2015 NCAA Tournament: Building a better bracket with different metrics
Following the Selection Show, our resident bracketologist decided to compare the Selection Committee’s work with brackets generated by the two most well-known metrics in college basketball. Here are the results.
Even before the Selection Show turned into true "What the heck was that?" experience, I had decided to bring this RPI/KenPom comparison post back after a few years away. I noticed that SB Nation's excellent Xavier Musketeers blog Banners on the Parkway has been running KenPom bracketology posts for awhile, so I've taken their baton on am running with it, with a twist.
Enter the RPI
Since the NCAA still insists on using its in-house Ratings Percentage Index as its primary tool in organizing the field of 68, I’ll start with this. I find it interesting to see how the Committee’s deliberations differ from what its metric suggests. For this exercise, I have added the 36 best available teams according to the official RPI ranking on NCAA.com to the 32 automatic qualifiers and seeded them as closely to the 1-68 list as possible, only making adjustments to account for bracketing rules. if a team’s seeding changed, I placed the difference next to its name in bold.
If you’re interested, I suggest reading Ken Pomeroy’s RPI primer before delving into the bracket. It does a great job of explaining what it is, and Pomeroy actually links to pieces defending it at the end.
After the matchups and rundown, I’ll point out the key differences between this bracket and the real deal.
| (1) MIDWEST Cleveland (Thu/Sat) | (2) WEST Los Angeles (Thu/Sat) | ||
|---|---|---|---|
| Louisville (Thu/Sat) | Pittsburgh (Thu/Sat) | ||
| 1 | KENTUCKY (SEC) | 1 | VILLANOVA (Big East) |
| 16 | HAMPTON/ROBERT MORRIS | 16 | UAB (C-USA) -2 |
| 8 | Butler -2 | *8 | Colorado State OUT |
| 9 | BYU +2 | *9 | Temple OUT |
| Columbus (Fri/Sun) | Seattle (Fri/Sun) | ||
| 5 | Oklahoma | 5 | VCU (A 10) +2 |
| 12 | VALPARAISO (Horizon) +1 | 12 | WOFFORD (SoCon) |
| 4 | NORTHERN IOWA (MVC) +1 | 4 | Wichita State +3 |
| 13 | EASTERN WASHINGTON (Big Sky) | 13 | HARVARD (Ivy) |
| Jacksonville (Thu/Sat) | Louisville (Thu/Sat) | ||
| 6 | West Virginia -1 | 6 | Providence |
| 11 | Iowa -4 | 11 | Tulsa/UCLA |
| 3 | North Carolina +1 | 3 | IOWA STATE (Big 12) |
| 14 | ALBANY (AE) | 14 | UC IRVINE (Big West) -1 |
| Seattle (Fri/Sun) | Charlotte (Fri/Sun) | ||
| 7 | San Diego State +1 | 7 | Oregon +1 |
| 10 | Cincinnati -2 | 10 | Ohio State |
| 2 | GONZAGA (WCC) | 2 | Duke -1 |
| 15 | NEW MEXICO STATE (WAC) | 15 | LAFAYETTE (Patriot) +1 |
| (4) EAST Syracuse (Fri/Sun) | (3) SOUTH Houston (Fri/Sun) | ||
| Omaha (Fri/Sun) | Omaha (Fri/Sun) | ||
| 1 | WISCONSIN (Big Ten) | 1 | Kansas +1 |
| 16 | COASTAL CAROLINA (Big South) | 16 | MANHATTAN/NORTH FLORIDA |
| 8 | Xavier -2 | 8 | Dayton +3 |
| 9 | Davidson +1 | 9 | STEPHEN F. AUSTIN (Southland) +3 |
| Portland (Thu/Sat) | Columbus (Fri/Sun) | ||
| 5 | Arkansas | 5 | Utah |
| 12 | St. John's -3 | 12 | Oklahoma State/Old Dominion |
| 4 | NOTRE DAME (ACC) | 4 | SMU (AAC) +2 |
| 13 | WYOMING (MW) -1 | 13 | GEORGIA STATE (Sun Belt) +1 |
| Jacksonville (Thu/Sat) | Pittsburgh (Thu/Sat) | ||
| 6 | Michigan State +1 | 6 | Louisville -2 |
| 11 | Boise State | 11 | Texas |
| 3 | Baylor | 3 | Maryland +1 |
| 14 | NORTH DAKOTA STATE (Summit) +1 | 14 | NORTHEASTERN (CAA) |
| Portland (Thu/Sat) | Charlotte (Fri/Sun) | ||
| 7 | BUFFALO (MAC) +5 | 7 | Georgetown -3 |
| 10 | N.C. State -2 | 10 | Georgia |
| 2 | ARIZONA (Pac-12) | 2 | Virginia |
| 15 | TEXAS SOUTHERN (SWAC) | 15 | BELMONT (OVC) |
| FIRST FOUR (Dayton) | |||
| Tuesday: To Louisville | Tuesday: To Louisville | ||
| 16 | HAMPTON (MEAC) | *11 | Tulsa OUT |
| 16 | ROBERT MORRIS (NEC) | 11 | UCLA |
| Wednesday: To Omaha | Wednesday: To Columbus | ||
| 16 | MANHATTAN (MAAC) | 12 | Oklahoma State -3 |
