Extra MustardSI On CampusFantasyPhoto GalleriesSwimsuitVideoFanNationSI KidsTNT
Get inside March Madness with SI.com's Luke Winn in the Tourney Blog, a daily journal of college basketball commentary, on-site reporting and reader-driven discussions.
3/12/2007 04:57:00 PM

Defending Your Bracket ... and Picking a Champ

Acie Law, Joseph Jones
Acie Law, Joseph Jones and the Ags are better than your average No. 3 seed.
Ronald Martinez/Getty Images
As you pore over your bracket picks this week, can you distinguish between the important factors and the irrelevant ones? You'll undoubtedly agonize over things such as, whether Texas A&M's home-court advantage in San Antonio will help the Aggies knock off Ohio State (yes); if UCLA's season-ending choke job should be considered a distress signal (not really); and how much you're interested in seeing an encore performance of the Jo Noah Boogie (as long as Lundquist joins in).

As much as all of that stuff -- location, momentum, dance moves, etc. -- matters, my most basic tenets of bracket-picking involve the laws of efficiency. I did a post at the outset of last year's Tourney Blog titled "Defending Your Bracket," which was reasonably successful in pinpointing the attractive offensive teams that were flawed on D -- and therefore risky bets in the dance. As the chart below shows, in the Modern Era of kenpom.com stats (2004-07), only one team outside the top 75 in adjusted defensive efficiency (sweet-shooting West Virginia, in '05) has reached the Elite Eight, and zero teams outside the top 25 in ADE have reached the Final Four.

Profile Of The Elite Eights
Off. Eff.
(Nat'l Rk.)
Def. Eff.
(Nat'l Rk.)
FloridaNC119.4 (2)87.2 (5)
UCLARU113.0 (28)85.1 (3)
LSUFF109.9 (50)85.7 (4)
George MasonFF110.0 (49)90.0 (18)
MemphisEE112.9 (29)87.4 (6)
TexasEE118.8 (4)88.7 (10)
UConnEE119.2 (3)89.9 (16)
VillanovaEE117.0 (11)89.9 (17)
North CarolinaNC122.1 (1)85.1 (7)
IllinoisRU120.0 (3)85.5 (11)
LouisvilleFF118.0 (6)87.8 (16)
Michigan St.FF118.0 (7)89.3 (25)
KentuckyEE111.4 (25)85.0 (6)
WisconsinEE110.6 (28)86.2 (12)
ArizonaEE116.3 (9)90.3 (34)
West VirginiaEE113.6 (12)94.6 (83)
ConnecticutNC115.7 (8)83.0 (3)
Georgia TechRU111.1 (27)83.7 (5)
DukeFF118.7 (2)83.5 (4)
Oklahoma St.FF116.3 (5)85.8 (9)
St. Joseph'sEE115.1 (9)85.8 (10)
KansasEE111.2 (23)86.5 (14)
XavierEE112.3 (17)87.9 (19)
AlabamaEE112.0 (18)92.2 (49)

Based on the data above, it behooves you to limit the number of defensively challenged teams you put in your Elite Eight and Final Four. When my SI.com Writer Bracket comes out on Tuesday, it'll only have one school outside the top 50 in defensive efficiency winning more than two games: Texas. And the 'Horns, at No. 59, are barely outside of that range.

It's easy to become infatuated with entertaining offensive teams that don't play D -- Gonzaga's Adam Morrison squads were the best example -- but recent history shows that poor defensive efficiency numbers are a red flag, no matter how well a school scores. Below is a list of the teams (from '05 and '06) that ranked in the top 25 in offensive efficiency but outside the top 75 in D. You'll notice that most of them didn't last too long in the dance:

2005's defensively challenged 1-6 seeds:
Team            Off. Eff       Def. Eff.     Out
(Seed) (Nat'l Rk.) (Nat'l Rk.) Rd. Out

LSU (6) 112.1 (22) 94.7 (84) 1
Gonzaga (3) 115.5 (10) 97.0 (119) 2
W. Forest (2) 120.8 (2) 94.0 (76) 2
The main attractions here were Brandon Bass, Morrison and Chris Paul -- all of whom could light up a scoreboard. None of them made it past the first weekend.

