands are tied somewhat because (1) there are four teams from each of the Big Ten and Big East and (2) Georgetown can't play in D.C.

We presume the committee won't tweak the No. 1 line to keep geographical preference, and the restrictions mean can't really do much on the No. 2 line. We also presume the committe won't swap teams between the No. 2 and 3 lines.

The big problem with tweaking the No. 3 line is that there are three Big Ten teams and they're currently in order of geographical preference. If we shuffle them around, we could be sending the second best Big Ten team to California while keeping the fourth best closer to home. What would the committee do in this case?

It looks like we might be able to make it good enough by tweaking a few things on the No. 4 line. Let's put Marquette in L.A., move Arizona to Dallas (their second-closest site) and put Syracuse in D.C. The downside is that the fourth-best Big East team is the closest to home, which seems unfair. The committee's bracketing principles give them flexibility, but it also leaves ambigutiy and makes it hard to know what they'd do in a situation like this.

Much better now:

East: 32

South: 33

Midwest: 34

West: 37

In “true seed” order, the committee then assigns each team (and, therefore, all teams in its bracket group—e.g., seeds 1, 8, 9, 16) to second-/third-round sites.

In other words, we need to assign pod sites for each of the 16 teams we've placed in regions. Most in the first several lines get their preferred site.

1. Duke - Philadelphia

2. Indiana - Dayton

3. Louisville - Lexington

4. Gonzaga - Salt Lake City

5. Kansas - Kansas City

6. New Mexico - Salt Lake City

7. Georgetown - Philadelphia

8. Miami - Lexington

9. Michigan State - Auburn Hills

10. Michigan - Auburn Hills

11. Florida - Dayton

12. Ohio State - Kansas City

13. Marquette - Austin

14. St. Louis - Austin

15. Arizona - San Jose

16. Syracuse - San Jose

Indiana is slightly closer to Lexington, but I doubt the committee will send the Hoosiers into Wildcat country. The Wolverines and Sparty hanging together in Auburn Hills could be fun. One or both has a pretty good chance of being there unless they slip a bit.

Time permitting, we'll tackle the remainder of the seed lines in Episode II. There are fewer restrictions in the 5-12 lines, and generally fewer conference conflicts in the 13-16 lines, so they may not be quite as exciting. There is also more chance of predicting the wrong seeds in the 5 to 12 range, so it may not be all that useful.

BubbleGrid for March 7: Oh, Virginia

See this to learn what the BubbleGrid all about.

Oh, Virginia. What are we going to do with you?

No team in recent years has a profile quite like the Hoos. After losing at Florida State Thursday, Virginia has won 3 games against RPI top 25 teams while losing 7 (7!) to teams outside the top 100. In the 13 years of data used to build the CTD selection model, no team with at least 3 top 25 wins has more than 4 losses outside the top 100. That makes it difficult for the model to draw a solid conclusion.

If we expand the window a bit, it doesn't get any clearer. 8 of the 9 teams with 3 or more top 25 wins and 4 101+ losses made the field, but only 1 (the infamous 2011 Southern Cal) of the 9 teams with 2 or more top 25 wins and 6 101+ losses made the field.

The problem with bubble teams that many forget is that you can't evaluate them in a vacuum. No matter how much you think we should shrink the field, there have to be 37 at-large teams. If you want to leave out Virginia, you have to replace them with another team with their own flaws.

That Southern Cal team was selected - and based on the seeding, they were not the last team in - despite 3 losses outside the top 200, and they didn't have a win as good as Virginia's best win. Selection is relative to each season, so that certainly doesn't guarantee anything for Virginia, but it's still too early to count them out.

Of course, now that I write this, they'll probably drop out of the field on Thursday's update.

  • Hover over a value for more information
  • Shaded teams are predicted to be in the field
  • Teams with a * are the projected automatic qualifier (AQ) for their conference. Bold shaded teams have clinched the AQ berth
  • Teams are listed in order of their selection confidence by the CTD model. In other words, the last shaded non-AQ team is the last team in, and the first non-shaded team is the first team out.

Data is through games of Wednesday, March 6.

