NAQT Prognostication Time

Old college threads.

How close do you think this prediction will be to the real ICT invitations (with reserved bids and host bids taken out of course)?

Poll ended at Tue Feb 10, 2004 10:35 pm

Close
5
28%
Not Close
13
72%
 
Total votes: 18

pakman044
Rikku
Posts: 262
Joined: Sun Nov 23, 2003 3:56 am
Location: Carrboro, NC
Contact:

NAQT Prognostication Time

Post by pakman044 »

Third edit: I have received the Southeast stats, but since they were done in Stats 99, it'll take me a little bit of time to reformat them into the formats I've been using (most of my Excel worksheets have come from the Stanford program outputs)

It's time for everyone's favorite pasttime until bids come out...that's right: ICT Bid Projections.

So for the sectionals I have, I calculated *MY* s-value on this metric:

s = (1/3)(win_pct + TCU + bonus_average / 30)

The TCU is equal to the number of powers times 15 plus the number of regulars times 10 plus the number of negs times -5. That sum is divided by the TUH stat (tossups heard) times 10. This is to reward people who get powers.

There are a few teams that get shafted on this, and I think a 9-4 team somewhere got pushed to the far bottom. It does not take into account schedule strength or section strength, and is vulnerable to small regions. (If you look at the UCLA D2 s-value and Berkeley Funky D1, you'd understand).

In the Canada sectional, I did not count the tiebreaker in stats, but did count it in winning pct.

The regions I have encoded so far: Northeast (NE), North Central (NC), West (W), Southwest (SW), Mid Atlantic (MA), Great Lakes (GL), Canada (CA), and Northwest (NW).

So with all of this in mind, here is my output. I won't interpret it for suspected outliers until I have all the results and which divisions the hosts will take their bids (if they can!):

Image
Image
Let the debates begin!

Patrick King[/img]

Edit: Mid-Atlantic Stats now are same as rest of stats.
Edit2: Northwest added to list.
Last edited by pakman044 on Tue Feb 10, 2004 2:59 pm, edited 1 time in total.
mujason
Wakka
Posts: 166
Joined: Fri Aug 15, 2003 12:47 am
Location: Missouri

Post by mujason »

Here's the formula I used last year for my predictions (and yes, it does include strength of schedule).

(Bonus conversion) + 50*(power per tu heard) + 25*(winning percentage) + .5(opponent bonus conversion) + 25*(opponent power per tu heard).

If I have time, I'll calculate these values and/or modify the formula.

Jason Mueller
mujason
Wakka
Posts: 166
Joined: Fri Aug 15, 2003 12:47 am
Location: Missouri

Post by mujason »

I just thought of something else for your formula, Patrick. For strength of schedule, you could include something like half of the opponents' TCU and bonus conversion. (Compensating for opponents' winning percentage wouldn't work, of course, as each sectional field has a winning percentage of .500 and would actually hurt the better teams because the rest of the field would have a lower winning percentage).

Jason Mueller
pakman044
Rikku
Posts: 262
Joined: Sun Nov 23, 2003 3:56 am
Location: Carrboro, NC
Contact:

Post by pakman044 »

I could, but unless there's a specific statistic for that in the releases, that would either require recursion and a computer program or lots of spare time (I built this in Excel).

Patrick King
mujason
Wakka
Posts: 166
Joined: Fri Aug 15, 2003 12:47 am
Location: Missouri

Post by mujason »

Patrick,

I'm going through and doing the strength of schedule improvement to your formula by brute calculation. For simplification, I'm assuming a round-robin for each sectional.
pakman044
Rikku
Posts: 262
Joined: Sun Nov 23, 2003 3:56 am
Location: Carrboro, NC
Contact:

Post by pakman044 »

Well, that would work except that a) some sections didn't have round robins, b) some had round robins and playoffs, c) some had doubles/triples/quadruples/quintuples??

But I'll take a look at what you get after your brute force, and if I think it looks fairly reasonable, I'll use it (fairly reasonable in that some team...say UCLA...doesn't have the hardest schedule strength. No hard feelings against them, but usually going 15-0 doesn't say *that* much on behalf of your opposition [other than the fact that they got steamrolled])

Patrick King
mujason
Wakka
Posts: 166
Joined: Fri Aug 15, 2003 12:47 am
Location: Missouri

Post by mujason »

Here are the results from Division I of the SW Sectional using Patrick's formula with a strength of schedule adjustment.
The new formula is: P=(win% + TCU + (1/30)*PPB + .5*OPPTCU + (1/60)*OPPPPB)/4. I'll calculate the rest of the sectionals (and maybe Div II) later.

A&M Grad 0.632980897
Rice UG 0.488788539
Okla A UG 0.437061966
Truman UG 0.413467369
A&M UG 0.409772316
Texas B UG 0.396431284
Texas A 0.351424309
Okla B 0.339572849

Jason Mueller
User avatar
QuizbowlPostmodernist
Wakka
Posts: 115
Joined: Wed Aug 13, 2003 12:22 am

Post by QuizbowlPostmodernist »

According to NAQT's host reporting requirements at http://www.naqt.com/official-reporting-results.html:
....the host must report full team statistics for each team; this consists of: division, institution, team name, team designation ("A", "B", etc.), coach name, coach (or representative) contact information, games won (noting number of wins by forfeit), games lost (noting number of losses by forfeit), total points scored, total power tossups, total non-power tossups, total interrupts, total tossups heard, and final order of finish. In addition, the host must submit a brief, but complete, description of the format used including the field's division into brackets and the results of any tie-breaker matches or criteria used to decide play-off spots or rebracketing.
NAQT is the black box. The list of ICT bids is the output. I don't know the NAQT formula, but I would guess that it's most likely similar to qb rating or BCS standings, where you create columns of various numbers and add them together to get a final statistic. Probably it involves bonus conversion, percentage of tossups answered, percentage of tossups powered, and the like.