2006's defensively challenged 1-6 seeds:
Team               Off. Eff.     Def. Eff.
(Seed) (Nat'l Rk.) (Nat'l Rk.) Rd. Out

Gonzaga (3) 117.6 (8) 100.7 (162) S16
Boston Coll. (4) 118.0 (6) 96.8 (100) S16
Mich. St. (6) 115.6 (14) 95.0 (85) 1
Tennessee (2) 116.6 (11) 94.8 (82) 2
Oklahoma (6) 113.8 (20) 94.8 (81) 1
West Va. (6) 116.3 (13) 94.6 (75) S16
This club included Morrison (again), Jared Dudley and Craig Smith, Shannon Brown, Chris Lofton and Kevin Pittsnogle. Two suffered first-round exits. Only West Virginia outperformed its seed expectation, reaching the Sweet 16 as a No. 6. None of them won more than twice.

The intriguing thing about the 2007 bracket -- and perhaps this is a result of smarter seeding by the selection committee -- is that only one school seeded 1-6, Vanderbilt, falls into the top-25 offense, below-75 defense category. There isn't a statistically fraudulent No. 1, 2 or 3 seed in the field. There are, however, four teams seeded either No. 7 or 8 that fit our profile, and each one has a talented scoring star:

2007's defensively challenged 1-8 seeds:
Team               Off. Eff.     Def. Eff.
(Seed) (Nat'l Rk.) (Nat'l Rk.)

Nevada (7) 116.8 (21) 99.7 (124)
BYU (8) 116.4 (23) 97.4 (101)
Boston Coll. (7) 118.9 (11) 96.7 (92)
Vanderbilt (6) 116.4 (22) 96.2 (81)
Arizona (8) 120.4 (7) 95.4 (76)
If you were thinking about going out on a limb and forecasting a Nick Fazekas-led sleeper run into the Elite Eight, I strongly urge you to reconsider. Nevada does seem like an attractive option at first glance; the Wolf Pack dominated the WAC for much of the year, and have three potential NBA players (Fazekas and guards Marcellus Kemp and Ramon Sessions), but they've been keeping a little secret in Reno: their D is suspect. While the teams in this group aren't guaranteed to be doomed -- I think Arizona and Vandy have one win in them apiece -- their potential as giant-killers is limited.

Since we're dealing with such a small sample size (just three seasons) of efficiency data, the conclusions we're making are admittedly unscientific. I wouldn't be surprised if Texas, a young team that got exponentially better in the backstretch of the Big 12, crashes the party in Atlanta with a defense that doesn't exactly fit the Final Four mold. But when it comes to picking a national champion, it's impossible to ignore that the profile of the past three winners includes a top-10 ranking in both offensive and defensive efficiency.

In the grid below, I've broken down the Nos. 1, 2 and 3 seeds from the 2007 bracket into four flights: "A" teams currently have national-champ efficiency resumes; "B" teams are great in one category, and right on the fringe in the other; "C" teams are another step down, and the "D" teams are unlikely to win the title. It's not surprising that three 1-seeds are in Flight A, take a look at Texas A&M: the third-seeded Aggies appear better-suited to cut down the nets in Indy than any of the No. 2s.
Team               Off. Eff.     Def. Eff.
(Seed) (Nat'l Rk.) (Nat'l Rk.)

Flight A
UNC (1) 123.4 (2) 84.1 (2)
Texas A&M (3) 120.4 (8) 85.9 (8)
Ohio St. (1) 121.7 (5) 86.2 (10)
Florida (1) 122.6 (3) 86.5 (11)

Flight B
Kansas (1) 118.0 (14) 81.1 (1)
Georgetown (2) 123.6 (1) 88.4 (17)

Flight C
Wisconsin (2) 117.2 (20) 84.9 (5)
UCLA (2) 117.6 (18) 86.7 (12)

Flight D
Memphis (2) 116.2 (24) 86.1 (9)
Pittsburgh (3) 118.9 (10) 90.5 (26)
Oregon (3) 118.8 (12) 93.0 (49)
Wash. St. (3) 111.8 (61) 89.7 (22)
Two Flight A teams will be meeting in my bracket's title game. To find out which ones ... you'll have to check back tomorrow.

