OFF SEASON DIVERSION: NCAA Tournament Simulation

[The Bofa on the Sofa]

OK, since we are in the lull between the end of the season and the spring schedule, I have been killing time by setting up an NCAA tournament simulation. I am going to use Pablo rankings to create match simulations, and then carry out a double elimination tournament of all the NCAA teams. I have noted here many times that this is what I would like the NCAA to do, so I set it up on my own. In the coming days, I will be posting summaries of the sub-regionals, regionals, and ultimately, the finals. Details of how I am doing things are below. Everything I am doing is using after the season (Dec 18) Pablo rankings.

1) a) There are 32 sub-regionals, hosted by the top 32 seeds. Each sub-regional has at least 8 teams and those teams are distributed throughout the top 256. Thus, the #2 seed in a sub-regional is from the group from 33 - 64, the #3 seed is from 65- 96, #4 seed is from 97-128, etc, through the top 8.

b) If there are more than 8 teams in a sub-region (there can be up to 12), then the 9 seeds and above are taken from those ranked over 256 without a lot of concern for their actual ranking.

c) As much as possible, sub-regionals are regionally created. I tried not to change any team's seeding in order to do it (and I only did it once, moving Georgia Tech up one notch, but that was because I screwed up in the original and didn't want to have to do it all over again). However, complete regionalization is impossible as long as I insist on having the top 32 teams hosting (which I do). Therefore, there are a boatload of lower seeded teams flying out of the east going to the west. My response to their complaints is to note that if east teams actually start getting competitive teams, then they will be able to get more host sites on the east coast. There are 4 sub-regions where I was mostly able to fill them up with local teams: Penn St, Texas, Duke, and Florida. (I had to bring Middle Tennessee to Duke for a #2 seed)

d) I tried to avoid putting a lot of the same conference in the same sub-regions, but that was a secondary consideration to regionalization. There are some cases, though, where the top two teams in a sub-region are from the same conference. However, given the double elimination tournament, this is less of a concern.

2) Each sub-region plays a double-elimination style tournament, with the two remaining teams (one in the winner bracket, one in the loser bracket) advancing to the Regional, which is held at the top 8 overall seeds. The sub-regional matchups are based on overall seeding (1-32, 2-31, etc)

3) The top 16 from the Regionals (8 from the winners, 8 from the losers bracket) will advance to the Finals, held at a neutral site.


Thus, the tournament is being held over the course of 3 weeks. The first week and Finals sessions are Four days, and the second week sessions are 3 days. I will describe more about Regionals and Finals when the time comes, but the Sub-regionals are broken down as the following:

Day 1: Preliminaries (for sub-regions of more than 8 teams), opening round of 8 teams, followed by the first set of elimination matches. At the end of the first day, there are 8 teams remaining. Depending on how many are in the sub-regional, that could mean as many as 12 matches on the day (obviously there are multiple courts playing)

Day 2: Second round elimination matches, Winner's bracket semi-finals, and the third elimination round. That is a total of 6 matches for the day, leaving 4 teams remaining

Day 3: The finals of the winner's bracket, with the loser facing the winner of the 4 elimination round. The winner of the Final, and the winner of the consolation match both advance to the Regionals in the following week.

Over the course of the next week, I will be posting the results from the sub-regionals, broken down by Region. I will usually only post highlights from Day 3, unless there is something interesting that happens on an earlier day.

[The Bofa on the Sofa]

Here are the top 32 seeds, the teams hosting the Sub-regionals

1 Nebraska
2 Stanford
3 UCLA
4 Washington
5 Texas
6 Penn State
7 Oklahoma
8 Minnesota
9 USC
10 Wisconsin
11 California
12 Hawaii
13 Ohio State
14 Iowa State
15 San Diego
16 Utah
17 Florida
18 Missouri
19 Colorado
20 Cal Poly
21 Pepperdine
22 Long Beach State
23 Kansas State
24 Arizona State
25 Michigan State
26 Northern Iowa
27 Colorado State
28 New Mexico State
29 Duke
30 Oregon
31 Santa Clara
32 Sacramento State

[The Bofa on the Sofa]

