The NCAA Tournament bracket has been announced! Lets take a look at how accurate my bracketology was. I have some thoughts below this list!
Correct Picks: (41 of 68)
1 – Alabama, Kansas, Purdue, Houston
2 – UCLA, Arizona, Texas, Marquette
3 – Gonzaga, Baylor
4 – None
5 – None
6 – TCU
7 – Missouri, Michigan State,
8 – Memphis
9 – Auburn, Illinois
10 – Boise State, Utah State, USC
11 – Providence, Mississippi State, NC State, Nevada
12 – Drake, VCU, Charleston, Oral Roberts
13 – Iona, Louisiana, Kent State, Furman
14 – Grand Canyon, UC Santa Barbara
15 – Vermont, UNC Asheville
16 – Northern Kentucky, Southeast Missouri State, Howard, Fairleigh Dickinson, Texas A&M-Corpus Christi, Texas Southern
TOO HIGH (Correct in parentheses)
3 – UCONN (4)
3 – Tennessee (4)
4 – Duke (5)
4 – San Diego State (5)
5 – Texas A&M (7)
5 – Iowa State (6)
6 – Northwestern (7)
8 – West Virginia (9)
8 – Florida Atlantic (9)
8 – Maryland (9)
9 – Penn State (10)
14 – Colgate (15)
14 – Princeton (15)
TOO LOW (Correct in parentheses)
4 – Kansas State (3)
4 – Xavier (3)
5 – Virginia (4)
5 – Indiana (4)
6 – Miami (5)
6 – Saint Mary’s (5)
7 – Creighton (6)
7 – Kentucky (6)
9 – Arkansas (8)
10 – Iowa (8)
15 – Kennesaw State (14)
15 – Montana State (14)
BUBBLE TEAM INCORRECTIONS:
11 – Oklahoma State (Out)
11 – Rutgers (Out)
Out – Pittsburgh (11)
Out – Arizona State (11)
THOUGHTS:
The 4-8 seed range destroyed me, but that’s what I was expecting. These ranges had little disparity in overall rating number, and inconsistencies in the selection committee tend to show up here that my ratings system can’t take into account. For example, Texas A&M seemed heavily punished for having two quadrant 4 (bad) losses, however Iowa wasn’t punished at all.
Meanwhile, UCONN ranked 8th in NET and 4th in Kenpom, but received a 4-seed while Xavier (22 in NET, 16 in Kenpom) shockingly received a 3-seed. The quality of wins and losses for both teams was relatively equal, and there was no mathematical case for Xavier receiving a higher seed than UCONN. We know the selection committee takes a look at NET and Kenpom rankings, however they seem to devalue those rankings in random instances, making an accurate formula all but impossible.
It’s always interesting to look at inconsistencies within conferences as well, the main one being Duke and Virginia in the ACC. Duke ranked higher in NET (16 vs 27), Kenpom (21 vs 34), had a similar win quality (5-6 vs 5-5 in Quad 1 games with no bad losses), equal non-conference strength of schedules, and Duke came off a controlling win over Virginia to win the ACC Championship. Duke is a team on the upswing with a lot of momentum, however Virginia managed a 4-seed with Duke getting a 5-seed. I would love to know what caused the selection committee to make this decision, as I have no idea what else to take into account to improve the formula. However, as I’ve discovered in my years of doing this, if there is something in Virginia’s favor that could give them a higher rating than Duke that could be taken into account, the formula would become even more inaccurate for other teams.
I seemed to undervalue the ACC as a whole outside of Duke as well, and that’s the other tricky part is valuing conference strength. The committee will seed teams dependent upon conference affiliation as much as anything else (see Florida Atlantic), and I instituted conference adjusters to try to accurately reflect this. However, there is no fact-based number that I can turn to and it’s completely subjective to personal opinion.
At this point, I’m viewing the formula as much of a predictive tool as anything else. I had higher ratings on Penn State and Texas A&M than the consensus on bracketmatrix.com and both teams performed well in conference tournaments. Doing some quick math, my bracket graded out at 345 points, which makes this my 2nd best year out of my 6 years of participation. However, the average score the past two years has been 346-347 points, and with a stricter eligibility criterion this season, that average is likely to be above 350, making this grade a failing one. I will review the data to try and optimize the formula with previous years to try to find something that would’ve graded highly the past two seasons with hope to find some more consistency, but as I’ve discovered, a formula that’s more accurate for one season will be wildly inaccurate for another no matter what is taken into account. This showcasing the seedings that likely have a set of context that can’t be reflected mathematically, but also shows the inconsistency of the selection committee from a year to year basis.