1-251-501-1001-200RoadGood road winsBad home losses
TeamScoreW-LWW+/-W+/-+/-+/-
Wichita St.> 99.924-703284127-5#27 VCU
#77 Indiana St.
#86 Air Force
#101 Evansville
St. Mary's> 99.925-501-174109-2#61 Brigham Young
#90 Santa Clara
#99 Utah St.
Cincinnati> 99.920-1014-48-255-5#37 Pittsburgh
Virginia> 99.920-93427563-7#40 Wisconsin
#85 Maryland
#126 Delaware
California> 99.920-101508-266-5#14 Arizona
#49 Oregon
#67 Denver
#96 Southern Cal
Villanova> 99.918-1245-27-346-7#47 Connecticut#265 Columbia
Belmont *> 99.922-611-16478-5#59 Stanford
#65 Eastern Kentucky
Temple> 99.922-81309498-4#53 Villanova
#55 Massachusetts
#81 Charlotte
#109 Canisius
#110 St. Bonaventure
#221 Duquesne
Boise St.> 99.918-923-37055-8#32 Creighton
#74 Wyoming
La Salle99.121-70205-198-5#27 VCU#197 Central Conn.
Iowa St.92.720-1023-46-223-8#69 Baylor
Louisiana Tech *8625-301153910-3#99 Utah St.
Kentucky72.420-901-36-334-6#57 Mississippi
#91 Texas A&M
Tennessee46.418-1112-18-104-7#91 Texas A&M#134 Georgia
Middle Tennessee *37.127-400-31-21211-4
Akron *2.623-511-14187-4#66 Ohio
Massachusetts< 0.118-901-48137-4#41 La Salle
#83 Xavier
#146 G. Washington
Stanford< 0.118-1303-76-715-6#50 California
#98 Arizona St.
Mississippi< 0.122-801-26055-6#56 Tennessee
Iowa< 0.119-1113-55-312-8
Southern Mississippi< 0.120-800-53-359-7#94 East Carolina
Alabama< 0.118-1100-47024-8#104 Dayton
#128 Mercer
#176 Tulane
Stephen F. Austin *< 0.120-311120611-3#24 Oklahoma
Bucknell *< 0.125-501032412-4#93 Loyola MD#105 Lehigh
Denver< 0.119-800-44-308-6#72 Northern Iowa
#99 Utah St.
Xavier< 0.117-121535-224-6#129 Vanderbilt
#235 Wofford
Brigham Young< 0.120-1000-55-306-5#89 Weber St.
#90 Santa Clara
#151 San Francisco
Baylor< 0.116-1301-74-6-44-7#52 Kentucky#144 Northwestern
#149 College of Charleston

History may be on their side

CTD was asked this yesterday on Twitter:

The question was presumably in response to Louisville jumping up to the last slot on the No. 1 seed line.

As I said last night:

Louisville is in the group of possible number 1 seeds, and one of the reasons why the CTD seeding model is high on them right now is their #1 overall ranking (as of Wednesday) in the Sagarin ratings. Obviously, the Sagarin ratings are but one of a set of tools confirmed to be used behind closed doors. (How they admit to using them is another story. Hopefully it's not only to resolve "descrepancies".)

Anyway, in the 13 years of historical data used to build the CTD model this season, there have only been 2 teams ranked 1st or 2nd in the Sagarin ratings at the end of the regular season that were not given a No. 1 seed. Cincinnati likely would have been a 1 seed in 2000 if not for Kenyon Martin's injury.

SeasonTeamSagarinSeed
2000Cincinnati12
2000Stanford21
2001Duke11
2001Stanford21
2002Duke11
2002Cincinnati21
2003Kentucky11
2003Arizona21
2004Duke11
2004Saint Joseph's21
2005Illinois11
2005North Carolina21
2006Duke11
2006Connecticut21
2007North Carolina11
2007Ohio St.21
2008North Carolina11
2008UCLA21
2009North Carolina11
2009Pittsburgh21
2010Kansas11
2010Duke21
2011Ohio St.11
2011Kansas21
2012Kentucky11
2012Ohio St.22
2013Louisville1?
2013Indiana2?

Any one predictor variable shouldn't be trusted by itself, but that's a pretty solid track record.

BubbleGrid is back

You know the deal. Selection Sunday is near. See this to learn what the BubbleGrid all about.

The big thing that stands out right now is that the CTD selection model thinks a little more highly of teams right around the cut line than it has in recent years.

For example, compare today's grid with last season's final version of the grid. The last four teams predicted in 2012 to be in the field (three of which did not make the field) all had scores of nearly zero, meaning their profiles did not match that of teams that historically were (or would have been) at-large selections.

This season, the last at-large team currently in the field as of today (Tennessee) has a score of 70.2 (out of 100), so the teams on the good side of the bubble right now have historically good profiles. Of course, there are still a few more regular season games and conference tournaments left to play, so thinks can and often do change.

