An approach to buzz point tracking

Old college threads.
Locked
User avatar
the third garrideb
Wakka
Posts: 101
Joined: Sat Oct 17, 2009 4:49 pm
Location: Waterloo, Ontario, Canada
Contact:

An approach to buzz point tracking

Post by the third garrideb »

Sometime in the past I had an idea which I've gotten a chance this week to put together some code for, so here it is. The basic idea is that you click on words in the packet as you read the questions, marking the tossups with correct/negged buzzpoints, and then generate a list of the buzzpoints at the end.

Here's a sample (using my team's submission to last year's VETO): http://csclub.uwaterloo.ca/~hzafar/buzz ... ssups.html

To quote from the github project's readme file:
TO USE: Click on the word that a player buzzes in on. The entire clue corresponding to that word will be highlighted. After one click, the clue will be blue, indicating a correct buzz. Another click will turn the clue red, indicating a neg. A third click will turn it brown, indicating the same clue was negged on and answered correctly with (i.e. if a team negs and the other immediately picks up the tossup). A fourth click will return the clue to normal.

When the packet has been read through and all tossups marked (dead tossups may have no marks), click the "generate stats" link at the top or bottom of the page to generate an XML of the buzzpoints.
Even though the questions are being read in a browser, no Internet connection is needed. The generated XML files can be saved locally and collected after the tournament.

I tested this by playing the packet through with my sister, to get a feel for how clicking on questions might interfere with moderating. I found it pretty usable and non-disruptive of the "game flow", but other feedback regarding this would be appreciated. (I also haven't tested functionality in IE or on a Mac.)

One issue with this is that it requires packets to be converted to XML (and then transformed into HTML). Depending on your resources, that can be a lot of work. Another is that I'm not really sure what should be done with the generated list of buzzpoints. If anyone has ideas for useful ways to present those stats (I have no statistics-related talents), I'd love to hear them.

If anyone thinks this is interesting enough that they might want to try it out for their upcoming tournament (even if only in one room or something), let me know, and I can help you get your questions formatted with the correct XML if you need (as long as it's not a tournament I'm planning on playing :p).

(Note: The bonuses in a packet would be in a separate file, just as normal bonuses. I didn't bother encoding and generating the bonuses for the sample packet, since no buzz point tracking is needed for those.)
Huma Zafar
University of Waterloo alumna
User avatar
Huntur
Lulu
Posts: 88
Joined: Wed Mar 14, 2012 8:49 am
Location: Blacksburg, VA

Re: An approach to buzz point tracking

Post by Huntur »

Statistically speaking, I think the only relevant stat would be "% of question heard" (%QH) This would be calculated as (# of words read)/(# of total words)*100%.

Making a guess: You'd probably see two peaks on a plot of that distribution. The first peak would be teams that are routinely powering a set and the second peak would be questions being buzzed on at completion. I would really caution looking at a simple average of %QH due to those peaks. It's not going to be a simple report-able number like most stats.
Therein, you could probably look at the correct vs incorrect buzzes and their respective %QH. However, you'd see outliers for good teams when they get to wait till the end of a question due to a neg by the other team.
For players/coaches, this would give them an idea of whether their team is potentially being too aggressive for the field and could sit for another clue.
For editors/writers, there'd be a statistical plot of how "hard" the set was for the field.

I think the more directly impactful usage could be for resolving protests where the location of buzz is important. You'd have it documented at the time of the buzz if all goes smoothly.
Brian Mongilio
Virginia Tech '16, Treasurer 2011-Present
Parkersburg South (WV) '11
User avatar
Habitat_Against_Humanity
Rikku
Posts: 476
Joined: Sun Jan 21, 2007 8:51 pm
Location: Syracuse, NY

Re: An approach to buzz point tracking

Post by Habitat_Against_Humanity »

I imagine clustered negs in a given question would indicate poorly worded and/or misplaced clues.
Rachel
UChicago 09
User avatar
the third garrideb
Wakka
Posts: 101
Joined: Sat Oct 17, 2009 4:49 pm
Location: Waterloo, Ontario, Canada
Contact:

Re: An approach to buzz point tracking

Post by the third garrideb »

Habitat_Against_Humanity wrote:I imagine clustered negs in a given question would indicate poorly worded and/or misplaced clues.
Yes, my main goal with this was to be able to provide more useful feedback for writers/editors than stats currently provide for tossups. (For bonuses, I think average conversion stats are usually a pretty good indication of how well-written the bonus was.) Things like which clues were repeatedly negged, which were never buzzed on, how buzzes were distributed throughout a question, etc. It's maybe not data you could put into a formula to quantify how good a set was (or could you?), but it could be really helpful when writers are reviewing their questions.
Huma Zafar
University of Waterloo alumna
Locked