| 16 | NORTH FLORIDA (A-Sun) | *12 | Old Dominion OUT |
* = new team
| BIDS BY CONFERENCE | AVOIDING DAYTON | ARRIVALS | DEPARTURES |
| Big 12: 7 | Boise State (41) | Colorado State | Indiana |
| ACC: 6 | Texas (42) | Old Dominion | LSU |
| Big East: 6 | Iowa (43) | Temple | Ole Miss |
| Big Ten: 5 (-2) | St. John's (44) | Tulsa | Purdue |
| AAC: 4 (+2) | LAST FOUR IN | ||
| MW: 4 (+1) | Tulsa (45) | ||
| Pac-12: 4 | Old Dominion (46) | ||
| A 10: 3 | UCLA (48) | ||
| SEC: 3 (-2) | Oklahoma State (50) | ||
| MVC: 2 | FIRST FOUR OUT | ||
| WCC: 2 | Iona (51) | ||
| C-USA: 2 (+1) | Purdue (53) | ||
| One-Bid Conferences: 20 | Green Bay (54) | ||
| Ole Miss (55) | |||
| NEXT FOUR OUT | |||
| LSU (56) | |||
| Richmond (57) | |||
| Louisiana Tech (58) | |||
| Indiana (59) |
Other notables: 62. Murray State, 63. Yale, 64. Illinois State, 65. Miami, 66. Rhode Island, 67. Texas A&M, 72. Illinois
The RPI bracket actually favors several of the bubble teams left out by the Committee, even going to the lengths of allowing for a Colorado State-Temple matchup in the Round of 64. And if you thought the Big Ten was a bit down this year, the RPI agrees with you, as Indiana and Purdue are both NIT-bound in this scenario. The weekend outbreak of #SECBasketballFever in Nashville is also punished by the metric, as both LSU and Ole Miss depart. Curiously, Georgia is in the same position on this bracket and the real thing, line 10.
At the top of the bracket, Kansas rise up to the top line, thanks to its typically impossible non-conference schedule, costing Duke its place. Maryland, North Carolina, and Northern Iowa receive slight seeding bumps, with all in the top 16. Georgetown and Wichita State, a four and seven in reality, respectively, swap places. But the biggest seeding winner is MAC Tournament champion Buffalo, rising from a 12 to a seven. On the flip side, Iowa slide from a seven to an 11.
Of the 68 teams, the RPI only gets 64 correct. In the seeding department, it pegged 29 teams correctly, placing another 17 within one line of their actual place. The RPI misses 11 by two lines, five by three, and one apiece by four and five.
In other words, the RPI needs to work on its bracketing skills a bit.
The KenPom Bracket
I’ve expanded upon Joel D.‘s work over at Banners on the Parkway by giving the KenPom projection the full treatment, with specific dates and sites and bracketing rules. This was a little easier to put together than the RPI bracket since KenPom.com highlights NCAA qualifiers and notes their seedings, making it a snap to determine where the automatic qualifiers, in particular, fell in his ranking. I used the same notation in this bracket to reflect seeding changes.
| (1) MIDWEST Cleveland (Thu/Sat) | (2) WEST Los Angeles (Thu/Sat) | ||
|---|---|---|---|
| Louisville (Thu/Sat) | Portland (Thu/Sat) | ||
| 1 | KENTUCKY (SEC) | 1 | ARIZONA (Pac-12) +1 |
| 16 | HAMPTON/TEXAS SOUTHERN | 16 | NORTH DAKOTA STATE (Summit) -1 |
| 8 | VCU (A 10) -1 | 8 | BYU +3 |
| 9 | Cincinnati -1 | 9 | Maryland -5 |
| Pittsburgh (Thu/Sat) | Seattle (Fri/Sun) | ||
| 5 | Michigan State +2 | 5 | Louisville -1 |
| 12 | Ole Miss/Vanderbilt | 12 | GEORGIA STATE (Sun Belt) +2 |
| 4 | Wichita State +3 | 4 | Baylor -1 |
| 13 | HARVARD (Ivy) | 13 | UC IRVINE (Big West) |
| Omaha (Fri/Sun) | Columbus (Fri/Sun) | ||
| 6 | Georgetown -2 | 6 | Butler |
| 11 | LSU -2 | 11 | Florida/Oregon |
| 3 | Oklahoma | 3 | NOTRE DAME (ACC) |
| 14 | NORTH FLORIDA (A-Sun) +2 | 14 | WYOMING (MW) -2 |
| Portland (Thu/Sat) | Seattle (Fri/Sun) | ||
| 7 | Providence -1 | 7 | San Diego State +1 |
| 10 | Boise State +1 | 10 | Oklahoma State +1 |
| 2 | Utah +3 | 2 | GONZAGA (WCC) |
| 15 | COASTAL CAROLINA (Big South) +1 | 15 | EASTERN WASHINGTON (Big Sky) -2 |
| (4) EAST Syracuse (Thu/Sat) | (3) SOUTH Houston (Fri/Sun) | ||
| Charlotte (Fri/Sun) | Omaha (Fri/Sun) | ||
