Sunday, March 31, 2019

Final Four per the MDS Model (2019)

Some decent upsets this weekend have left us with only the best team (per my model) remaining: Virginia, which had a 16.33% chance to win it all at the start, now is up to better than 2 in 5 odds.


SeedTeam1st RoundChampionshipChampion
Final Four
2Michigan St.100.00%57.98%28.75%
3Texas Tech100.00%42.02%17.58%
1Virginia100.00%70.88%42.49%
5Auburn100.00%29.12%11.18%

No other 1 seed made it though! So Virginia will have to carry the banner for the 54% combined favorite that all 1 seeds encompassed going into the Sweet 16.

Potential area of concern for Viginia fans: I've only picked the champion correctly twice going into the Final Four, and both times it was Villanova.

Monday, March 25, 2019

Sweet 16 per the MDS Model (2019)

I caught some flack for my model being so chalky going into the tournament, and well, here we are:


SeedTeamElite 8Final FourChampionshipChampion
Sweet 16
1Duke67.55%40.85%23.65%13.52%
4Virginia Tech32.45%13.91%5.58%2.20%
3LSU32.63%11.01%3.91%1.37%
2Michigan St.67.37%34.23%18.13%9.48%
1Gonzaga66.97%39.31%21.62%12.03%
4Florida St.33.03%13.65%5.20%2.01%
3Texas Tech45.10%20.03%8.78%3.90%
2Michigan54.90%27.01%13.13%6.46%
1Virginia82.75%52.68%33.34%19.30%
12Oregon17.25%4.77%1.32%0.31%
3Purdue44.18%17.47%8.12%3.33%
2Tennessee55.82%25.07%13.11%6.12%
1North Carolina64.01%37.41%18.68%9.42%
5Auburn35.99%15.95%5.79%2.12%
3Houston40.75%16.80%6.13%2.25%
2Kentucky59.25%29.84%13.52%6.18%

Virginia is the only team with better than a 50/50 chance of making the Final Four at this point, and no team has even a 1 in 5 chance of winning it all. Interestingly, they are also the only team to see their odds improve from prior to the tournament. The chalkyness of results has left Duke, Carolina, and Gonzaga relatively unchanged. Even so, there is a 54% likelihood that one of the 1 seeds ends up winning it all.

Tuesday, March 19, 2019

"What are the odds?" Duke and Carolina Will Meet in the NCAA Tournament, and When

Duke/Carolina is arguably the fiercest rivalry in all of college sports. I can understand if you choose The Game (since they did fight an actual war at one point), but results this century don't encompass a compelling argument:



And between North Carolina and Duke, the two schools rank #3/#4 in all-time wins, have 48/42 NCAA tournament appearances, 20/16 Final Fours, and 11/11 championship game appearances (all figures are respective). So it's fairly surprising that we've never met in the NCAA tournament (we met once in the NIT, in 1971).

The committee never puts us in the same region, so the Final Four would be the earliest meeting. And even for the best teams, it's more likely than not that they won't make the Final Four. So this proposition is definitely a long shot, but not impossible; it almost happened in 1991 in the title game. So when is it expected to happen?

I started by gathering the seedings for both us and Dook over the past 20 seasons (we also missed the tournament 3 times...):


AverageMedianModeMaxMinStandard Dev
Carolina321182.53
Dook1.921161.33

We've had more variance, but the most likely seed for both teams is a 2. We also have been on the same side of the bracket only 5 times in this period, or 25% of the time (in other words, 25% of the time we could have met in the Final Four, and the rest of the time would have been in the title game).

Next, I drew a fairly deep sample of the odds a generic 1 through 8 seed makes the Final Four using the past 5 seasons of my model:



SeedProb FF
128.93%
223.73%
312.51%
410.59%
57.67%
64.34%
72.98%
82.20%

Note:
I threw out 
Kentucky's region from 2015. Talk about an outlier: 78.5% to make the Final Four!

So now for the math on an outcome I certainly hope to never see; I actually agree with Coach K, who stated in '91 "that he never wanted the two schools to play in the last game of the season because it would be too devastating for the loser." (Chansky, Blue Blood: Duke-Carolina

Monday, March 18, 2019

The Bump (or Lack Thereof) From Winning Your Conference Regular Season

Stemming from my earlier post on whether teams see a bump from winning their conference tournament, I have received demands to follow this up with the same analysis but for conference regular season titles.

I can now recreate the chart from the last post with both types of conference titles considered:

YearChampionConferenceConf TourneyConf Reg Season>= 1Both
2018VillanovaBig East1010
2017North CarolinaACC0110
2016VillanovaBig East0110
2015DukeACC0000
2014ConnecticutAAC0000
2013LouisvilleBig East1111
2012KentuckySEC0110
2011ConnecticutBig East1010
2010DukeACC1111
2009North CarolinaACC0110
2008KansasBig 121111
2007FloridaSEC1111
2006FloridaSEC1010
2005North CarolinaACC0110
2004ConnecticutBig East1010
2003SyracuseBig East0110
2002MarylandACC0110

8/17 won their conference tournament, 11/17 won their conference regular season, 15 /17 won at least 1 of these, and 4/17 won both.

I can see how far each team progressed similar to last time as well:


FinishCount%Cumul %
Missed11.0%100.0%
First Round54.9%99.0%
Second Round2221.6%94.1%
Sweet 161817.6%72.5%
Elite 82524.5%54.9%
Final Four1312.7%30.4%
Championship87.8%17.6%
Champion109.8%9.8%
Total102100.0%100.0%