The Quizbowl Resource Center

Posted: **Fri Aug 17, 2012 5:37 pm**

This past summer, I worked in sports rankings, and this led me to have a pretty good understanding of how some sports rankings work. From it, I also gained a number of computer algorithms to calculate rankings, and some rank optimization algorithms that I would like to apply to quizbowl. So here I am announcing these ranks, which I plan to update regularly (maybe 2-4 times a month). Right now, I'm still in a bit of transition so I need to find out where I can get some webspace, but I have the basic algorithms down for it. If management is OK with it, I'd like to use this thread as a place to collect the links to SQBS results pages that you, the community, want me to include in my ranks (I'll be using the second post and third posts of this thread to keep track of the links that I've included already). Right now, I can only parse from SQBS scoreboard reports, so any results that are only on NAQT's page won't work for me. When posting results of a tournament, please also include the state of the tournament, and the state of any schools at that tournament not from that state so that I can keep it straight in the ranks.

One of the nice things about the method I'll be using (Hodge Rank with an l2 norm) is that it takes score differential into account, and allows for different matches to be weighted more heavily.

As for future improvements to this, I plan to change how the score differential weighting is generated, where instead of the differential used being the actual game differential, a team's effective score in a game is TU Correct*BC for tournament+Pts on Tossups. One potential problem with this is that some tournaments change difficulty midway through or for playoffs, but I see that happening less and less.

If you have any questions or suggestions on the technical side of things, let me know at the email listed below this post.

Edit: One thing to add, this ranking is all pairwise comparison based, so it won't work until the graph of teams is connected, which probably won't happen until late October or early November, but I can split the graph and do regional rankings until then.

Edit2: Just wanted to update everyone that with some help from Jeff Hoppes, I can now parse scores from NAQT's website, I just need the tournament ID. I also have everything up and running, just waiting to collect data. I'm going to try and pull a bunch of NAQT tournaments from the past year and post my results.

Edit3: Because there's no reason not to, I'll be doing this for the college circuit as well. I'll be having a number of "regional" rankings on the disconnected parts of the graph until the nationwide graph of teams becomes connected. I'm still looking for some webspace, as I don't want to overload anyone. Right now, I've got all my scripts up and running on a localhost, but I don't think that's accessible to the general public.
Anyways, here's to hoping that this rank will work well at predicting the national champions.

Edit4 from staff: Updated spreadsheets should be available at https://www.dropbox.com/sh/1u81z47hf8gms6k/ppHvqarBcu

Posted: **Fri Aug 17, 2012 5:37 pm**

Posted: **Fri Aug 17, 2012 5:38 pm**

This post reserved for further results links.

Posted: **Fri Sep 28, 2012 3:18 pm**

Looking over my current data, it's nonsensical to publish anything yet. Expect my first rank to come out around the same time as the first BCS ranking.

Posted: **Tue Oct 23, 2012 2:59 pm**

Just as an FYI, I'd expect meaningful results to start around when HFT posts their results since that tournament will connect the northeast to Texas, and I'd guess that by that point we'll be able to connect the northeast to Illinois through Northmont.

Posted: **Fri Nov 02, 2012 4:22 pm**

The file of scores I've been working with has gotten too big in order to upload to this site (limit 256 kB), so if anyone wants it, let me know. The highlights of the rankings to this point:

-The Northeast has connected to the Mid-Atlantic, allowing Dorman to take over the highest ranked spot in our most connected region.

-The South seems much stronger than the north. This may be due to the fact that these regions are only weakly connected by 1 or 2 teams.

-High Tech has 3 teams in the top 36, which is pretty cool.

I welcome any comments and suggestions to help improve these rankings.

Posted: **Fri Nov 02, 2012 5:38 pm**

What do the numbers mean?

