Bubble Analysis: Did last year teach us anything?

Success in projecting brackets begins with being able to correctly identify who is, and who is not, 'On the bubble', and then, by selection Sunday correctly identifying which of those bubble teams the committee will select for the field of 68, and which one's it will leave out.  The selection committee, because it has changes year over year, both in membership and leadership, tends to have its own unique criteria for winnowing the bubble and arriving at the final field, and therein lies the challenge ... are they going to apply similar thinking to this year's field, or will this committee break with past tendencies and start new ones?

There is obviously no way of knowing this for certain, but its still critical to look back at the previous season's selection and attempt to identify clues as to what the committee may do this season.  So let's take a look back and see what, if anything we can learn about last year's selection, and which bubble teams got in, which were left out.  I think it is fair to say that last year was one of the most difficult fields to project in some time based on the number of people who missed on multiple teams either getting in or not making it.  

The NCAA does provide some insight into their thinking by providing a true seed list after seeding and pairings have been revealed.  This seed list ranks all 68 teams from 1 to 68, and through this we can identify the 8 at-large teams that made up the last 8 teams selected, and how the committee ranked them. 

Here is the list of the final 8 teams to make the field last season, as ranked by the selection committee:

1.  Pittsburgh
2.  Temple
3.  Syracuse
4.  VCU
5.  Vanderbilt
6.  Michigan
7.  Wichita State
8.  Tulsa

While we don't know for certain what other teams were under consideration that in the end did NOT make the field, as the NCAA does not provide that, the following teams were widely regarded as teams most likely to be selected to make the field by the largest number of 'prognosticators', in no particular order

1.  St. Boneventure
2.  San Deigo State
3.  St. Mary's
4.  Monmouth
5.  Valparaiso

In trying to identify the characteristics that the teams that were selected had in common, I found some pretty interesting characteristics that seemed to have absolutely no bearing on selecting the bubble teams that many have believed have been important in the past. 

RPI  was meaningless:

• The average RPI rank of the 8 teams above that got in was 56.5.  
• The rank of those 5 that didn't was 43.6.  

St. Boneventure was the 30th ranked team by RPI and DIDN'T get in ... Tulsa was the 70th ranked RPI team and did. 

Road/Nuetral Record was meaningless -  Every year we hear pundits talk about the importance of winning away from the home gym as an indicator of tournament worthiness.  The logic ... you don't play any NCAA tournament games on your home floor, so you have to show the committee you can play well on the road.  But when it came to bubble teams, R/N record appeared to mean nothing last year

• Winning Pct in R/N Games for 8 teams above who got in - only 48.1%
• Winning Pct in R/N Games for 5 Teams above who missed - a very good 68.3%

Vanderbilt was awful away from it's place - going 5-11 (31.3%)

Monmouth was great on the road - going 17-6 (73.9%).  They played 67% of their games away from home, won 74% of them, and three times as many as Vanderbilt and still were left out.  Clearly, road record meant nothing to the committee when it came to selecting the final at large teams.

Winning Percentage vs RPI Top 100 or 150 was meaningless

• Winning Percentage of select bubble teams was just 42.6% and 47.5% respectively
• Winning Percentage of our 5 Teams not selected were 53% and 63% respectively.

This was among the most surprising finding.  Emphasis in the past seemed to have placed a premium on those teams who were successful more often than not against teams in the RPI top 150 and top 100.  Vanderbilt was a dismal 9-13 vs the RPI top 150, while Monmouth was 10-4.   Michigan was an awful 4-12 vs the top 100 ... St.Mary's a solid 6-3. 

So, those things didn't matter ... what did?

Getting wins against RPI Top 25 and Top 50 Opponents

• Of the 8 bubble teams selected by the committee all of them had at least 1 win vs. the RPI top 25, and all but one (Wichita State) had at least 2 wins vs. the RPI top 50.
• Of the 5 teams that didn't make the field, 3 did not have a win vs. the RPI top 25 and all but 1 had 2 wins or less vs the RPI top 50.
• The winning percentage of the selected teams vs the RPI Top 50 was 31.6% while of those left out, the percentage was 45%.   Raw wins, trumped winning percentage. 
• A team was better off being 2-7 vs top 50 (Pitt, VCU) than being 2-2 (St, Mary's, Monmouth) so long as they had a least 1 win against the Top 25.

Getting wins against the rest of the selected field

• Each of the teams that made the field won at least 4 games against the other 67 teams in the field (except Wichita State who had 3)
• Each of the teams that missed the field failed to win more than 4 (Except St. Boneventure who had 5).

Minimizing bad losses (Against teams outside RPI top 150)

• 7 of the 8 teams selected had 1 or fewer losses outside the RPI 150
• 4 of the 5 teams left out of the field had 1 or more losses outside the RPI 150 and 3 had losses to teams outside the top 200.

