Reviewing 2007 (a.k.a., Oh no they didn't!)
As I noted after the field was announced, last year's CTD oracle (or, Ron as I'll call it this year) performance was a mixed bag, Some misses (see Arkansas) couldn't have been predicted with an automated approach, and some came down to a close call between several teams.
But 2007, more than any year in recent memory, was the Year of the Committee Head Scratcher. Here's a look at the good, the bad, and the ugly from last year. One argument you will not see me use is how a team performed in the tournament. Anything can happen in a one-and-done situation.
Drexel seemed to be a problem for many bracketologists, but CTD correctly slayed the Dragons. ESPN's Andy Katz says about Drexel:
Drexel was done in more by the way it played in the Colonial than by how it performed out of conference. The committee celebrated the Dragons' nonconference road wins -- like Syracuse, Villanova and Creighton -- but going 1-5 against the top teams in the CAA (Old Dominion, VCU and Hofstra) essentially eliminated the Dragons.Indeed, CTD correctly weighed Drexel's #5 non-conference RPI against their #88 conference RPI along with the rest of their similarly hit (8 games over .500 on the road) or miss (2-4 vs. RPI top 50; #74 Sagarin ranking) resume.
Here, in geeky tabular form with equally geeky yet endearing commentary, is where I and/or CTD messed up:
I let in, they left out
|I let in, they left out||CTD||Web||Rank||Conf||NC||Road||L10||1-25||1-50||1-100||101+|
Ah yes, Syracuse. There wasn't much uproar about their omission, was there? While 97% of the Bracket Project participants included Syracuse, this was the least surprising of the three CTD missed in terms of the confidence (91%) assigned to the pick.
Missouri State and Air Force
CTD sure didn't get a whole lotta love from the wisdom of the bracketology crowds with these two. Despite the 99+% (Missouri State) and 96% (Air Force) confidence values assigned, they were only included on 27% (MSU) and 17% (AF) of the brackets surveyed for the Bracket Project.
You may remember that Missouri State was also a problem for CTD in 2006. I'm not sure I understand this one. Their RPI rankings (conf/non-conf and overall) were essentially top 40, and their Sagarin raking of 26 also supports at-large selection. However, they only had one RPI Top 25 win (Wisconsin) and records against RPI Top 50 and Top 100 were sub-.500.
Interestingly (I don't know whether it's intentional) the at-large split point for games +/- .500 against RPI Top 50 teams seems to be right at .500. Since 2000, 77% of the eligible (those with RPI rank 105 or better) teams with better than .500 records against such teams were (or we estimate would have been based on their seed) selected, while only 34% of the teams at or below .500 were selected.
I left out, they let in
|I left out, they let in||CTD||Web||Rank||Conf||NC||Road||L10||1-25||1-50||1-100||101+|
Again, here's Andy Katz:
Not all Sunday games are created equal. The committee had one spot open Sunday, and it was going to go to NC State, if it won the ACC tournament with a win over North Carolina, or to Arkansas. NC State lost, so the Hogs were in -- even though they were losing to Florida in the SEC tournament final while the selection committee was making its decisions.I can live with Arkansas (they were CTD's first team out), but it seems that making the SEC final was enough to get them in.
On the other hand, I can't live with Stanford. CTD interpreted their profile as less than a 1% match of an at-large selection.
One more time, the Katz writes:
Unbalanced scheduling is common in the high-majors (save the Pac-10), yet that's the reason some teams -- like Syracuse and Kansas State -- didn't get into the field... Unbalanced scheduling hurt those teams, but a true round-robin, and going 10-8 in the Pac-10, actually helped Stanford, according to (selection committee chairman Gary) Walters.RPI rankings for conference games only: Syracuse 32, Kansas State 35, Stanford 49. So Stanford benefitted simply because they had a true round-robin schedule? They had a worse non-conference RPI than the supposedly dreadful Syracuse and they were demolished by Air Force.
The favorite team of toddlers everywhere, the Red Raiders were our third team left out, so I guess I'm also OK with them. Only 14 teams had more than their 3 wins over RPI Top 25 teams (though Syracuse also had 3), but their 3 games under .500 against RPI Top 100 and rough finish (6-7) didn't help.