REGION 1: Nebraska
Sub-Regionals: Nebraska Sacramento State Utah Florida

NEBRASKA SUB-REGIONAL
Nebraska
Creighton
Kansas
South Dakota State
Tulane
IUPUI
William & Mary
Sam Houston State
Southern
Grambling State

Highlights: Nebraska easily rolls over Creighton (19, 11, 21) to move to the regionals. #4 seed Tulane pulls a 4 game win over South Dakota State in the other early match (30-25, 18-30, 30-27, 30-18). Then, Tulane pulls off another upset, this time beating Creighton in 4 games (30-28, 30-22, 23-30, 30-27). Amazingly, Tulane had barely squeaked by Kansas on Day 2 (31-29, 23-30, 29-31, 30-24, 19-17) just to stay alive for Day 3.

Winner: Nebraska
Consolation: Tulane

SACRAMENTO STATE SUB-REGIONAL
Sacramento State
Saint Marys
Georgia Tech
Oregon State
Fairfield
Rice
Birmingham-Southern
Central Connecticut State
Campbell

Highlights: Not a lot of local teams in this sub-region - only St Marys. Oh well, Sacramento St beats St Marys in the finals in 4 games, but St Marys beat Fairfield in 4 games in the consolation match. Fairfield had pulled out a 5 game upset over Oregon St in the early elimination match.

Winner: Sacramento State
Consolation: St Marys

UTAH SUB-REGIONAL
Utah
BYU
Alabama
Weber State
Utah State
Maryland
New Hampshire
Stony Brook
Holy Cross

Highlights: Not a lot of options here, so Utah and BYU end up in the same sub-regional. Utah controls BYU in the final (24, 16, 27), but BYU sweeps by Alabama in the consolation (23, 18, 24)

Winner: Utah
Consolation: BYU

FLORIDA SUB-REGION
Florida
Florida International
Jacksonville State
Florida State
Miami
UNF
Jacksonville
UCF
Stetson
Bethune-Cookman
Alabama State
Savannah State

Highlight: Most of the state of Florida is here for jam-packed action. Florida beats #2 seed Florida International (29, 21, 24), who then pulls out a 5 game win over Jacksonville St, winning 15 - 11 in game 5. Not a single upset the entire weekend.

Winner: Florida
Consolation: Florida International

[The Bofa on the Sofa]

REGION 2: Minnesota
Sub-Regionals: Minnesota Michigan State USC Arizona State

MINNESOTA SUB-REGION
Minnesota
Notre Dame
Western Kentucky
North Dakota State
Eastern Kentucky
Bradley
George Washington
Austin Peay
DePaul
Chicago State

Highlights: Lots of midwest teams here, but Minnesota and Notre Dame are the cream of the bunch. Minnesota beats the Irish in the final (30-18, 30-19. 27-30, 30-17), and Notre Dame beats #5 seed Eastern Kentucky in the consolation (27-30, 30-23, 30-20, 30-16). Eastern Kentucky had pulled off a couple of 5 game upsets over Western Kentucky and North Dakota St to get to Day 3.

Winner: Minnesota
Consolation: Notre Dame

MICHIGAN STATE SUB-REGIONAL
Michigan State
Ohio
IPFW
Western Michigan
Eastern Michigan
Central Michigan
Kent State
Butler
Oakland
Indiana State

Highlights: A good set of local teams. The NCAA is happy. Michigan St and Ohio make the finals as expected, but Ohio upsets the Spartans on their home court, winning in 3 (24, 28, 21). Michigan St regroups, though, and sweeps #3 seed IPFW 32-30, 30-22, 30-16.

Winner: Ohio
Consolation: Michigan St

USC SUB-REGION
USC
UCSB
Cal State Fullerton
Delaware
Princeton
Wake Forest
Syracuse
St. Francis-PA
Alabama A&M
Marist


Highlights: USC has no problems, sweeping their way to the win, beating UCSB 27,24,26 in the final match. UCSB then sweeps away CSFullerton (16, 21, 25) to win the consolation.