Just quit making up random factors and saying you have S-values.

--Anthony
NotBhan
Rikku
Posts: 306
Joined: Tue Dec 16, 2003 12:30 pm
Location: Parts Unknown

Post by NotBhan »

If you (Patrick and Jason) need a couple more stats to pump directly into your vein, I can give you some unofficial stats for UF from the southeast SCT. My stat sheet has us with 41/161/28 on tossups, 6130 total points in 14 matches, a 14-0 record, 299 tossups heard, and 20.02 points per bonus.

Again, that's strictly off of my running scoresheet, so the stats compiled by UGa will probably differ slightly. I don't think we'll benefit from the emphasis on powers in your formula, though -- we were sitting back a good bit against the less experienced teams, figuring it's better to get 10 and a bonus than to pick up a stupid neg.

--Raj Dhuwalia, UF
"Keep it civil, please." -- Matt Weiner, 6/7/05
pakman044
Rikku
Posts: 262
Joined: Sun Nov 23, 2003 3:56 am
Location: Carrboro, NC
Contact:

Post by pakman044 »

I'm still waiting for southeast, but seeing that I was there and know how much h!@@ went on with the missing teams and the facility practically shutting down during rounds 14 & 15, I'm not surprised that the stats are slow. Heck, I'm tired and have gotten sick to boot! ;-)

Yeah, the S-value I posted is just my interpretation of what NAQT says. It's probably wrong, but it's something to keep myself occupied with. If no one was interested in prognostication, then why are there all sorts of sportswriters trying to project the NCAA field (and seedings!) now?

Jason, I can send you the Excel spreadsheets, so you don't have to do it all completely by hand if you want.

Patrick King
mujason
Wakka
Posts: 166
Joined: Fri Aug 15, 2003 12:47 am
Location: Missouri

Post by mujason »

Patrick,

I'm actually using Excel, also. And I don't mind doing the math; I don't have any homework due until Thursday and doing the math is fun and keeps me fresh.

Jason
samer
Lulu
Posts: 75
Joined: Wed Jun 18, 2003 3:01 pm

Post by samer »

From the NAQT web site:
S-values
NAQT does not release the formula for the S-value, but it combines statistical measures of tossup performance and bonus performance and includes corrections for opponent strength. The final result is adjusted by overall finish so that it is rare for teams to be invited ahead of those that finished ahead of them at their Sectional.

It is always better to win, even if this puts you into a higher (and harder) playoff bracket, and it is always better to convert a tossup, given the opportunity, and to convert bonus points, given the opportunity. Tossup statistics are computed per tossup heard (and bonus statistics per bonus heard), so there is no benefit to running up the score by trying to rush through a huge number of tossups near the end of a game.
samer dot ismail -at- gmail dot com / Samer Ismail, PACE co-founder, NAQT editor
Jaberwocky
Kimahri
Posts: 2
Joined: Tue Feb 10, 2004 8:05 pm
Location: Rochester, NY
Contact:

Post by Jaberwocky »

Based on the NAQT website and what they say they use, one thing I would change in your calculation is giving a heavier weight to Winning Percentage instead of giving it an equal weight with the other components.
pakman044
Rikku
Posts: 262
Joined: Sun Nov 23, 2003 3:56 am
Location: Carrboro, NC
Contact:

Post by pakman044 »

It is now in place!

Southeast results are now compensated by average tossups heard per game; actual values for TUH/game are available at http://www4.ncsu.edu/~paking/2004naqtsct-southeast-uga/ . This did shift a few teams around; most notably, Wofford (southeast) ahead of Pikeville (KCQRL) for last in in D2.

In D2, apparently the Berry v. UTC game in the southeast was contested, but the game is not in the results because the scoresheet wasn't found. Robin Richards (southeast TD) thinks that Berry won, but I have not changed the stats until there is more specific info. Based on my ratings, Berry is not very far down in the D2 ratings, and a win could raise their winning percentage to the point where they could break into the top 5 wait list (or higher).

So without further ado...the final ratings (unadjusted; Jason might have something to say about this):

D1:
Image
D2:
Image

Hosts. Unless a host has otherwise specified, if they attended the ICT last year, they will take a D1 team. If a host qualifies a D2 team, their host bid must be a D1 team [NAQT rules]; if the host qualifies a D1, they may NOT use their host bid [NAQT rules].

D1:
HOSTS [7]--
Western Ontario
DePauw
Princeton
Brandeis
Simon Fraser
Stanford
Arkansas

RESERVED [1]--
British Student Quiz Bowl Champion

AUTOMATIC (D1 Champions) [9]--
Rochester B (Canada)
Michigan A (Great Lakes)
Maryland A (Mid-Atlantic)
Yale A (New England)
Carleton (North Central)
British Columbia A (Northwest)
Florida A (Southeast)
Texas A & M Grad (Southwest)
Cal-Berkeley Funky (West)

AUTOMATIC (D1 UG Champions; if and only if section > 4 UG teams) [3]--
Swarthmore (Mid Atlantic)
Harvard A (Northeast)
Rice (Southwest)

AT-LARGE [12]--
1. Chicago A (Great Lakes)
2. Emory A (Southeast)
3. Yale B (Northeast)
4. Iowa State A (North Central)
5. South Florida A (Southeast)
6. Michigan B (Great Lakes)
7. Virginia (Mid Atlantic)
8. Illinois A (Great Lakes)
9. Berkeley Krusty (West)
10. MIT (Northeast)
11. Indiana A (Great Lakes)
12. Rochester A (Canada)