Posted: **Fri Nov 02, 2012 6:16 pm**

The number is a rating, a scale-able number that, the closer numbers are together, the closer they are. I've just kept it on a 0-100 scale, with a higher number being better, normalized so the worst team is a 0, and the best team is a 100. The ranking then becomes your order by rating.

Posted: **Fri Nov 02, 2012 6:21 pm**

Golran wrote:The number is a rating, a scale-able number that, the closer numbers are together, the closer they are. I've just kept it on a 0-100 scale, with a higher number being better, normalized so the worst team is a 0, and the best team is a 100. The ranking then becomes your order by rating.

Sorry for being unclear, what I meant to ask was if there was any meaning in the numbers besides as strictly as an ordering (i.e. a 2 point differential means the average voter has 75% chance of favoring the higher team, or something)?

Posted: **Fri Nov 02, 2012 7:34 pm**

The method it seems you're describing would be a random walker. What I've done is created a multigraph between the teams (nodes) where each game played is an edge weighted by the score differential. We then have an m by n matrix (B) of the game results, where m is the number of games and n is the number of teams, where the entry B_i,j is 1 if team j won game i, -1 if team j lost game i, or 0 if they didn't play. We also have an m by 1 vector, y where y_i is the absolute score differential in the game. Finally, we have an n by 1 vector of ratings, phi, which will be the result of a minimization. From the data, we are able to create B and y uniquely, and the goal is to minimize ||B*phi - y|| over the set of all vectors phi, under a given norm, which I've been using the l-2, Euclidean norm. We can further add a stronger weighting term for specific games to, say have IS-A sets weighted at 0.5, IS sets weighted at 1, SCT at 1.25, and all other games at 0 to predict HSNCT.

How we had been interpreting the ratings for college basketball was using them in the Bradley-Terry model of pairwise comparisons, where if a team of rating r1 plays a team of rating r2, the probability team 1 wins is (r1+1)/(r1+r2+2) .

Posted: **Tue Dec 11, 2012 12:13 am**

Will this be updated?

Posted: **Tue Dec 11, 2012 2:51 pm**

Yes... I've got 5 finals in 5 days this week, so I'll be putting together an update next week once I've finished these finals.

Posted: **Wed Jan 16, 2013 6:50 pm**

Posted: **Wed Jan 16, 2013 7:14 pm**

Schools with apostrophes in their names are sometimes also listed without them, causing St. John(')s (TX) and Maggie Walker Governor(')s School, and perhaps others, to be listed twice.

Posted: **Wed Jan 16, 2013 7:28 pm**

Making this change by just deleting all the single quotes.

Edit: And I went through and did it. Current set should reflect this, let me know if you see any more like that.

Posted: **Wed Jan 16, 2013 8:22 pm**

Golran wrote:Making this change by just deleting all the single quotes.

Edit: And I went through and did it. Current set should reflect this, let me know if you see any more like that.

Athens Academy shows up twice, 13th and 15th.

Posted: **Wed Jan 16, 2013 10:17 pm**

Western Albemarle A (VA) is 30th and 65th. They are misspelled once with an extra "r" Albe"r"marle A

Also, Montgomery Blair A is 21st and 32nd; listed once as VA, once as MD. There is no Montgomery Blair HS in VA.

Posted: **Wed Jan 16, 2013 10:59 pm**

mithokie wrote:Western Albemarle A (VA) is 30th and 65th. They are misspelled once with an extra "r" Albe"r"marle A

Also, Montgomery Blair A is 21st and 32nd; listed once as VA, once as MD. There is no Montgomery Blair HS in VA.

Fixed.

Posted: **Thu Jan 17, 2013 3:51 pm**

My recommendation would be to do a sort-by team name and look through for repeat names. There are some I'm not certain of, but I do see these in looking through the A's and B's.
American Christian & American Christian Academy
Bainbridge & Bainbridge Middle
Belvidere & Belvidere North (?)
Ben Franklin A-B & Benjamin Franklin A-B
Bettendorf (IA) & Bettendorf (IL) (is this a typo?)
Blue Ridge & Blue Ridge Middle
Boerne & Boerne-Champion
Whatever 'Bridgeport A Tie' is

Obviously in a file with 12000 game results it could be tedious, but games involving scab teams or unidentifiable pseudonyms should probably be removed.