Winner: USC
Consolation: UCSB

ARIZONA STATE SUB-REGION
Arizona State
Arizona
Belmont
South Carolina
Texas State
Northern Arizona
Louisiana-Lafayette
Brown
Iona


Highlights: Arizona vs Arizona St in the finals, and ASU wins in 4 (21-30, 31-29, 33-31, 30-17). Arizona handily beats an overmatched South Carolina team for the second spot.

Winner: Arizona State
Consolation: Arizona

[GatorVball]

You lost all credibility with me ranking Florida 17. Good idea, but 17? On what planet?

[The Bofa on the Sofa]

Please read the first post:
"Everything I am doing is using after the season (Dec 18) Pablo rankings"

Did you forget that Florida got blown out at home by Minnesota, who finished #8? Given that, it is not a surprise at all that they dropped to 17.


I don't seem to remember your name as one of those who does as well as Pablo in picking winners. So, given your biased opinion, and Pablo, I'm going with Pablo.

[GatorVball]

I have never entered the pick em contest. I read that it's based on the Pablo, but Pablo has Hawaii and Utah ahead of Florida, 2 teams the Gators beat. And Oklahoma, Iowa State, San Diego, Ohio State ahead of them? Come on. Pablo is flawed. Minnesota didn't blow out FLorida. They beat them in 4. That means Florida falls to 17? Get real.

I appreciate what you are trying to do, a playoff like this would be great, but you'd be better off using RPI. When Florida isn't one of the top 16 teams, I have a hard time taking anything else seriously.

[The Bofa on the Sofa]

gatorvball said:

I have never entered the pick em contest. I read that it's based on the Pablo, but Pablo has Hawaii and Utah ahead of Florida, 2 teams the Gators beat. And Oklahoma, Iowa State, San Diego, Ohio State ahead of them? Come on. Pablo is flawed.

Blah, blah, blah.

It also has Florida ahead of South Carolina, a team they lost to.

[The Bofa on the Sofa]

REGION 3: WASHINGTON
SUB-REGIONS: Washington Duke Ohio State Cal Poly

WASHINGTON SUB-REGION
Washington
Eastern Washington
Gonzaga
Montana
Boston College
Idaho State
Binghamton
Coastal Carolina
Louisiana-Monroe
Nicholls State

Highlights: Washington rolled, along more than 21 points in a game only once, in game 3 against EWU in the finals. EWU swept Boston College, who had eliminated #3 seed Gonzaga on day 2, to win the consolation.

Winner: Washington
Consolation: Eastern Washington

DUKE SUB-REGION
Duke
Middle Tennessee
Winthrop
North Carolina
East Carolina
Appalachian State
UNC Greensboro
Gardner-Webb
North Carolina A&T
Winston Salem State
The Citadel
South Carolina State

Highlights: Middle Tennessee was the only team from not right nearby, and they make the most of it, upsetting Duke in the finals in 5 (15 - 12 in game 5). Duke regrouped to beat Winthrop in 4 games to win the consolation. Recall that Middle Tennessee beat Louisville in the first round of the real NCAA tournament, so they aren't bad.


Winner: Middle Tennessee
Consolation: Duke

OHIO STATE SUB-REGION
Ohio State
Cincinnati
Dayton
Xavier
Cleveland State
Toledo
Wright State
Youngstown State
Akron
Coppin State

Highlights: Ohio State beats Cincinnati in the finals by scores of 30-22, 28-30, 30-26, 30-22. Cincinnati had made the finals with a 5 game win over Dayton in the semis. However, Dayton turns the tables in the consolation match, winning 26-30, 30-24, 30-21, 18-30, 15-11.

Winner: Ohio State
Consolation: Dayton

CAL POLY SUB-REGION
Cal Poly
LSU
Clemson
San Diego State
Furman
McNeese State
Southern Miss
UC Riverside
NC State
High Point

Highlights: LSU wins a tough 4 game match against Cal Poly to win the final, 30-25, 23-30, 36-34, 31-29. Cal Poly recovers to beat Clemson in 4 games to win the consolation.

Winner: LSU
Consolation: Cal Poly

[Phaedrus]

p-dub, how do you determine who wins the match? I know that one of the refs who lives in St. Louis did a probabilistic simulation using the team stats to create a match simulation program based on the statistical match ups in the 1990's in order to see how the matches would turn out if we went to the rally scoring format. I know the USAV used his results to make their decision on going with the rally format.