| 1 | Virginia +1 | 1 | WISCONSIN (Big Ten) |
| 16 | LAFAYETTE/ROBERT MORRIS | 16 | BELMONT (OVC) -1 |
| 8 | Davidson +2 | 8 | Arkansas -3 |
| 9 | Georgia +1 | 9 | STEPHEN F. AUSTIN (Southland) +3 |
| Jacksonville (Thu/Sat) | Jacksonville (Thu/Sat) | ||
| 5 | SMU (AAC) +1 | 5 | Texas +6 |
| 12 | VALPARAISO (Horizon) +1 | 12 | BUFFALO (MAC) |
| 4 | IOWA STATE (Big 12) -1 | 4 | North Carolina -1 |
| 13 | WOFFORD (SoCon) -1 | 13 | NEW MEXICO STATE (WAC) +2 |
| Columbus (Fri/Sun) | Louisville (Thu/Sat) | ||
| 6 | Iowa +1 | 6 | Ohio State +4 |
| 11 | St. John's -2 | 11 | UCLA |
| 3 | NORTHERN IOWA (MVC) +2 | 3 | Kansas -1 |
| 14 | UAB (C-USA) | 14 | NORTHEASTERN (CAA) |
| Pittsburgh (Thu/Sat) | Charlotte (Fri/Sun) | ||
| 7 | West Virginia -2 | 7 | Xavier -1 |
| 10 | N.C. State -2 | 10 | Dayton +1 |
| 2 | VILLANOVA (Big East) -1 | 2 | Duke -1 |
| 15 | MANHATTAN (MAAC) +1 | 15 | ALBANY (AE) -1 |
| FIRST FOUR (Dayton) | |||
| Tuesday: To Louisville | Tuesday: To Pittsburgh | ||
| 16 | HAMPTON (MEAC) | 12 | Ole Miss -1 |
| 16 | TEXAS SOUTHERN (SWAC) -1 | *12 | Vanderbilt OUT |
| Wednesday: To Charlotte | Wednesday: To Columbus | ||
| 16 | LAFAYETTS (Patriot) | *11 | Florida OUT |
| 16 | ROBERT MORRIS (NEC) | *11 | Oregon -3 |
* = new team
| BIDS BY CONFERENCE | AVOIDING DAYTON | ARRIVALS | DEPARTURES |
| Big 12: 7 | Dayton (40) | Florida | Indiana |
| SEC: 7 (+2) | UCLA (41) | Vanderbilt | Purdue |
| ACC: 6 | St. John's (42) | ||
| Big East: 6 | LSU(43) | ||
| Big Ten: 5 (-2) | LAST FOUR IN | ||
| Pac-12: 4 | Ole Miss (44) | ||
| A 10: 3 | Florida (45) | ||
| MW: 3 | Oregon (46) | ||
| AAC: 2 | Vanderbilt (47) | ||
| MVC: 2 | FIRST FOUR OUT | ||
| WCC: 2 | Stanford (48) | ||
| One-Bid Conferences: 21 | Purdue (49) | ||
| Texas A&M (50) | |||
| Richmond (51) | |||
| NEXT FOUR OUT | |||
| Miami (52) | |||
| Indiana (53) | |||
| Rhode Island (55) | |||
| Temple (56) |
Other notables: 64. Old Dominion, 65. Illinois, 68. Colorado State, 72. Tulsa, 73. Murray State
The first thing that struck me when putting the KenPom bracket together was how easily the first few lines flowed. Its ideal East Regional final, for example, features No. 4 overall Virginia and No. 5 Villanova. Gonzaga is paired with Arizona in the West with no manipulation. Even that Midwest 1 vs. 2 -- Kentucky vs. Utah -- doesn't look too awful, despite the Utes elevated seed.
However, this bracket is clearly based on a metric that is used to evaluate team's efficiency on the court, not necessarily its wins and losses. Therefore, the variance between it and reality is quite wide. Consider the First Four, which includes a pair of SEC teams that weren't even under consideration, Florida and Vanderbilt. The 16-17 Gators weren't even the last team in the field here, as they actually ranked ahead of both the Commodores and Oregon.
Again, the metric is not a fan of the Big Ten, as Indiana and Purdue are both gone, replaced by the two SEC schools. That means #SECBasketballFever’s total in this field is seven, matching the Big 12, which seems a bit off.
As you might suspect, KenPom doesn't do a great job with seeding. Of the 66 teams KenPom projected into the field, it got the seeds of just 15 right. It was off by one line for 29 teams, and by two for 13 more. Five teams were three lines lower or higher than reality, while one apiece were off by four lines (Ohio State, up from a 10 to a six), five lines (Maryland, down from a four to a nine), and six lines (not surprisingly this was Texas, bumped from an 11 all the way to a five).
By the way, that UCLA Bruins squad that was a surprise Committee selection made both of these fields -- though not exactly comfortably.
So, there’s some post-Selection Show, pre-First Four food for thought. In what ways would you like to see the selection, seeding, and bracketing process change?
See More:
