Sunday, March 30, 2014

Final Four by the Aggregate Model

Basically the path has been cleared for Florida to win the whole thing. Which, if the tournament so far has been any indication, means someone else will pull it off. This is when I default to my (non-sports fan) friend's saying: "You can't predict sports!"

But I'll give it a shot anyway:


SeedTeamChampionshipChampion
South-1Florida71.08%47.46%
East-7Connecticut28.92%13.03%
West-2Wisconsin56.67%23.89%
Midwest-8Kentucky43.33%15.62%

Wednesday, March 26, 2014

Seeding DOES Matter: In the Short-run

I've referenced this article by KenPom on why seeding doesn't matter multiple times in some recent posts.

About a month ago, before the field was set, the UNC Band Truck tweeted this:
It's true, seeding doesn't really affect a team's overall odds of winning it all. But as the tweet indicates, it does affect a team's chances of getting to the next round in the short-run, and subsequently, their chances of getting to the next weekend. From my (albeit biased) perspective, this has major implications: seeding could be the difference between whether I (and the rest of the band, team, and cheerleaders) get to travel to both San Antonio, TX and New York City, or just the first/second round site.


North Carolina received a 6-seed in this year's tournament, while some pundits/talking heads/fans expected us to receive as high as a 4-seed. (I personally think the 6-seed was fair. We were 22nd in KenPom's ratings going into the tournament, predicted to be a 6 per the Bracket Matrix, and quite frankly, getting dropped to the 6 instead of the 5 was beneficial: we had to play Iowa St in the second round as opposed to Michigan St, AND the 12-seed of Harvard would've been a tougher draw than 11-seeded Providence). So let's assume we could've gone anywhere between a 4 and a 7.

I calculated North Carolina's odds of advancing from each seed, going up against the "average" opposing seed in the first two rounds:

Seed7-seed6-seed5-seed4-seed
Prob Round 3258.71%55.20%60.96%68.51%
Prob Sweet 1626.78%28.02%27.18%38.05%

These probabilities ascend with the seed, with the exceptions of the 6-seed in the first round and the 5-seed in the second. This is due to the relative strengths of the possible opposing 11-seeds (Tennessee, Iowa) and then the possible opposing 4-seeds (Michigan St, Louisville) in this year's tournament. So once again, matchups trump seeding: but being a higher seed generally does have a positive impact on your chances of making it to the second weekend.

What if the NCAA Tournament was Reseeded?

Each year, the first round of the NCAA tournament offers multiple upsets, and this year was no different, with 7 double-digit seeds winning in the first round. As KenPom has pointed out multiple times, seeding doesn't matter, but matchups do. So consider the question: what if the NCAA tournament was reseeded after each round, like the NFL playoffs, the NHL playoffs, and some conference tournaments (such as the NEC)?

Monday, March 24, 2014

Sweet 16 by the Aggregate Model

Two rounds in, and we're at the Sweet 16. In my personal bracket I picked Creighton to win it all, and we know how that worked out. I was (un?)lucky enough to see both that and North Carolina's chances end in person, in the same building, in San Antonio. Not the best day.

The Aggregate Model is performing the best out of my three brackets (beating the MDS Model and my personal picks (so long McDermott)), so here are the updated probabilities for each team. Louisville is still the overwhelming favorite. Also, Tennessee at 2%! This is a team that was one of the last ones in to the tournament and they have a 1/50 chance of winning the whole thing!

Tuesday, March 18, 2014

Bracket Math for the MDS Model

I finally finished my own MDS Model for college basketball, and got it done just in time to predict the bracket. Since it wasn't in action for any of the regular season, I have no baseline to measure its performance against. Here goes nothing!


Bracket Math for the Aggregate Model

In all its glory, the Aggregate Model is finished for the NCAA Tournament. It is actually fairly similar to Nate Silver's composite (but was created independently, and in my case, originally implemented for college football). Mine is a central mean (calculated the same way as Jeff Sagarin's) of the following models:

KenPom
Kenneth Massey
Sagarin's Predictor
Statfox
Dokter Entropy
Sonny Moore

These were chosen based on personal preference and their performance throughout the season (per the Prediction Tracker). So far in 2013-14, this Aggregate Model has picked 74.35% outright winners for NCAAB.


Monday, March 17, 2014

ACC Tournament Projections after Semis

A little late, I know, but the final was exactly as predicted, Virginia vs Duke. Which was as absolute close to 50/50 as you could get, 0.9112 (Virginia) vs 0.9111 (Duke), so Virginia was the slight favorite (and won the title).

FinalSeedTeam2nd RoundQuartersSemisFinalChampion
1Virginia100.00%100.00%100.00%100.00%50.00%
8Maryland100.00%0.00%0.00%0.00%0.00%
9FSU100.00%100.00%0.00%0.00%0.00%
4North Carolina100.00%100.00%0.00%0.00%0.00%
5Pitt100.00%100.00%100.00%0.00%0.00%
12Wake Forest100.00%0.00%0.00%0.00%0.00%
13Notre Dame0.00%0.00%0.00%0.00%0.00%
2Syracuse100.00%100.00%0.00%0.00%0.00%
7NC St100.00%100.00%100.00%0.00%0.00%
10Miami FL100.00%0.00%0.00%0.00%0.00%
15Va Tech0.00%0.00%0.00%0.00%0.00%
3Duke100.00%100.00%100.00%100.00%50.00%
6Clemson100.00%100.00%0.00%0.00%0.00%
11Ga Tech100.00%0.00%0.00%0.00%0.00%
14Boston College0.00%0.00%0.00%0.00%0.00%
Totals128421

NCAA Tournament Odds are coming soon! Both the MDS Model for NCAAB is set, as well as the (better performing) Aggregate Model!

Friday, March 14, 2014

ACC Tournament Projections after Qtrs

Well, Syracuse Clemsoned, Clemson Clemsoned, Pitt tried unsuccessfully to Clemson, and Virginia steamrolled FSU.

Now Duke is your new predicted champion thanks to their easy matchup with State. Damn it Clemson.

QtrsSeedTeam2nd RoundQuartersSemisFinalChampion
1Virginia100.00%100.00%100.00%69.09%34.55%
8Maryland100.00%0.00%0.00%0.00%0.00%
9FSU100.00%100.00%0.00%0.00%0.00%
4North Carolina100.00%100.00%0.00%0.00%0.00%
5Pitt100.00%100.00%100.00%30.91%15.45%
12Wake Forest100.00%0.00%0.00%0.00%0.00%
13Notre Dame0.00%0.00%0.00%0.00%0.00%
2Syracuse100.00%100.00%0.00%0.00%0.00%
7NC St100.00%100.00%100.00%18.62%9.31%
10Miami FL100.00%0.00%0.00%0.00%0.00%
15Va Tech0.00%0.00%0.00%0.00%0.00%
3Duke100.00%100.00%100.00%81.38%40.69%
6Clemson100.00%100.00%0.00%0.00%0.00%
11Ga Tech100.00%0.00%0.00%0.00%0.00%
14Boston College0.00%0.00%0.00%0.00%0.00%
Totals128421

Thursday, March 13, 2014

ACC Tournament Projections after Day 2

The odds of the favorites (teams with byes) winning the title continue to drop as the favored teams (in this model) in the first two rounds all won (with the lone exception of Wake Forest over Notre Dame). As these teams move into the later rounds, they take away from the byes as their respective chances of winning the title increase.

If Georgia Tech had been able to pull out the OT victory over Clemson, Duke would be the new favorite to win it all, taking yet another ACC Tournament title 27.97% of the time.

It will be interesting to see how things shake up if FSU, NC State, or Clemson can pull off the large upset tomorrow (UNC-Pitt is basically a coin flip).

Day 2SeedTeam2nd RoundQuartersSemisFinalChampion
1Virginia100.00%100.00%78.27%53.21%26.61%
8Maryland100.00%0.00%0.00%0.00%0.00%
9FSU100.00%100.00%21.73%8.07%4.03%
4North Carolina100.00%100.00%52.43%20.84%10.42%
5Pitt100.00%100.00%47.57%17.89%8.94%
12Wake Forest100.00%0.00%0.00%0.00%0.00%
13Notre Dame0.00%0.00%0.00%0.00%0.00%
2Syracuse100.00%100.00%75.67%36.32%18.16%
7NC St100.00%100.00%24.33%5.95%2.98%
10Miami FL100.00%0.00%0.00%0.00%0.00%
15Va Tech0.00%0.00%0.00%0.00%0.00%
3Duke100.00%100.00%80.84%51.73%25.87%
6Clemson100.00%100.00%19.16%5.99%3.00%
11Ga Tech100.00%0.00%0.00%0.00%0.00%
14Boston College0.00%0.00%0.00%0.00%0.00%
Totals128421

Wednesday, March 12, 2014

ACC Tournament Projections after Day 1

Day 1 is in the books, with Wake Forest pulling off the slight upset over Notre Dame. After Miami's poor performance in the 1st half and their seemingly cursed play against Va Tech, someone texted me "Miami must be built on a hokie burial ground". They managed to finally beat Va Tech though, which is the only team in the ACC in the bottom half of Division 1. Looks like firing Seth Greenberg worked out exactly as they hoped.

Not much changed for the teams that haven't played (yet).


SeedTeam2nd RoundQuartersSemisFinalChampion
1Virginia100.00%100.00%78.72%54.20%27.10%
8Maryland100.00%48.61%10.11%3.76%1.88%
9FSU100.00%51.39%11.17%4.30%2.15%
4North Carolina100.00%100.00%58.50%23.24%11.62%
5Pitt100.00%75.88%36.10%13.57%6.78%
12Wake Forest100.00%24.12%5.40%0.93%0.46%
13Notre Dame0.00%0.00%0.00%0.00%0.00%
2Syracuse100.00%100.00%76.82%36.53%18.27%
7NC St100.00%53.48%13.01%3.16%1.58%
10Miami FL100.00%46.52%10.17%2.24%1.12%
15Va Tech0.00%0.00%0.00%0.00%0.00%
3Duke100.00%100.00%83.32%53.27%26.64%
6Clemson100.00%62.19%11.92%3.73%1.87%
11Ga Tech100.00%37.81%4.76%1.06%0.53%
14Boston College0.00%0.00%0.00%0.00%0.00%
Totals128421

Monday, March 10, 2014

2014 ACC Tournament Projections

Using the ratings given by my Aggregate Model (composed of the following NCAAB models: KenPomMasseyJeff Sagarin7OTsCompughter, and Moore), I've projected the likelihood of each team winning the ACC tournament in two ways.
Note: All games are considered neutral site

First, here are the final ratings:


RankTeamRate
1Virginia0.9112
2Duke0.9111
3Syracuse0.8795
4North Carolina0.8350
5Pitt0.8212
6FSU0.7403
7Maryland0.7295
8Clemson0.7085
9NC St0.7012
10Miami FL0.6712
11Notre Dame0.6416
12Ga Tech0.5964
13Wake Forest0.5935
14Boston College0.5100
15Va Tech0.4641


Top 6 Conference Tournaments Picks

The Aggregate Model picks each conference tournament as follows:
ACC, Big 10, Big 12, Big East, Pac 12, SEC