Here is an itemized list of the model results. Obviously picking all these upsets in the first round is not a smart strategy. Utah State has failed in this model for 3 years. Clemson you have to question playing late Tuesday and then early on Thursday.
We wanted to get these out there so we can discuss them after the games.
Our First Round Statistical Upsets:
(12) Utah State over Kansas State
(12) Clemson over West Virginia
(11) Gonzaga over St. John’s
(11) Marquette over Xavier
(10) Michigan State over UCLA
(10) Florida State over Texas A&M
(9) Illinois over UNLV
(9) ODU over Butler
Our Conditional First Round Upsets:
(12) Richmond over Vanderbilt
(11) Gonzaga over St. John’s
(11) Missouri over Cincinnati
(10) Michigan State over UCLA
(9) Tennessee over Michigan
Note: Conditions nullify the Illinois-UNLV upset
Our Second Round Statistical Upsets
Our Second Round Conditional Upsets(5) K State over Wisconsin
Need 2 double digit seeds from Gonzaga, Marquette, Missouri, Clemson, and Michigan State
Our Third Round Statistical Upsets
(3) Syracuse over (2) North Carolina
(3) BYU over (2) Florida
Our Third Round Conditional Upsets
(3) UConn over San Diego State
(4) Wisconsin over (1) Pitt – realize this doesn’t correspond with 2nd round conditional
Our Fourth Round Statistical Upsets
Our Fourth Round Conditional Upsets
(1) Select at least 3 teams that have been to a Final 4 since 1985
(2) Pick no team that gets more than 30.5% of their points from 3 (Louisville, ND, Wisconsin)
(3) Pick Duke or two teams from one conference
(4) Make sure the seeds add up to between 7 and 13
The model likes Ohio State over Kansas, but the Irving coming back to Duke, all of this could change. Also, be aware of Purdue having some issues on their team.
What isn’t a rule in our models, but could be with one more year of data?
BracketAnalytics.com is constantly re-evaluating our rules and with each tournament, we get new data to evaluate our rules. Right now there are two rules that are close to being added: Road Record and 3 point dependency.
You have probably heard me talk a lot about three point dependency. It is an excellent indicator of cumulative advancement, but it is not a good indicator in the early rounds because teams get hot (think Cornell last year).
The bottom line is that these rules are not yet statistically significant and passing our tests, so they don’t go in the model.
Here is a miniature version of what teams may be vulnerable because of these almost rules.
Clemson over UAB
Clemson was in the tournament last year, UAB was not. Clemson has a higher Points Per Possession (PPP) ratio than UAB. UAB relies way too heavily on the three-point FG. Clemson has a stronger Strength of Schedule.
Southern Cal over VCU
Souther Cal has a higher PPP ratio than UAB. VCU relies way too heavily on the three-point FG. USC has a stronger Strength of Schedule.
UTSA over Alabama State
UTSA has a better PPP ratio. They have a better Strength of Schedule. Alabama State is dreadful on the free throw line.
UNC-Ashville over Arkansas-Little Rock
This game will be dreadful and I am the least confident in this one. UNC-A has a higher PPP ratio, but there are no other indicators hitting on model.
Commentary on the Model and the Bracket
First thing first. We have an ethical responsibility to disclose any conflicts of interest. We are in Richmond, Virginia, and one of the co-founders went to two of the schools in the tournament and a spouse went to two of the schools so we want to reveal those schools to you. Those schools are Virginia Commonwealth University, University of Richmond, Old Dominion, Connecticut, Pitt, and Penn State.
Now on to the brackets. We feel the committee did a great job SEEDING the teams, but could have done better SELECTING the teams. Simply picking Virginia Tech over Georgia would have improved the brackets exponentially. Last year, 24 teams were mis-seeded according to our models, this year it is considerably less (more later on this).
Utah State. Utah State has our models fooled. The models like them, but the eyeball test does not. They were in a similar position last year with Texas A&M and just didn’t compete.
When you are using the Interactive Bracket Builder, I would start by picking all the low risk teams in the first round, and then team’s without the “Lower Expected Point Production Risk”, then hit the Seed Evaluator button. If your distributions show up as yellow or red, investigate the “Not in Last Year’s Tournament” risk, and then hit the Seed Evaluator again. If you still have some red or yellow, here is our prioritized list of upsets by seed. * represents schools that our teams has a relationship with (disclosure above, BOLD INDICATES THE MODEL PICKS AN UPSET WITHOUT CONDITIONAL FILTERS
9 Seeds1. Old Dominion*
1. Michigan State
2. Florida State
3. Penn State*
2. Utah State
3. Morehead State
Our Final 4 Short List has some major changes
Our ‘Green List’ has gotten smaller, and our ‘Yellow List’ has had some names drop off and a perennial power jump back on. Let’s see who is peaking at the right time and who isn’t.
Georgetown, West Virginia, BYU, and UConn have fallen off our Final 4 short list, which isn’t surprising. The Huskies have disappointed time and time again as of late. They haven’t been the same team since they beat Villanova in mid-January. BYU and Georgetown you can attribute to suspensions and injuries, respectively. West Virginia has not has the bench strength, literally and figuratively , to be part of the equation this season.
Our Green List consists of Ohio State, Kansas, Duke, Texas, and Syracuse. Duke is the weakest team on the Green List due to three point dependence, but they are on the right side of the line for now. Syracuse is the weakest team in terms of net points per possession.
Kentucky is the strongest team on our ‘Yellow List’ as our doubts persist as they dance from one side to the other of our road performance and three point dependence thresholds. Florida is the new team on our ‘Yellow List’.
Remember, we feel 3 teams will come from our ‘Green List’ and one team will come from our ‘Yellow List’ based on tracking this over the past 12 years. We have one more week for this to change, so let’s see what happens.