Labels: , ,

posted by Luke Winn | View comments |


Posted: March 12, 2007 8:17 PM   by Anonymous Chris

Thank you for pointing out something that many fans of the Big 12 already know...Texas A&M AND Kansas are great defensively...Kansas is 2nd in the nation in defensive efficiency and the analysts on ESPN are saying UCLA will trump them with THEIR defense?! Is it possible that in the effort to take the "east coast bias" out of college basketball, the sports media has overcompensated and gone all the way to the west coast, skipping the midwest?

Don't mean to ruin the suspense of waiting until tomorrow, but it appears your title game matchup will be Florida vs. Texas A&M. Florida is on the opposite side of the bracket of the other Flight A teams, and I don't think you would pick North Carolina with Pyscho T sporting the facemask the entire tournament

Using the post-season stats from prior years skews your predictions because teams that win in the tourney move up in Pomeroy's rankings. UCLA and Florida were 8th and 7th last year before the tourney, and they rose to 2 and 1 before the final game. Mason rose significantly during the tourney. Kansas was 2nd and fell to 6th by losing to Bradley and watching others pass it by reaching the Elite 8 and beyond. The list goes on and on.

If you can find Pomeroy's pre-tourney ratings, those would be much more useful in determining whether the ratings have held up during the tourney, versus teams rated lower rising to the top of the rankings by playing well in March.
I know they have to deal with a tough draw and the psycho-mask, but I was surprised that none of the experts picked UNC to get to the final four. After their finish in the ACC tourney and looking at these numbers they look pretty strong to me.
Much better use of stas, Luke. Your post on consistency was completely flawed! Florida ended up being "inconsistent" after a 17 winning run!
Try that consistency formula on the Dallas Mavs today, and they will come as inconsistent, only because they lost one game after a long run of victories.
By the way, I don't see anybody playing as well as the Gators now.

I really don't think there is much difference between the teams in your A, B, and C flights (and yes, I am a UCLA fan, but that doens't completely bias me).

You've ranked the teams into flights on the basis of their rankings which artifically inflates the minor differences in their ratings.

For example, over the course of a typical game (with an average number of possessions) against best offense this year (Georgetown with an offensive efficiencies of 123.6), the defensive efficiency rating for the best defense in your A flight (UNC, with a rating of 84.1 and a ranking of 2) predicts that they will give up 70.3 points. In a similar game, the 12th ranked Def. Eff.team (UCLA, the worst defense in your C flight with a rating of 86.7) is predicted to give up 72.4 points. That's a difference of only 2.1 points.

Likewise, the best offense in your A-flight (UNC, with a rating of 123.4) is predicted to score 67.7 points in an average game against this year's best defense (Kansas at 81.1). The worst offense in your C flight (Wisconsin, ranked 20 with a rating of 117.2) is predicted to score 64.3 points, a difference of only 3.4 points.

You can argue in high-pressure important games that 2 or 3 points is significant, but I suspect that random factors are much more important than that. Really, by the ratings (not the rankings), only a basket or two separates the best of A flight from the worst of C flight. Only the artificial separation caused by the rankings is what makes them look farther apart. And aren't the seductive lies of rankings why we turned to statistics in the first place?
Alex said he will torment Duke (a.k.a. Luke Winn) for the rest of his life, both physically, but mainly in a psychological sense if he does not bow down and recognize the triumph of the Buckeye Nation that will proceed over the next three weekends. Being that Alex is impartial and has finished in the top-10 in the March Madness pools for the the past 10-years, Luke should recognize Alex's prowpess in picking the champ. Oden is going to come out of his shell these next few weeks. As some bygone rock-band once said, "you ain't seen nothing yet...duhn-duh....ba-ba-ba-baby, no you ain't seen nothing yet"

I'm surprised, given the focus on Kenpom's ratings, that I haven't seen anyone pick up on the Luck and Consistency ratings as potential predictors yet. Of the two, luck, WHICH ACTUALLY MEASURES SKILL IN WINNING CLOSE GAMES, is the more important. In 2006, 40% (8 teams) of the top 20 got into the tournament (7 in 2007), of which 6 won their first game and 3 were in the Sweet 16. This makes for an interesting hypothesis for 2007, especially for the better teams in this mix (UNLV, GW).