Posted: **Thu Jan 17, 2013 3:55 pm**

jonpin wrote:Belvidere & Belvidere North (?)

These are different schools.

jonpin wrote:Bettendorf (IA) & Bettendorf (IL) (is this a typo?)

Probably.

Posted: **Thu Jan 17, 2013 4:19 pm**

jonpin wrote:Boerne & Boerne-Champion

Boerne High and Boerne Champion are the two public high schools in Boerne, Texas.

Posted: **Thu Jan 17, 2013 5:17 pm**

jonpin wrote: Bainbridge & Bainbridge Middle

Bainbridge is the Middle School feeder to Bainbridge High School.

Posted: **Thu Jan 17, 2013 5:58 pm**

jonpin wrote:My recommendation would be to do a sort-by team name and look through for repeat names. There are some I'm not certain of, but I do see these in looking through the A's and B's.
American Christian & American Christian Academy
Bainbridge & Bainbridge Middle
Belvidere & Belvidere North (?)
Ben Franklin A-B & Benjamin Franklin A-B
Bettendorf (IA) & Bettendorf (IL) (is this a typo?)
Blue Ridge & Blue Ridge Middle
Boerne & Boerne-Champion
Whatever 'Bridgeport A Tie' is

Obviously in a file with 12000 game results it could be tedious, but games involving scab teams or unidentifiable pseudonyms should probably be removed.

Fixed these, the ones with "middle" are a different school according to NAQT, likely the middle school associated with the high school. I went through the rest and think I got them all, let me know if there are any more you catch.

Also, I'm effectively removing most pseudonyms by requiring a 10 games played minimum to have a rank other than the default average, though psuedonym teams that do play more are useful since they provide more connections in the graph of teams. The new "unranked" rating is 438.7922.

Posted: **Thu Jan 17, 2013 8:10 pm**

delete this I can't read above me

Posted: **Sat Jan 26, 2013 1:53 am**

New ranks are up at the dropbox, I'll probably be keeping current by updating every Friday.

Posted: **Sun Jan 27, 2013 5:42 pm**

You might want to link to that dropbox in the first post of this thread.

Posted: **Sun Jan 27, 2013 6:03 pm**

Or not. Ctrl+F is a wonderful thing, my friends.
Besides, there's some rather wonky things that are happening in here.

Like, besides questionable numbers that I guess unavoidably happen between unconnected regions, does anyone know how exactly Dunbar B got to be like 130 points higher than Dunbar A?

Posted: **Sun Jan 27, 2013 6:05 pm**

thyringe_supine wrote:Or not. Ctrl+F is a wonderful thing, my friends.

.

User was warned for emptyquoting. --the mgmt

Posted: **Sun Jan 27, 2013 6:38 pm**

thyringe_supine wrote:Or not. Ctrl+F is a wonderful thing, my friends.
Besides, there's some rather wonky things that are happening in here.

Like, besides questionable numbers that I guess unavoidably happen between unconnected regions, does anyone know how exactly Dunbar B got to be like 130 points higher than Dunbar A?

Looks like a combination of a few things, but it's mainly because Dunbar "B" actually beat Dunbar "A": http://www.naqt.com/stats/team-performa ... m_id=48163 (round 9)

Posted: **Sun Jan 27, 2013 8:04 pm**

Posting the link in the first post saves everyone time and takes ten seconds effort.

Posted: **Sun Jan 27, 2013 11:15 pm**

But which one gives you the satisfaction?

Anyway, is there anyone who actually saw this Dunbar match happen? Because I'm sort of mystified.
And like I said, there's a lot of stuff in here that isn't as obviously strange as that.
For one thing, where in Sam Hill is Etowah?
My admittedly limited knowledge of Georgia quizbowl doesn't recall that name.
Also, I haven't really ever paid attention to Cali quizbowl, but since they're the only California team in the top 50, I have no idea how good Arcadia is, especially since most likely none of the other teams near them (rankings-wise) would have gone to any of the same or maybe even similar events.

Posted: **Sun Jan 27, 2013 11:41 pm**

thyringe_supine wrote:Anyway, is there anyone who actually saw this Dunbar match happen? Because I'm sort of mystified.

Dunbar's "B" team is good. Just like most schools that have traditionally strong quiz bowl programs.

Posted: **Sun Jan 27, 2013 11:44 pm**

thyringe_supine wrote:For one thing, where in Sam Hill is Etowah?
My admittedly limited knowledge of Georgia quizbowl doesn't recall that name.

Etowah's in northern Georgia (Cherokee County, to be specific). In my experience playing them, they got good at the beginning of 2012 and have been steadily improving ever since.

Posted: **Sun Jan 27, 2013 11:46 pm**

But that's not what I mean. If that's legit and the Dunbar teams are in the correct order, that still doesn't make sense.
Shouldn't the definition of a B team be 'the team that is worse than the A team'? That's always been my understanding of how it goes.

EDIT: Oh, okay. But are they really that good, and I've just heard about it?

Posted: **Sun Jan 27, 2013 11:53 pm**

thyringe_supine wrote:But that's not what I mean. If that's legit and the Dunbar teams are in the correct order, that still doesn't make sense.
Shouldn't the definition of a B team be 'the team that is worse than the A team'? That's always been my understanding of how it goes.

EDIT: Oh, okay. But are they really that good, and I've just heard about it?

Plenty of B-teams have beaten/out-placed their respective A-teams over the past few years. Thomas Jefferson and DCC at the 2011 HSNCT are a couple that come to mind.

Also, having played Etowah multiple times, I think I can say with confidence that they are a legitimate top 25-30 contender, and could place even higher with practice.

Posted: **Mon Jan 28, 2013 12:11 am**

Well, yes, but you just mean at a certain event. These rankings would seem to suggest that their B team was consistently significantly better than their A team. I don't really think this is the case, which is why I wonder under what circumstances it happened

Posted: **Mon Jan 28, 2013 12:53 am**

thyringe_supine wrote:Well, yes, but you just mean at a certain event. These rankings would seem to suggest that their B team was consistently significantly better than their A team. I don't really think this is the case, which is why I wonder under what circumstances it happened

Dude, it's called jobbing. When a team has two strong teams, it comes down to who the packet hits better. That's why you play the game.

At 2011 HSNCT, DCC B finished ahead of DCC A. At 2008 HSNCT, Dorman B finished ahead of Dorman A.

Posted: **Mon Jan 28, 2013 3:24 am**

In this particular case, Dunbar B has a history of finishing higher than Dunbar A, although sometimes it has to due with splitting their regular lineup among two teams for various reasons.

Posted: **Fri Feb 08, 2013 12:13 pm**

thyringe_supine wrote:EDIT: Oh, okay. But are they really that good, and I've just heard about it?

Etowah has one player, Duncan, who is very good. I don't recall even seeing anyone else in a chair on their side first-buzz a question in the 2-3 games I have watched him play my team this year. He is not at the level where Adam Silverman was last year, but he is easily one of the best - if not the best - individual players in GA. Similar to Adam, he has great depth in lots of areas, he is very fast to what he knows, and he doesn't make a lot of mistakes (certainly not egregious ones). Assuming he registers, I am not predicting he finishes in the top 4 at HSNCT (a la Adam), but a deep Sunday run isn't out of the question.