So can anything be gleaned from this?  Honestly we won't know until selection Sunday if these criteria weighed as heavily in selection this year as they seemed to last year.  But if it does, here are some of the teams currently on the bubble that would stand to benefit ... and others who may be in some trouble if they don't get some big wins.

By the way, I have not calculated the 'vs the teams under consideration' data, and won't until near the end of Conference tournaments.  Its a long and arduous task that becomes obsolete the moment a team plays itself out of bubble consideration, or a new team emerges as a conference tournament champion that changes the field.  That analysis will have to wait. 

So back to seeing what last year's criteria (minus vs. Field) tells us the bubble looks like today:  

First, lets define who our bubble teams are.  Based on the cumulative seed average found on projected brackets at http://bracketmatrix.com/, here are the 24 teams widely regarded to be bubble teams as of Monday morning fighting it out for the last 8 at large spots, in no particular order:

Marquette

Oklahoma State

Seton Hall

Syracuse

Tennessee

Vanderbilt

Georgia Tech

Kansas State

California

Providence

Illinois State

Rhode Island

Pittsburgh

Indiana

TCU

Clemson

Alabama

Georgia

Ole Miss

Boise State

Wake Forest

Texas Tech

Colorado State

Houston

• 6 Teams with at least 2 wins vs. RPI Top 25:  Marquette, Georgia Tech, Pitt, Syracuse, Providence, Indiana
• 5 Teams with at least 4 wins vs RPI top 50:  Marquette, Syracuse, Georgia Tech, Vanderbilt, Clemson
• 10 Teams without a loss outside RPI top 150:  Marquette, Georgia Tech, Seton Hall, Kansas State, Tennessee, TCU, Georgia, Ole Miss, Cal, Wake Forest

Only Marquette and Georgia Tech appear on all three lists ... Applying last year's criteria, they would appear to be well positioned today to be selected.

Perhaps it would be easier to find the teams missing all the criteria of last years selected bubble teams so we can rule them out:

• Teams without a win vs Top 25:  TCU, Georgia, Ole Miss, California, Wake Forest, Clemson, Oklahoma State, Alabama, Houston, Illinois State, Colorado State
• Teams with 1 win or less vs Top 50:  Houston, Colorado State, California, Wake Forest, Illinois State, Rhode Island, Boise State
• Teams with 1 loss or more outside RPI Top 200, or 2 losses outside RPI top 150:  Colorado State, Providence, Pitt, Vanderbilt, Syracuse, Texas Tech.

When looking at last year's at large field (not just bubble) only two teams made the field without a win against the Top 25 ... USC and Dayton.  

USC had 4 Top 50 wins though, were above .500 against top 150 and did not lose a game outside the RPI top 150.  Of this year's crop of bubble teams without a top 25 win, the only team possessing the same 'profile' as USC last year is California.  

Dayton's resume was a bit different ... they only had 3 top 50 wins, had a loss outside the RPI top 200 but were stellar against RPI top 150 (13-6) and top 100 (9-5).  They also seemed to get rewarded for playing a difficult non conference schedule (9th NC-SOS) and having success (11-3).  Oklahoma State is about the closest profile among bubble teams matching Dayton's from last year.  Oklahoma State has 0 top 25 Wins, but a winning record against the RPI Top 150 (.529) and a 10-2 NC Record against the 31st toughest NC schedule.

Wichita State was the only team a year ago with less than 2 wins against the top 50.  So what was unique about the rest of their resume that seemed to overcome the shortcoming of having so few Top 50 wins.  Well, of the bubble teams, they were the only one with a win against an RPI Top 10 (A home win vs. Utah).  They also did not have a loss outside the RPI top 150 and only 1 loss outside the top 100.  No team in this year's bubble field has a profile that matches ... Rhode Island has just one Top 50 Win, but it is against a top 25, though not top 10 ... and have 0 losses outside the RPI top 100. Tennessee has 2 Top 50 wins, and 1 against a top 10, but they have a loss outside the top 100 (Though not top 150).  Rhode Island appears to be the closest match

Finally, 5 At Large teams made the field with losses outside the top 200 ...  and 2 were bubble teams (Temple and Syracuse).  The others were Maryland, Dayton and Wisconsin.  We have explored Dayton's unique profile, and found our closest match, so lets focus on what Maryland and Wisconsin seemed to have in common.  Both won at least 62.5% of games against Top 150, 53% against top 100 and had at least 5 wins vs. the top 50.   None of our bubble teams this year share those criteria.

So ... of our 24 teams, it appears we can find 4 (Marquette, Georgia Tech, California, Oklahoma State) as teams with all or most of profiles of bubble or at large teams that qualified a year ago, suggesting that if last year's criteria were applied, they all may have done enough to get in already. 

The best thing the remaining teams can do, short of winning thier respective Conference Tournament and automatic bid,  is to do the obvious - get wins against the best opponents left on their schedule and avoid losses to teams over 100+ RPI.  If you don't have any RPI top 25 games left, and haven't gotten a win against one yet, chances are you are going to get left out.