Also, on luck, note that no top 21 team in 2006 with a luck rating less than -0.025 (Kansas, Illinois, Pitt, Georgetown, South Carolina, Arkansas, West Virginia, Arizona) made it into the Elite 8 (and many unexpectedly lost in the first or second rounds). This offers another interesting hypothesis regarding the fates of UNC, Texas A&M, Duke, Georgia Tech, Louisville, Kentucky, Villanova, Indiana.

Consistency is also not entirely irrelevant: last year's #1 (Gonzaga, #41 overall, Sweet 16) and this year's #1 (SIU, #35 overall) are both very dangerous outside picks. The fact that Northwestern St. (#2 in 2006, #105 overall) won its first round game over Iowa (#19 overall) in 2006 raises an interesting question about New Mexico St. (#2 in 2007, #96 overall) who play Texas (#16 overall) in round one (especially given Texas's marginal -0.019 luck rating).

With respect to Geoffrey's comments, I did a quick calculation, summing the offensive and defensive rankings (a very crude measure of a teams combined offensive and defensive prowess in comparison to other teams).

In 2006, the top 10 teams had ranking sums of: 7, 14, 31, 19, 18, 40, 28, 30, 35, 54.

In 2007, the top 10 teams had ranking sums of: 4, 15, 14, 15, 16, 18, 25, 30, 33, 42.

The results suggest that Geoffrey is right that there is a great deal of parity among the best teams this year. They also suggest that the best teams are even better than the rest of the field, this year, than they were last year (when overall kenpom.com ratings showed all 8 Elite 8 teams coming from the top 10).
I think the stats illustrating the necessity of defense to make it to the final four are poignant indeed. Based upon this logic, which teams were left out of the dance that probably should have been in and who should they have replaced?
That's a very interesting question. If you go over to kenpom.com/rate Ken Pomeroy has all the teams lined up. Then you can click on defensive efficiency and see them ranked by defense.

I forget exactly how many at large bids are available, but it's somewhere in the range of 30-35. So, that would suggest that your question asks whether any of the top 35 defensive teams were left out, and the answer is yes:

Connecticut (7)
Oklahoma (16)
Clemson (21)
Michigan (23)
Kansas St (25)
Syracuse (27)
Hawaii (30)
San Diego St (32)
Depaul (35)

Several of these teams were bubble teams.

Another interesting piece of information about these teams: six of these teams are in the NIT (except for the two highest rated, UConn and Oklahoma, and Hawaii), and all six won their first round game.

So why are these teams not in the NCAA tournament? The reasons vary but the big one is this: of them all, Syracuse has the best rated offense at #51 in the country. You can't win with defense alone.
The other part of your answer is a little more complicated. You have to take the at large bids and figure out which 8 had the worst defense:

Virginia #56 (#39, #47)
Texas #59 (#4, #16)
Xavier #63 (#15, #36)
Arizona #75 (#7, #23)
Vanderbilt #81 (#24, #52)
Texas Tech #85 (#58, #66)
Boston College #92 (#10, #41)
BYU #98 (#22, #54)

I've put their offensive rankings and overall rankings in that order in parentheses. At this point it becomes a judgment call about who should get in for whom being left out. To me, Virginia, Texas Tech, Vandy, and BYU look most vulnerable, followed by Boston College and Xavier. But there are good reasons to include many of these teams in. We shall see how they survive.
Posted: March 15, 2007 1:38 PM   by Anonymous Anonymous
Gotta love Boston College sending Bobby Knight home. Great pick Luke.
Posted: March 15, 2007 5:15 PM   by Anonymous Anonymous

I hate to burst the VCU bubble, but you can't pick against Duke before the sweet 16. They may be a flawed team that will go down once they play a talented power, but Mike will have them coached and ready for the first two rounds as they make yet another 16 for 10 in a row.
Reporting From ...
Contact The Blog
The Blog Pool
Recent Posts
The Style Archive
divider line