Posted: **Sat Feb 09, 2013 1:07 am**

New ranks are up (link in first post), minimum games pushed up to 11 to try to force teams out of the ranked category if they've only played 1 tournament and therefore are not very connected to the rest of the country. (Cistercian, as a heads up, when the minimum was at 10 and I was looking at them, you were ranked 3rd as a result of only your most recent tournament and moving all your previous performances down a letter).

Posted: **Fri Feb 15, 2013 12:37 am**

Why did Arcadia drop 17 places in the most recent ranking? We haven't been to any new tournaments since the last update.

Posted: **Fri Feb 15, 2013 12:55 am**

Other teams played games during that time, and as a result their ranks went up. Also, since you're not really connected to other regions, you'll be very volatile.

Posted: **Sat Mar 02, 2013 11:13 pm**

Golran wrote:New ranks are up (link in first post), minimum games pushed up to 11 to try to force teams out of the ranked category if they've only played 1 tournament and therefore are not very connected to the rest of the country. (Cistercian, as a heads up, when the minimum was at 10 and I was looking at them, you were ranked 3rd as a result of only your most recent tournament and moving all your previous performances down a letter).

With the conclusion of the NAQT tournament (IS-124) at St. Mark's School of Texas today, hopefully Cistercian will now have enough games to provide a sufficiently large statistical basis for comparison.

Posted: **Sat Mar 16, 2013 9:17 pm**

New ranks are now in the dropbox covering over 20,000 games between 2481 teams in 239 tournaments. The list is available in the dropbox as "Entered Tournaments.txt". Of note is Ladue jumping up to #1, and Texas teams occupying 4 of the top 7 spots. Biggest drop by a former top-10 was Ezell-Harding from #4->#12, while Edmund Memorial (OK) jumps up to #11 from #29 with their 26-1 record on the year.

Posted: **Sun Mar 17, 2013 1:35 pm**

Bellarmine B and C are both shown twice as Bellarmine B/C and Bellarmine College Prep B/C

Posted: **Tue Apr 30, 2013 2:42 pm**

Is there a reason that Glasgow High School (KY) is not in the list. Not a lot of data, but we did play at Ezell-Harding's tournament this past September.

Todd Garrison
Glasgow High

Posted: **Tue Apr 30, 2013 4:21 pm**

Peabody wrote:Is there a reason that Glasgow High School (KY) is not in the list. Not a lot of data, but we did play at Ezell-Harding's tournament this past September.

Todd Garrison
Glasgow High

I believe Ian's method uses individual game scores, which Ezell-Harding did not report.

Posted: **Fri May 03, 2013 1:16 pm**

Got it. Thanks!

Posted: **Sun May 12, 2013 3:48 pm**

Finally done with finals etc, I'll be updating this in the next couple of days... I've just got about 80-100 tournaments left to enter (I fell behind really badly), and I'd expect to have the update done by Tuesday night at the latest.

In relation to Bellarmine being double listed, that will be fixed in the next update, and Jeff is exactly right in that I use scores(or more importantly score differential) as the entire source of my rankings, so just knowing if a team won or lost a game is not enough to input. If you have scores that I don't, send me a message and I'll add them in.

Posted: **Mon May 13, 2013 3:42 pm**

New ranks are now available in the dropbox at:

https://www.dropbox.com/s/8dnqgpkia07zr ... %2013.xlsx

Notable in the newest set of rankings:
-The "zero" score for insufficient games (still need 12 games to be ranked) is now 426.0912.
-Arcadia, once it connected to the rest of the circuit through Bellarmine B when they played up north, has jumped up really high in the rankings. On the other hand, Escobar's C team dominating one tournament and then losing to Arcadia in a different one may have contributed to this. While I personally believe Arcadia is a top 25 team, I don't see them as a top 3.
-Texas once again dominates the top 10 with 4 of my top 10 teams.
-Edmond Memorial from Oklahoma has played some very weak opposition, leading to their high ranking being inflated by large point differentials when their opponents just don't score much.