WAIT LIST:
1. Delaware (Mid Atlantic)
2. Georgetown (Mid Atlantic)
3. Minnesota (North Central)
4. Toronto A (Canada)
5. South Carolina (Southeast)

D2--
HOSTS [1]
Georgia

RESERVED [8]
Bevill-Jasper (Alabama CC)
Florida CC winner
South Georgia Gold (Georgia CC)
Hutchinson CC Blue (Kansas CC)
Marion Military Black (Mississippi CC)
Two more unnamed CC teams (as per NAQT tradition)

AUTOMATIC [10]
Carnegie Mellon A (Canada)
Chicago C (Great Lakes)
Penn A (Mid-Atlantic)
Harvard B (New England)
Carleton (North Central)
British Columbia C (Northwest)
Emory B (Southeast)
Kansas (Southwest)
UCLA (West)
Georgetown College (KCQRL)

AT-LARGE [13]
1. Illinois B (Great Lakes)
2. St. Thomas (North Central)
3. Pitt Baltimore CFL (Canada)
4. MIT B (Northeast)
5. Harvard C (New England)
6. Athens State (Southeast)
7. Columbia (Mid Atlantic)
8. Wichita (Southwest)
9. Virginia (Mid Atlantic)
10. Tulsa (Southwest)
11. Maryland (Mid Atlantic)
12. Swarthmore A (Mid Atlantic)
13. Wofford (Southeast)

WAIT LIST:
1. Pikeville (KCQRL)
2. Truman B (Southwest)
3. Florida B (Southeast)
4. Stanford Labe (West)
5. Case Western (Canada)

Lists can fluctuate, especially since we don't know how many CC teams will accept, we don't where hosts will put their bids, and we don't know where the KCQRL will stand in this. Expect one to two more teams out of the KCQRL to be on the D2 list.

Let the fun begin....

Patrick King
edit: forgot to turn on BBCode
Edit2: Fixed D1 Virginia Bavg. Added automatic bids for D1 UG champions
Edit3: Removed D1 UG champs that didn't have autos, added Rochester A, moved Arkansas to host-bid of D1.
Edit4: Added KCQRL SCT results.
Edit5: Added Illinois A to D1 prediction.
Edit6: Changed D2 Southeast TUH stats.
Last edited by pakman044 on Wed Feb 11, 2004 9:35 pm, edited 5 times in total.
User avatar
Dan Greenstein
Yuna
Posts: 848
Joined: Sun Apr 20, 2003 8:26 pm
Location: Takoma Park, MD

Post by Dan Greenstein »

Patrick, you left out that Undergraduate champions are automatic qualifiers.
User avatar
ezubaric
Rikku
Posts: 369
Joined: Mon Feb 09, 2004 8:02 pm
Location: College Park, MD
Contact:

Post by ezubaric »

pakman044 wrote: WAIT LIST
1. Wofford (Southeast)
2. Truman B (Southwest)
3. Stanford Labe (West) //this invite would force Stanford D1 host bid
4. Florida B (Southeast)
5. Case Western (Canada)
Since I know the West sectionals best, I thought I'd throw in my two cents.

This has Stanford Labe being invited before Berkeley, Stanford Incon, and Caltech; all finished ahead of Stanford Labe (something that the NAQT text says shouldn't happen). Stanford Labe was in a lower bracket and had fun with the lower-ranked teams. Perhaps adding something to track the final rank of the team would help?
Jordan Boyd-Graber
UMD (College Park, MD), Faculty Advisor 2018-present
UC Boulder, Founder / Faculty Advisor 2014-2017
UMD (College Park, MD), Faculty Advisor 2010-2014
Princeton, Player 2004-2009
Caltech (Pasadena, CA), Player / President 2000-2004
Ark Math & Science (Hot Springs, AR), Player 1998-2000
Monticello High School, Player 1997-1998

Human-Computer Question Answering:
http://qanta.org/
sobek22
Lulu
Posts: 18
Joined: Mon Feb 09, 2004 11:30 pm

Post by sobek22 »

I can already tell you that something's wrong here. Virginia simply was a better team than Georgetown. Both were involved in the four-way tie at 5-4, and both lost out. Virginia then went and beat Georgetown in the playoffs. More so, Virginia's bonus conversion was by a large margin the second best in the tournament -- closer to ours than to any other team's.

I'm not saying Georgetown was a bad team; they played admirably well at the SCTs (and led us mid-second half). I would switch Virginia and Georgetown's positions on that list, however.

-Adam
pakman044
Rikku
Posts: 262
Joined: Sun Nov 23, 2003 3:56 am
Location: Carrboro, NC
Contact:

Post by pakman044 »

UG champs are autos? Didn't know that. Now that I checked D1 rules, that is correct. Will reedit my post to reflect this.

Patrick King
User avatar
bucktowntiger
Lulu
Posts: 11
Joined: Thu Sep 11, 2003 10:39 pm
Location: Arkansas
Contact:

Post by bucktowntiger »

Arkansas will take its automatic bid in D1. Also, Undergrad teams win automatic bids only if there are at least four Undergrad teams in the sectional in which it participated.

Edit: Also, I think British Columbia does not get an automatic invite for D1 because there were not four teams in D1 at Northwest.

--Josh, P. t. altaica
Last edited by bucktowntiger on Thu Feb 12, 2004 5:48 pm, edited 1 time in total.
pakman044
Rikku
Posts: 262
Joined: Sun Nov 23, 2003 3:56 am
Location: Carrboro, NC
Contact:

Post by pakman044 »

sobek22 wrote:I can already tell you that something's wrong here. Virginia simply was a better team than Georgetown. Both were involved in the four-way tie at 5-4, and both lost out. Virginia then went and beat Georgetown in the playoffs. More so, Virginia's bonus conversion was by a large margin the second best in the tournament -- closer to ours than to any other team's.