[The Bofa on the Sofa]

phaedrus said:

p-dub, how do you determine who wins the match? I know that one of the refs who lives in St. Louis did a probabilistic simulation using the team stats to create a match simulation program based on the statistical match ups in the 1990's in order to see how the matches would turn out if we went to the rally scoring format. I know the USAV used his results to make their decision on going with the rally format.

I am using Pablo rankings to determine point probabilities, and then simulate the matches from there, applying a 0.05 sideout correction. Talking to people who know these things, they tell me that 0.05 is about the right correction. Thus, if a team scores. .53 of the points overall, it ends up being .48 when they are serving, 0.58 when the opponent serves. Matches are simulated point by point.

Scores depend on how far separated the teams are. For example, Washington vs Eastern Washington, in Seattle, ends up with scores 30-18, 30-21, 30-28. Meanwhile, teams that are closer like Dayton and Cincinnati end up being a lot closer (they went 5 twice, but that is not always going to happen).

It was when I started doing this type of simulation that I discovered that Pablo ratings were linear with respect to point percentage.

[Phaedrus]

pablo said:

phaedrus said:

p-dub, how do you determine who wins the match? I know that one of the refs who lives in St. Louis did a probabilistic simulation using the team stats to create a match simulation program based on the statistical match ups in the 1990's in order to see how the matches would turn out if we went to the rally scoring format. I know the USAV used his results to make their decision on going with the rally format.

I am using Pablo rankings to determine point probabilities, and then simulate the matches from there, applying a 0.05 sideout correction. Talking to people who know these things, they tell me that 0.05 is about the right correction. Thus, if a team scores. .53 of the points overall, it ends up being .48 when they are serving, 0.58 when the opponent serves. Matches are simulated point by point.

Scores depend on how far separated the teams are. For example, Washington vs Eastern Washington, in Seattle, ends up with scores 30-18, 30-21, 30-28. Meanwhile, teams that are closer like Dayton and Cincinnati end up being a lot closer (they went 5 twice, but that is not always going to happen).

It was when I started doing this type of simulation that I discovered that Pablo ratings were linear with respect to point percentage.

I am not sure what this other guy did for correction but that sounds pretty close to what he was doing in his simulation.

So what is the sigma between Pablo and point percentage?

Interesting stuff.

[The Bofa on the Sofa]

phaedrus said:

pablo said:

I am using Pablo rankings to determine point probabilities, and then simulate the matches from there, applying a 0.05 sideout correction. Talking to people who know these things, they tell me that 0.05 is about the right correction. Thus, if a team scores. .53 of the points overall, it ends up being .48 when they are serving, 0.58 when the opponent serves. Matches are simulated point by point.

Scores depend on how far separated the teams are. For example, Washington vs Eastern Washington, in Seattle, ends up with scores 30-18, 30-21, 30-28. Meanwhile, teams that are closer like Dayton and Cincinnati end up being a lot closer (they went 5 twice, but that is not always going to happen).

It was when I started doing this type of simulation that I discovered that Pablo ratings were linear with respect to point percentage.

I am not sure what this other guy did for correction but that sounds pretty close to what he was doing in his simulation.

Actually, I think I know who you are talking about. It's the same guy who I talked to about my sims.

So what is the sigma between Pablo and point percentage?

Based on 100s of thousands of simulated matches (actuially more than a million overall), Pablo ratings are as perfectly linear in point percentage as you can get between point pcts of .4 and .6. I think it also works outside that range, but you need too many matches to get statistical significance.

Of course, that is what happens if you have a very large data set with well controlled conditions. The normal NCAA season isn't nearly large enough, and there is a lot of scatter expected. However, I have been able to basically reproduce that scatter and give a result that more or less mirrors actual data. It's not easy, though (there's a huge schedule effect that needs to be accounted for).

[IdahoBoy]

I also saw Oklahoma blown out by UCLA, USC defeated by Hawaii in the round of 8, but isn't reflected as much as Florida's loss to Minnesota.

[The Bofa on the Sofa]

1) Home
2) The overall bad performance by the SEC probably didn't help, either. The SEC overall went way down.