Included in this set of rankings are 29,851 games between 3,119 teams over the course of 360 tournaments
I'll have one more set pre-HSNCT, one more between HSNCT and PACE-NSC, and one final set after both nationals. For the HSNCT and PACE rankings, I'll put out two sets: one will contain all the games as all of my rankings have, while the other will contain only the teams that are attending the relevant tournament.

Once again, I welcome any corrections and comments on the system or any of the data I've collected. I've got some ideas on how to improve this process for next year, but for a first effort, I think it's been as good as expected. These plans include integrating the ratings into the web browser as opposed to an excel sheet using PHP/MySQL, using more than an L2 ranking method such as having the Massey and Colley rankings available, and having a way for users to input scores and tournaments into the rankings. This will all depend on how much time I have available over the summer, but I'll do what I can. In any case, I plan to run these again in at least the same level as I did this year.

The Quizbowl Resource Center

I'm making [yet another] new high school quizbowl ranking

I'm making [yet another] new high school quizbowl ranking

Re: I'm making [yet another] new high school quizbowl ranking

Re: I'm making [yet another] new high school quizbowl ranking

Re: I'm making [yet another] new high school quizbowl ranking

Re: I'm making [yet another] new high school quizbowl ranking

Re: I'm making [yet another] new high school quizbowl ranking

Re: I'm making [yet another] new high school quizbowl ranking

Re: I'm making [yet another] new high school quizbowl ranking

Re: I'm making [yet another] new high school quizbowl ranking

Re: I'm making [yet another] new high school quizbowl ranking

Re: I'm making [yet another] new high school quizbowl ranking

Re: I'm making [yet another] new high school quizbowl ranking

Re: I'm making [yet another] new high school quizbowl ranking

Re: I'm making [yet another] new high school quizbowl ranking

Re: I'm making [yet another] new high school quizbowl ranking

Re: I'm making [yet another] new high school quizbowl ranking

Re: I'm making [yet another] new high school quizbowl ranking

Re: I'm making [yet another] new high school quizbowl ranking

Re: I'm making [yet another] new high school quizbowl ranking

Re: I'm making [yet another] new high school quizbowl ranking

Re: I'm making [yet another] new high school quizbowl ranking

Re: I'm making [yet another] new high school quizbowl ranking

Re: I'm making [yet another] new high school quizbowl ranking

Re: I'm making [yet another] new high school quizbowl ranking

Re: I'm making [yet another] new high school quizbowl ranking

Re: I'm making [yet another] new high school quizbowl ranking

Re: I'm making [yet another] new high school quizbowl ranking

Re: I'm making [yet another] new high school quizbowl ranking

Re: I'm making [yet another] new high school quizbowl ranking

Re: I'm making [yet another] new high school quizbowl ranking

Re: I'm making [yet another] new high school quizbowl ranking

Re: I'm making [yet another] new high school quizbowl ranking

Re: I'm making [yet another] new high school quizbowl ranking

Re: I'm making [yet another] new high school quizbowl ranking

Re: I'm making [yet another] new high school quizbowl ranking

Re: I'm making [yet another] new high school quizbowl ranking

Re: I'm making [yet another] new high school quizbowl ranking

Re: I'm making [yet another] new high school quizbowl ranking

Re: I'm making [yet another] new high school quizbowl ranking

Re: I'm making [yet another] new high school quizbowl ranking

Re: I'm making [yet another] new high school quizbowl ranking

Re: I'm making [yet another] new high school quizbowl ranking

Re: I'm making [yet another] new high school quizbowl ranking

Re: I'm making [yet another] new high school quizbowl ranking

Re: I'm making [yet another] new high school quizbowl ranking

Re: I'm making [yet another] new high school quizbowl ranking

Re: I'm making [yet another] new high school quizbowl ranking

Re: I'm making [yet another] new high school quizbowl ranking

Re: I'm making [yet another] new high school quizbowl ranking

Re: I'm making [yet another] new high school quizbowl ranking