I'm not saying Georgetown was a bad team; they played admirably well at the SCTs (and led us mid-second half). I would switch Virginia and Georgetown's positions on that list, however.

-Adam
Correct. I misenterred Virginia's Bavg, which should actually be 18.51. That would make a considerable difference, and will be corrected once I rerun my spreadsheet, fix the UG auto bid error, and do other fixing.

If anyone is significantly interested in this (ha ha), and see an error, post it and I will check and fix it if necessary.

Patrick King
sobek22
Lulu
Posts: 18
Joined: Mon Feb 09, 2004 11:30 pm

Post by sobek22 »

Patrick, you also may have accidentally skipped over Fred "The Truth" Bush's Rochester A when compiling the list of at-large teams.

-Adam
pakman044
Rikku
Posts: 262
Joined: Sun Nov 23, 2003 3:56 am
Location: Carrboro, NC
Contact:

Post by pakman044 »

sobek22 wrote:Patrick, you also may have accidentally skipped over Fred "The Truth" Bush's Rochester A when compiling the list of at-large teams.

-Adam
Yup. I skipped over them because I didn't think they had performed since Rochester B got the autos (another classic case of ghosting!). Will fix on the *NEXT* edit.

And that is correct about D1 undergrads having min of 4 per region. This means I will have to go back and check, but the only team I think that hurts is South Florida B.

Patrick
mps4a_mps4a
Lulu
Posts: 48
Joined: Wed Sep 03, 2003 12:42 am
Location: University of Virginia
Contact:

Post by mps4a_mps4a »

thanks for the post about Virginia, but we did lose a few too many times to make me feel comfortable in our getting a bid... and what's more, I don't know if we actually can get in ahead of VA Tech, Delaware, and Columbia because of the circle of death pain.

But on a more important note, when does NAQT actually announce the bids?
User avatar
einstein must die
Lulu
Posts: 21
Joined: Sun May 04, 2003 11:50 pm
Location: greenville, sc

Post by einstein must die »

during our mediocre performance in athens on saturday, we beat wofford. i hope that doesn't keep them from receiving an invitation to nationals...
in 1984 i was hospitalized for approaching perfection
-silver jews
samer
Lulu
Posts: 75
Joined: Wed Jun 18, 2003 3:01 pm

Post by samer »

ezubaric wrote:This has Stanford Labe being invited before Berkeley, Stanford Incon, and Caltech; all finished ahead of Stanford Labe (something that the NAQT text says shouldn't happen). Stanford Labe was in a lower bracket and had fun with the lower-ranked teams. Perhaps adding something to track the final rank of the team would help?
Not only shouldn't, but *cannot* happen: all four teams from the upper bracket *must* get invites before Stanford Labe can get an invite.
samer dot ismail -at- gmail dot com / Samer Ismail, PACE co-founder, NAQT editor
pakman044
Rikku
Posts: 262
Joined: Sun Nov 23, 2003 3:56 am
Location: Carrboro, NC
Contact:

Post by pakman044 »

einstein must die wrote:during our mediocre performance in athens on saturday, we beat wofford. i hope that doesn't keep them from receiving an invitation to nationals...
We almost beat Wofford, we were about 30 points away, and I gave it the old heave-ho with about 2 seconds remaining (I called a timeout with 13 to go to warn everyone of the time since my guys are newbies at the college quiz bowl).

Did you play for Furman A or B? You might've run into me or run of my teammates from NC State.

Patrick King
mujason
Wakka
Posts: 166
Joined: Fri Aug 15, 2003 12:47 am
Location: Missouri

Post by mujason »

Edit: I now used a new SOS for the Mid-Atlantic Sectional. It helped a little, but every little bit helps.
Another Edit: Now with the new SE TUH stats. SE teams improved.

Well, I finally got around to finishing the Div I stats. I still have a few Div II sectionals to do.

Rank Team P Sect
1 Berkeley Funky 0.745852695 W
2 Florida A 0.699454378 SE
3 Yale A 0.686877958 NE
4 Maryland A 0.686298849 MA
5 Michigan A 0.662673788 GL
6 A&M Grad 0.632980897 SW
7 Rochester B 0.613233672 CA
8 Chicago A 0.609972926 GL
9 Emory A 0.595677482 SE
10 Carleton 0.583730581 NC
11 Harvard A 0.58194199 NE
12 Yale B 0.580153735 NE
13 South Fl A 0.56628654 SE
14 Michigan B 0.565169192 GL
15 Berkeley Krusty 0.564773375 W
16 Illinois A 0.562187234 GL
17 Iowa State A 0.557941295 NC
18 Virginia 0.555485584 MA
19 MIT 0.551999263 NE
20 Indiana A 0.547795534 GL
21 Rochester A 0.536817657 CA
22 Swarthmore 0.532932042 MA
23 UBC A 0.526519856 NW
24 Delaware 0.515289733 MA
25 Georgetown 0.506503828 MA
26 Toronto A 0.493121153 CA
27 Rice UG 0.488788539 SW
28 Minnesota 0.488077115 NC
29 Virginia Tech 0.481217157 MA
30 South Carolina 0.47493445 SE
31 South Fl B 0.450485857 SE
32 SFU A 0.445904249 NW
33 Michigan C 0.437273192 GL
34 Oklahoma A UG 0.437061966 SW
35 Columbia A 0.42460092 MA
36 Caltech 0.420454752 W
37 Pitt Shreveport Pirates 0.41916681 CA
38 Truman UG 0.413467369 SW
39 A&M UG 0.409772316 SW
40 Northwestern A 0.400551733 GL
41 Texas B UG 0.396431284 SW
42 Rutgers 0.378473088 MA
43 Iowa State B 0.371701972 NC
44 Brandeis 0.364915607 NE
45 Williams 0.352367028 NE
46 Texas A 0.351424309 SW
47 Iowa 0.349166707 NC
48 Tulane A 0.347012679 SE
49 Oklahoma B 0.339572849 SW
50 Columbia B 0.325598711 MA
51 Chicago B 0.315061768 GL
52 Maryland B 0.308728568 MA
53 Vanderbilt 0.303455587 SE
54 USC 0.284093436 W
55 BC 0.28238664 NE
56 Grinnell 0.275046401 NC
57 McGill A 0.274063453 CA
58 Ohio State A 0.262026997 GL
59 Tennessee 0.253701487 SE

Feel free to debate. Of course, I understand if the results have glitches due to bracketed playoffs (such as the Mid-Atlantic sectional, now partly corrected).

Jason Mueller
Last edited by mujason on Wed Feb 11, 2004 10:57 pm, edited 2 times in total.
mujason
Wakka
Posts: 166
Joined: Fri Aug 15, 2003 12:47 am
Location: Missouri

Post by mujason »

Oh, one more thing. For the southeast sectional, I used 300 tossups heard, because I think Raj said Florida heard about 300 tossups. I'll correct the stats if the TUH stats are ever released.

Jason Mueller
pakman044
Rikku
Posts: 262
Joined: Sun Nov 23, 2003 3:56 am
Location: Carrboro, NC
Contact:

Post by pakman044 »

Okay, for *my* stats, I have pulled back the UG champs that did NOT have at least 4 UG teams. Mid-Atlantic, Northeast, and Southwest were left in; SW was the only one I could confirm that followed the at least 4 rule.

Rochester A was put in my predictions.

As for the person who said about Stanford Labe, I couldn't find an explicit NAQT policy about this, but a strength of schedule measurement would flush those out (which I haven't done, but Jason worked on).

KCQRL results added.

Patrick
Edit: added part about KCQRL. Didn't want to make another post.
Last edited by pakman044 on Wed Feb 11, 2004 2:14 am, edited 1 time in total.
User avatar
einstein must die
Lulu
Posts: 21
Joined: Sun May 04, 2003 11:50 pm
Location: greenville, sc

Post by einstein must die »

pakman044 wrote:
einstein must die wrote:during our mediocre performance in athens on saturday, we beat wofford. i hope that doesn't keep them from receiving an invitation to nationals...
We almost beat Wofford, we were about 30 points away, and I gave it the old heave-ho with about 2 seconds remaining (I called a timeout with 13 to go to warn everyone of the time since my guys are newbies at the college quiz bowl).

Did you play for Furman A or B? You might've run into me or run of my teammates from NC State.

Patrick King
i captained furman's A team. sorry if i came off as a little rude after we played you all, but i don't take losing very well :D
in 1984 i was hospitalized for approaching perfection
-silver jews
mujason
Wakka
Posts: 166
Joined: Fri Aug 15, 2003 12:47 am
Location: Missouri

Post by mujason »

As for the Stanford DII bracket discussion, my strength of schedule was based on a round robin, not bracketed playoffs. Although I do agree that teams that made the upper bracket should be seeded before the lower bracket teams, I wasn't able to correct it and I'm not even sure if I could correct it even with weighting the strength of schedule for actual opponents played.

Jason Mueller
User avatar
ezubaric
Rikku
Posts: 369
Joined: Mon Feb 09, 2004 8:02 pm
Location: College Park, MD
Contact:

Post by ezubaric »

pakman044 wrote:As for the person who said about Stanford Labe, I couldn't find an explicit NAQT policy about this, but a strength of schedule measurement would flush those out (which I haven't done, but Jason worked on).
I don't see how it would; Jason's proposed SoS measurement would reward all teams in the Western sectional for playing against difficult teams (since he's assuming a round robin). These bid predictions would still be impossible (as Samer pointed out, since it skips over three more qualified teams).
NAQT's Webpage wrote:NAQT does not release the formula for the S-value, but it combines statistical measures of tossup performance and bonus performance and includes corrections for opponent strength. The final result is adjusted by overall finish so that it is rare for teams to be invited ahead of those that finished ahead of them at their Sectional.
Here's a way to correct it without using Perl or Python to go through all the matches:

Rank all of the teams in all the regions by bonus conversion (since this is pretty independent). Rank the lowest bonus conversion as #1 and so on ... Then for a team finishing at position n in a region, compute its SoS by taking the average of the ranks of teams 1 through n-1. This wouldn't make it impossible for the Stanford Labe thing to happen, but it would make it far more difficult (since this indirectly takes the finishing place into account).
Jordan Boyd-Graber
UMD (College Park, MD), Faculty Advisor 2018-present
UC Boulder, Founder / Faculty Advisor 2014-2017
UMD (College Park, MD), Faculty Advisor 2010-2014
Princeton, Player 2004-2009
Caltech (Pasadena, CA), Player / President 2000-2004
Ark Math & Science (Hot Springs, AR), Player 1998-2000
Monticello High School, Player 1997-1998

Human-Computer Question Answering:
http://qanta.org/
Jaberwocky
Kimahri
Posts: 2
Joined: Tue Feb 10, 2004 8:05 pm
Location: Rochester, NY
Contact:

Post by Jaberwocky »

sobek22 wrote:Patrick, you also may have accidentally skipped over Fred "The Truth" Bush's Rochester A when compiling the list of at-large teams.

-Adam
Now that's one nickname for Fred I had not heard before, perhaps I'll have our team webmaster add it to his profile if we ever update that section again.

Glad to see Rochester A got added in their defense 3 of their losses came to Rochester B, so hopefully we don't ruin the chance of getting 2 teams in.

-Jack
User avatar
Theory Of The Leisure Flask
Yuna
Posts: 761
Joined: Fri Aug 29, 2003 11:04 am
Location: Brooklyn

Post by Theory Of The Leisure Flask »

Patrick-

Did you accidentally leave out Illinois A in the at-large bids?

As for Virginia, I don't know whether they'll be able to overcome their lower-bracket finish and get an ICT bid, but I'll second Adam and say that they certainly deserve it (probably even more than us, with our crap-ass powers and bonus conversion.) They convincingly beat us the one time we played them; that 18+ ppb makes me think that they most likely deserved to be second and just got unlucky.
Last edited by Theory Of The Leisure Flask on Wed Mar 17, 2004 12:37 am, edited 1 time in total.
Chris White
Bloomfield HS (New Jersey) '01, Swarthmore College '05, University of Pennsylvania '10. Still writes questions occasionally.
pakman044
Rikku
Posts: 262
Joined: Sun Nov 23, 2003 3:56 am
Location: Carrboro, NC
Contact:

Post by pakman044 »

benjaminthedonkey wrote:Patrick-

Did you accidentally leave out Illinois A in the at-large bids?

As for Virginia, I don't know whether they'll be able to overcome their lower-bracket finish and get an ICT bid, but I'll second Adam and say that they certainly deserve it (probably even more than us, with our crap-ass powers and bonus conversion.) They convincingly beat us the one time we played them; that 18+ ppb makes me think that they most likely deserved to be second and just got unlucky.
Yup, they go under Virginia I think.
PK
mujason
Wakka
Posts: 166
Joined: Fri Aug 15, 2003 12:47 am
Location: Missouri

Post by mujason »

Edit: I incorporated a new SOS calculation for the West, incorporating the bracketed double round-robin. Although it helped to increase the ratings of the top bracket teams and lowered the ratings of the lower bracket teams, Stanford Labe still holds a slight lead. Another adjustment is necessary.
Another Edit: New SE TUH stats included. Helped many teams.

Here are my Division II calculations. Notes: there are two major hiccups in these calculations: the West Coast bracketed playoff and the lack of TUH for the Southeast. For the Southeast, I've heard they had a low TUH; if that's true, the reduction in TUH should improve the ratings of the SE teams. Also, as noted before, Stanford Labe is overrated based on beating up on the bottom playoff bracket and the top teams beating up on each other (especially UCLA beating up on those top teams). A more refined SOS calculation (one that takes into account playing some opponents more than others) may help alleviate this problem, but I doubt that a new SOS would fix the problem completely. Perhaps a ratings cap could be used instead, but I'm sure NAQT has a good solution to the problem. Anyway, here are my Div II ratings. Enjoy!

Rank Team P Sect
1 ucla 0.745271734 W
2 chicago c 0.646622193 GL
3 illinois b 0.632842835 GL
4 carleton 0.629472407 NC
5 harvard b 0.622179062 NE
6 penn a 0.620879299 MA
7 cmu a 0.615972312 CA
8 emory b 0.610394716 SE
9 st. thomas 0.610340169 NC
10 pitt bal colts 0.599379191 CA
11 kansas 0.592415234 SW
12 mit b 0.586284336 NE
13 harvard c 0.584301028 NE
14 columbia 0.572006832 MA
15 athens state 0.568773941 SE
16 wichita 0.567972693 SW
17 virginia 0.567827877 MA
18 maryland 0.558120794 MA
19 tulsa 0.554494018 SW
20 swarthmore a 0.551493047 MA
21 wofford 0.542526134 SE
22 truman b 0.535094579 SW
23 florida b 0.531923519 SE
24 stanford labe 0.530774773 W
25 georgetown college a 0.530409047 K
26 pikeville 0.528689734 K
27 berkeley well 0.52783247 W
28 case western 0.522155546 CA
29 lsu a 0.52033367 SW
30 yale c 0.516811982 NE
31 caltech 0.516592879 W
32 berry 0.507284596 SE
33 toronto b 0.504486658 CA
34 mcgill b 0.503642231 CA
35 furman a 0.500405928 SE
36 rolla c 0.490460997 SW
37 stanford incoln 0.489694548 W
38 kentucky 0.48331424 K
39 ottawa b 0.482161634 CA
40 queen's b 0.478921975 CA
41 chicago d 0.476870693 GL
42 st. olaf a 0.474338469 NC
43 north carolina state 0.46781569 SE
44 virginia tech 0.462450621 MA
45 bu 0.462168944 NE
46 uc davis 0.461927282 W
47 villanova 0.46145141 MA
48 osu a 0.458981347 SW
49 queen's a 0.454418741 CA
50 michigan d 0.444684796 GL
51 murray state 0.440452524 K
52 ubc c 0.43951557 NW
53 transylvania a 0.438543958 K
54 dartmouth 0.435393821 NE
55 grinnell a 0.427148251 NC
56 tulane b 0.418841362 SE
57 osu b 0.414182811 SW
58 mcmaster a 0.410898414 CA
59 macalester 0.408092267 NC
60 utd 0.404225297 SW
61 rolla a 0.40087936 SW
62 iowa 0.396334619 NC
63 swarthmore b 0.388330826 MA
64 grinnell b 0.387385456 NC
65 lou-laf a 0.379524039 SW
66 northwestern b 0.378210116 GL
67 st olaf b 0.376621207 NC
68 transylvania b 0.374038979 K
69 ohio state b 0.370142156 GL
70 ill-springfield a 0.365090525 GL
71 macon state b 0.361060786 SE
72 oklahoma c 0.35977835 SW
73 ottawa a 0.358758888 CA
74 delaware 0.355519857 MA
75 cincinnati a 0.354745349 GL
76 bucknell 0.354098467 MA
77 penn b 0.349456238 MA
78 georgetown college b 0.349134069 K
79 rochester c 0.335579795 CA
80 tenn-chattanooga 0.33177139 SE
81 monmouth a 0.329368672 GL
82 robert morris 0.319314685 MA
83 toronto c 0.31292559 CA
84 penn state 0.309614027 MA
85 mcmaster b 0.305203866 CA
86 cmu b 0.301795872 CA
87 ubc b 0.301187881 NW
88 mcgill c 0.300536399 CA
89 rutgers-newark 0.291593434 MA
90 berkeley tito 0.290740992 W
91 berkeley deficit 0.284653563 W
92 furman b 0.280126758 SE
93 bc b 0.27792619 NE
94 sw missouri st 0.274923666 SW
95 sfu b 0.262771039 NW
96 morehouse 0.259267595 SE
97 cincinnati b 0.257352708 GL
98 lexington cc 0.254850791 K
99 lsu b 0.249022217 SW
100 lou-laf b 0.242387183 SW
101 bc c 0.232793734 NE
102 langara 0.222324957 NW
103 somerset cc 0.195842795 K
104 macon state a 0.190574324 SE
105 monmouth b 0.169198287 GL