I'm not as sold on Kentucky's position as the model is. They haven't looked great since losing Nerlens Noel for the season. After posting a 21.5 Net Efficiency Margin with Noel, their Noel-less NEM has been been 7.0 - roughly 100th best in the country - in an admittedly small five-game sample. They still have a chance for a committee-convincing signature win this weekend against Florida, but time's running out for any shot at a repeat.

  • Hover over a value for more information
  • Shaded teams are predicted to be in the field
  • Teams with a * are the projected automatic qualifier (AQ) for their conference. Bold shaded teams have clinched the AQ berth
  • Teams are listed in order of their selection confidence by the CTD model. In other words, the last shaded non-AQ team is the last team in, and the first non-shaded team is the first team out.
1-251-501-1001-200RoadGood road winsBad home losses
TeamScoreW-LWW+/-W+/-+/-+/-
St. Mary's> 99.925-501-174109-2#61 Brigham Young
#92 Santa Clara
#98 Utah St.
Virginia> 99.920-93427563-7#41 Wisconsin
#70 Maryland
#127 Delaware
Cincinnati> 99.920-1014-48-245-5#39 Pittsburgh
Belmont *> 99.922-611-16478-5#64 Eastern Kentucky
#72 Stanford
Temple> 99.921-81309497-4#54 Massachusetts
#55 Villanova
#82 Charlotte
#105 St. Bonaventure
#108 Canisius
#216 Duquesne
Boise St.> 99.918-823-27165-7#35 Creighton
#66 Wyoming
La Salle97.820-70205-188-5#31 VCU#194 Central Connecticut
Louisiana Tech *86.725-301153910-3#98 Utah St.
Villanova8617-1234-36-426-7#40 Connecticut#267 Columbia
Kentucky76.420-901-36-334-6#58 Mississippi
#83 Texas A&M
Tennessee70.217-1112-18-113-7#83 Texas A&M#129 Georgia
Middle Tennessee *39.927-400-31-21211-4
Iowa St.819-1022-55-313-8#69 Baylor
Akron *6.422-511-14187-4#73 Ohio
Massachusetts< 0.118-901-48137-4#44 La Salle
#93 Xavier
#149 George Washington
Southern Mississippi< 0.120-700-53-359-6#100 East Carolina
Mississippi< 0.121-801-25-135-6#56 Tennessee
Maryland< 0.120-922-13-453-6
Alabama< 0.118-1000-47144-7#104 Dayton
#132 Mercer
#175 Tulane
Bucknell *< 0.124-501032412-4#96 Loyola MD#103 Lehigh
Stephen F. Austin *< 0.120-311120611-3#25 Oklahoma
Arkansas< 0.118-1123-16-3-21-8
Stanford< 0.117-1302-85-804-6#95 Arizona St.
Denver< 0.119-800-44-308-6#75 Northern Iowa
#98 Utah St.
St. John's< 0.116-1203-24-414-6#49 Cincinnati#110 Rutgers
#203 NC Asheville
Brigham Young< 0.120-1000-55-306-5#89 Weber St.
#92 Santa Clara
#154 San Francisco
Baylor< 0.116-1301-74-6-44-7#51 Kentucky#143 Northwestern
#150 College of Charleston
Valparaiso *< 0.123-700-22-1710-5#57 Detroit#220 Loyola Chicago

Narrative busters

Hopefully, this will be the first in an intermittent series of posts poking holes in the popular narratives using facts and data and stuff.

After Ryan Kelly's triumphant return Saturday, all people want to talk about is how his return will help the Blue Devils on offense. Indeed, the Devils have been a much better team with with the senior in the lineup. However, using our Net Efficiency Margin metric, we see that Kelly's absence manifested itself far more on defense than on offense.

All numbers are prior to March 2 game against Miami.

DukeOff NEMDef NEMTotal NEM
With Kelly (15 games) 16.3-17.734.0
Without Kelly (13 games) 18.5-6.224.7

In other words, after accounting for tempo and opponent strength Duke was actually slightly better as a team on offense during Kelly's 13-game absence, but they were significantly worse (11.5 points per 100 possessions) on defense. (Lower defensive NEM values are better.)

Let's temper the title favorites talk based on a single game where they needed a performance described - by Coach K, no less - as one of the best ever by a Duke player in Cameron to win by one possession at home. However, if they can regain that defensive form and continue to get performances like Kelly's 36-point outburst, the Devils will be contenders in Atlanta.