Big seed misses
|Big seed misses||Seed||Rank||Conf||NC||Road||L10||1-25||1-50||1-100||101+|
UNLV (too low) and Virginia (too high) seem to attract the biggest complaints in the seeding department. In most of these cases, the Web consensus agrees that Virginia, USC and Vandy were overseeded, and UNLV was underseeded. Several employees of the Worldwide Leader also agreed:
The following teams were measurably underseeded in the 2007 Tournament: UNLV, Niagara, UCLA, Marquette, Texas, Nevada, Creighton. And the following teams were demonstrably over-seeded: Gonzaga, George Washington, Texas Tech, Long Beach State, Virginia, Albany, Stanford. (Joe Lunardi - ESPN Insider required)
Virginia is a 4-seed. Texas also is a 4-seed. Huh? ... Look closer at Virginia and you will see that this team is supremely lacking in its qualifications for such a high seed... Texas, on the other hand, makes sense; the Horns very well could be underseeded at 4... If the 4-seed is supposed to be better or more worthy than, say, a 7-seed such as UNLV, I must have missed something. (Doug Gottlieb - ESPN Insider required)
Virginia, USC and Vandy
We seem to have overestimated the importance of non-conference RPI and underestimated the importance of conference RPI and good wins with these three. While each had 4 wins over RPI Top 25 teams, most of their other attributes didn't support their 4 (Virginia), 5 (USC) and 6 (Vandy) seeds. Notably, each had bad non-conference RPIs (100 or worse) and Sagarin rankings not in line with their seeds. While their conference RPIs were nearly correlated with their seeds, so the committee sure didn't apply that consistently to UNLV (#14 conference RPI = #7 seed line?).
It seems like one factor that we don't consider may have been part of it. Virginia won 7 games against RPI Top 50 teams. (CTD only considers games +/- .500 against RPI Top 50 teams.) Only 14 teams won 7 or more such games, so that attribute supports their #4 seed, as does does their record of 3 games over .500 against the same teams (only 13 teams did so). However, with most of their other attributes not supporting a #4 seed, it looks like the committee is selectively applying their principles. That is not only difficult to predict, but unfair to the teams.
Their RPI (#27) and non-conf RPI (#25) support a higher seed than #12, but their single win over an RPI Top 25 team, 5 games under .500 against RPI Top 50 teams and 3 games under .500 against RPI Top 100 teams, and conference RPI (#47) don't. Sounds like more arguments against than for CTD's #8 seed prediction, so I (and the Web consensus) tend to agree the committee was closer.
The committee gave UNLV a #7 because... Anyone? Anyone? Yeah, they blew this one. Most writers (see above), the Web consensus and all but the Sagarin ranking had UNLV much higher than the #7 the committee gave them. I still don't understand this one. Wow.
|Top 2 seed lines||Seed||Rank||Conf||NC||Road||L10||1-25||1-50||1-100||101+|
|Ohio St.||B10||30-3||1 (#3)||1 (#1)||16.00||1||2||1||1||1||17||20||4||10||5||7||12||0|
|UNC||ACC||28-6||1 (#2)||1 (#2)||15.96||1||1||7||2||7||9||3||2||7||6||8||13||0|
Not much about this one, just that Florida makes absolutely no sense as the #1 seed overall, and the mainstream and non-mainstream media agree. Not to mention the #2 seed given by the Web consensus. I think the committee violated their principle that neither previous season performance nor potential should enter into selection or seeding.
Admittedly, it was a pretty tight race this year. Several complications with the Sunday games likely also had an effect. I couldn't find any links either way, but with Ohio State and Wisconsin playing in the Big Ten championship (the game started at 3:30 ET), the committee my have decided to slot the winner as a #1 and loser as a #2 to minimize bracket complications. (To my credit, I did note this hours before the brackets were unveiled.)
ESPN's Joe Lunardi basically ripped the committee a new one: " was, quite simply, the worst seeding/pairings performance we've seen in my 13 years of Bracketology." Wow. Seriously, if you have an ESPN Insider subscription, Lunardi's review of the committee's performance since 1995 is a thorough and accurate (though slightly self-back-patting) analysis.
Let's hope for a less ugly 2008.