Jason Mueller

Another note: I call the S-value P and the teams are in lowercase because in my spreadsheet, I lower-cased the Div II teams to help keep them apart from Div I. Also, it doesn't hurt that kansas was in Div II, and I don't like to capitalize kansas.
Last edited by mujason on Wed Feb 11, 2004 10:59 pm, edited 2 times in total.
User avatar
Dan Greenstein
Yuna
Posts: 848
Joined: Sun Apr 20, 2003 8:26 pm
Location: Takoma Park, MD

Post by Dan Greenstein »

By "visual analysis," here is my pronostication.

NAQT Division I Field Pronostication
Legend for Automatic Bids: Overall Champion Undergraduate Champion* Host Bid (LFI) = Last Four In**

* if there was 4+ UG teams at that sectional
** Basically, the at-large teams who would be left out if the field were 32 instead of 36

Reserved (1)
British Champion

Canada (3)
Rochester B
Rochester A
Western Ontario

Great Lakes (6)
Michigan A
Chicago A
Michigan B
Illinois A
Indiana A
Depauw

Mid-Atlantic (6)
Maryland A
Swarthmore
Delaware
Virginia (LFI)
Georgetown (LFI)
Princeton

New England (5)
Yale A
Harvard
Yale B
MIT A
Brandeis

North Central (3)
Carleton
Iowa State A
Minnesota (LFI)

Northwest (2)
British Columbia
Simon Fraser

Southeast (3)
Florida
South Florida A
Emory

Southwest (3)
Texas A&M Grad
Rice
Arkansas

West (4)
Berkeley A
Berkeley B
Cal Tech (LFI)
Stanford
Rog
Kimahri
Posts: 2
Joined: Wed Feb 11, 2004 6:53 pm

Order of selection from a sectional

Post by Rog »

Mid-Atlantic (6)
Maryland A
Swarthmore
Delaware
Virginia (LFI)
Georgetown (LFI)
Princeton
Dan,
UVA and Georgetown can't get bids until Virginia Tech and Columbia do. See Samer's post from above. At least, I'm pretty sure that's the way it works. Can you clarify this Samer?

-Roger
mujason
Wakka
Posts: 166
Joined: Fri Aug 15, 2003 12:47 am
Location: Missouri

Post by mujason »

Check out the edit in my Div II calculations.

Jason Mueller
User avatar
Dan Greenstein
Yuna
Posts: 848
Joined: Sun Apr 20, 2003 8:26 pm
Location: Takoma Park, MD

Post by Dan Greenstein »

As the only Sectional in which D1 was split for playoffs, Mid-Atlantic is naturally the most difficult to predict. Maryland, Swarthmore and Delaware are no-brainers. However, the other four teams are more trouble. Once again, a 6-7 team could be selected ahead of a 9-4 team.

Virginia's bonus conversion made me put them in the field of 36, but the division complication had me make them Last Four In. Georgetown gave us our second closest game, so that also influenced me.
NotBhan
Rikku
Posts: 306
Joined: Tue Dec 16, 2003 12:30 pm
Location: Parts Unknown

Post by NotBhan »

Dan, I'm not sure if I'm reading your notation correctly, but in the Southeast, Emory would go in as an at-large team, not an undergrad team. They had planned to play as an undergrad team but changed their minds, I think.

USF B was the top undergrad team in the Southeast Div I, but I don't know if there were enough undergrad teams for them to get an automatic spot. Tulane was undergrad, I think, but I'm not sure about Vanderbilt or Tennessee-Knoxville -- if both of those two were undergrad teams, then that would make 4 undergrad teams in the southeast.

--Raj Dhuwalia
"Keep it civil, please." -- Matt Weiner, 6/7/05
Rog
Kimahri
Posts: 2
Joined: Wed Feb 11, 2004 6:53 pm

Post by Rog »

Ok, Dan, I see your point. It's just that Virginia Tech and Columbia A had losing records because they were in the top bracket. UVA and Georgetown were in the lower bracket and consequently faced weaker competition. If I understand this correctly, NAQT takes this into consideration in two different ways. First, they account for the different schedule strengths and secondly, they only invite teams from an SCT in the order of their final placement. This is because of the Maryland-Princeton ICT invites of a few years ago when the Princeton team which had a worse record by about 3 games, but slightly better stats was invited ahead of the Maryland.

To sum up, the D-I teams invited from the Mid-Atlantic regional would probably be invited in this order:
(Maryland, Swarthmore, and Princeton are automatic)

Delaware
Columbia A or Virginia Tech (I think they tied, but Columbia did win 2 out of 2)
UVA or Georgetown (tied)

Based on stats, I would imagine this order would end up being Delaware, VT, Columbia, UVA, Georgetown.

One thing that is interesting to point out is that these stats do not include the three 8 tossup mini-games, which amount to another game played or about 8% more data. UVA was clearly one of the better teams at this SCT, but they lost 2 of their mini-games. I should point out that all three of Delaware's mini-games went to overtime or the last tossup. Sampling error clearly comes to mind.

-Roger
mujason
Wakka
Posts: 166
Joined: Fri Aug 15, 2003 12:47 am
Location: Missouri

Post by mujason »

If I recall correctly, I think that NAQT does, on rare occasions, invite a lower-finishing team over a higher-finishing one. I think Northwestern over Minnesota last year was an example. However, NAQT will probably try to invite the higher-bracket teams first.

Jason Mueller

(Not speaking for NAQT because he doesn't work for NAQT).
User avatar
Theory Of The Leisure Flask
Yuna
Posts: 761
Joined: Fri Aug 29, 2003 11:04 am
Location: Brooklyn

Post by Theory Of The Leisure Flask »

IIRC, for a team to be invited ahead of a team that finished higher in the standings, their S-value has to be a certain (rather high) value above the team ahead of them. While I'm pretty certain that Virginia's S-value is higher than VT's and Columbia's, I'm not sure if it's high enough.

I think Delaware will almost certainly get in; and I imagine that VT and Columbia will both be relatively high on the wait list (but I doubt they'll get bids outright). Georgetown won't be too far behind the above teams; there was'nt much difference in ability between any of the teams between 2nd and 7th last Saturday.

Of course, one thing sure to throw a monkey wrench into all this calculating is the lack of TUH in the Southeast, which looks like it'll throw a bit of unwanted guesswork into what's already been a controversial ICT bid process, even before the bids are announced...
Last edited by Theory Of The Leisure Flask on Wed Mar 17, 2004 12:39 am, edited 1 time in total.
Chris White
Bloomfield HS (New Jersey) '01, Swarthmore College '05, University of Pennsylvania '10. Still writes questions occasionally.
BigFlax
Lulu
Posts: 66
Joined: Wed Feb 04, 2004 10:04 pm
Location: Chicago, IL
Contact:

Post by BigFlax »

mujason wrote:If I recall correctly, I think that NAQT does, on rare occasions, invite a lower-finishing team over a higher-finishing one. I think Northwestern over Minnesota last year was an example. However, NAQT will probably try to invite the higher-bracket teams first.

Jason Mueller

(Not speaking for NAQT because he doesn't work for NAQT).
This is, in fact, true. We went 6-8 in Div I last year in the Midwest and were astonished to find that we were selected for the ICT (especially after not qualifying at 8-6 in a larger, if in some ways easier, sectional in 2002). Minnesota went 7-7 and in fact defeated us twice, though both times in very close games (the second came in overtime).

In terms of using that to see how NAQT might make exceptions to the "in order of finish" rule, I believe that our stats were, except for in the win-loss column, significantly better than Minnesota's in terms of stuff like bonus conversion and total points (I can't verify that at the moment because NAQT's website seems to be having issues, at least from my computer).

At any rate, we finished 19th out of 32 teams and sixth out of 13 undergrad teams, so at least you can't say we clearly didn't belong there. (Not that you were, but I felt I should point it out.)

EDIT: http://www.naqt.com/Results/2003-sct-re ... ml#Midwest

Indeed, we had more than 500 points more than Minnesota on fewer tossups heard (only seven fewer, but even so). Of course, this is presumably not the only stat they use to calculate S-values, but I don't have other stats at my fingertips.
mujason
Wakka
Posts: 166
Joined: Fri Aug 15, 2003 12:47 am
Location: Missouri

Post by mujason »

As for the case for Virginia, I think they had the 7th-highest bonus conversion rate in all of Division I. That should merit some consideration. I agree that the Mid-Atlantic competition was quite close.
My formula has Virginia at 17th, but that SOS was based on a round robin. A correction for the SOS would probably put them at about 20th.

Jason Mueller
pakman044
Rikku
Posts: 262
Joined: Sun Nov 23, 2003 3:56 am
Location: Carrboro, NC
Contact:

Post by pakman044 »

I made a recalculation based on new average TUH/game stats for the southeast, and edited (once again) my predictions above.

Patrick King
mujason
Wakka
Posts: 166
Joined: Fri Aug 15, 2003 12:47 am
Location: Missouri

Post by mujason »

I edited Div I for the Mid-Atlantic SOS. Virginia is down to 18, with Swarthmore at 22, Delaware at 24, Georgetown at 25, Virginia Tech at 29, and Columbia A at 35.

Jason Mueller
mujason
Wakka
Posts: 166
Joined: Fri Aug 15, 2003 12:47 am
Location: Missouri

Post by mujason »

I edited Div I and Div II to reflect the new TUH stats for the SE.

Jason Mueller
User avatar
quizbowllee
Auron
Posts: 2180
Joined: Thu Feb 12, 2004 2:12 am
Location: Alabama

Post by quizbowllee »

Hey, where did you get the Southeast "tossups heard" stats? I haven't seen them anywhere. I'm the captain of Athens State's team and I know our tossups heard number was low... Several rounds we didn't even reach 20. Thanks.

-Lee Henry
Athens State Quiz Bowl
Locked