On Fairness

Quizzes are unfair competitions. In a running race, for example, all the athletes run on the same track, and the fastest among them wins. In golf, all the competitors play on the same course under the same conditions, and the most skilled of them wins. In a quiz, however, each team of finalists is asked a different question each time that other teams may or may not get to answer. Hence, many factors such as the seating of teams, the passing format, bias in question setting, etc skew the outcome in a way that may not reflect the ability of the teams.

Okay, so all of that is old news. Today, stuff like Infinite Rebounds as a passing format and drawing lots for seating are pretty standard in quizzes. However, this is not enough. More is needed. Althought the combination of IR and drawing lots has made quizzes more fair than before, it is felt that we should go further to ensure fairness. So the questions is, what more can be done?

The Centaurian System

The Centaurian System is a passing format derived from Infinite Rebounds which is, IMO based on the assumption that equalising the number of Direct question to each team is essential for fairness. Although I have never really understood how it works, anyone wanting to do so should read this excellent description by it's creator, "Centaurian" Abhishek Nagraj. To the best of my knowledge, no quiz has ever been conducted on the Centaurian system, so no data is available

Criticism: The main Criticism of the Centaurian System is that few accept its central assumption, that the number of directs per team really matter.

Drawing lots for Questions

This method seems to be prevalent outside Pune (correct me if I'm wrong). It basically consists of making chits with every question number on them, and having the teams draw them. The teams are asked the question corresponding the number they pick. The beauty of this systyem is that teams have no one but themselves to blame for the questions they are asked.

Criticism: If your quiz is on Infinite Rebounds and teams draw lots for seating, this procedure is basically redundant. You can draw lots either for seating or for questions, but the quantum of fairness is the same (provided the order of questions is fixed). Its just that drawing lots for seating is easier and takes far less time.

Choice of Seating by Qualfying Order

This method is championed by VIT quizzers since it was first used in Quiz-o-mania '05 the SCIT Software Quiz. The finalists are asked to choose their position in the order of qualification. This means that the team qualifying first will get first pick of seats, the second team will get second and so on. The rationale is that the seating arrangement, while suitably random, allows teams to carry forward to the finals the advantage of their performance in the qualifying round. In other quizzes, teams once in the finals are on an equal footing irrespective of their performance in the elims.

Criticism: This method does not really give the first qualifiers an advantage because they have no say in the seating of other teams. Subsequent teams engage in a competitive game, as all teams jockey to get a favourable position. Even so, the middle teams are likely to get the greatest advantage.

Choice of seating by First Qualifiers

This method calls for the team qualifying first to decide the seating order for all teams. The rationale is the same as the previous method.

Criticism: This method gives an advantage only to the first qualifiers, all other teams being on an equal footing. It also relies on the first qualifiers knowing the capbilities of the other teams. A team from another city, for example would not be able to fully take advantage of it.

Round Reversals

As Abhishek has coverd this in great detail in a previous article, I shall not go into it again. Suffice to say, there should be 0>R>N round reversals, R being the no. of Round Reversals and N being the number of questions (with R preferably being far close to 0 than N).

Post Scipt

I have only ennumerated here the techniques that I have come across. If I missed any, please feel free to point them out.

Round Reversals - Are they really that effective?

I was reading JR's anniversary article archives written in a seemingly alternately illegible hand and experienced firsthand Gaurav's pole-position & 4 quarter system at the S.E.Q. This post is a result of these two things.

The point is this, whatever may be the fallacies in a particular system of scoring these are ironed out due to round reversal. I.e. if a certain team is getting large number of directs due to another team sitting before them, this advantage is immediately transferred to the team on the otherside of that team. So seemingly all is well.

Then comes Quiz-O-mania '05, which witnessed super Vibhendu-Anand first half performance followed by an amazing catch-up in the second half by The Moops. (ala CSKA Moscow). Reason? It was surmised that the first half of questions had a concentration of Hindi-film questions and due to these questions being absent and/or due to round reversal Gaurav's team getting to answer before them they could perform better. So it is like the quiz being in two parts. Two absolutely different styled halves (?) thereby making round reversal ineffective.

So Gaurav decides that we'll go one step further and reverse 3 times. Now as his quiz was mostly done by him and JR (70-30 as he put it) and probably due to dispersion of JR's questions throughout the quiz this didn't have any major difference. I think the result would have been pretty similar had the normal system been adopted.

So does the 'success' of Gaurav's idea as many have claimed mean that the more we reverse the better it is. Actually a good way of testing a particular method is to take it to the hilt and then test it's effectiveness. Funnily it turns out that such a system would effectively result into an 'ox-cart'/Boustrophedon system which JR linked to recently. That is we go along in a particular order and then reverse once the cycle for a question is over.

So consider a particular case.

A - B -C -D. D gets it right. Now it is C's turn by this system. So we go C - B - A - F. F gets it right. Fair enough?

Now consider the following :

Team A : 2 attempts 0 points
Team B : 2 attempts 0 points
Team C : 2 attempts 10 points
Team D : 1 attempts 10 points
Team E : 0 attempts 0 points
Team F : 1 attempts 10 points

Now looking here you would say that logically the next question should be to E. But - as F got it right the next question will go back to A and so on. Those teams, which are around the leading teams, will get maximum number of attempts due to continuos reversals. Now this is an extreme case - but whenever we reverse in a quiz the team that was supposed to get the next question but don't is extremely hard done by. This is like omitting them from the current cycle of questions much like being a lap behind in F1. So whenever we reverse we are heavily discriminating against such teams. Now that I think about it, this happened to us at least once at SEQ. So consider a snapshot of the passing during a reversal: D-E-F-A-F and so on. How unfair to B!

Ideally as most agree 'fair'ness in a system is characterized by equal number of directs and more importantly attempts. So playing around with IR doesn't naturally imply a better system. More importantly I wanted to point out that 'reversal' is not necessarily the answer to all problems. I am not saying that reversals should be done away with - maybe one would be okay. I think next time someone is cataloguing a quiz he could note when the rounds reverse and so on. BTW did anyone manage to track SEQ statistically? That would be very interesting. Pls. post stats if available.

! gnizziuq yppah neht lliT

~ Abhishek

The Pune Season Ender Quiz 2004-05 14th May 2005

Results:

1st: Niranjan Pednekar (TRDDC) & Sudarshan Purohit (PSPL)
2nd: Amit Varma and Leslie (Wisden Cricifo)
3rd: Rachit Lahoti (GSSL) and Abhinav Sharma (IIM Lucknow)
Joint 4th : Aditya Udas (MESCOE) & Meghashyam Shirodkar, Ganesh Hegde (VIT) & Shivaji Marella (BJMC)
6th: Vivek Venkataramani (BJMC) and Abhishek Nagaraj (COEP)

Organised by: Gaurav Sabnis, J. Ramanand & B.V. Harish Kumar

QM: Gaurav Sabnis

Report:

The Season Ender Quiz (thats the official name) had all what everyone desires in a perfect quiz. The elims had fine audio questions spanning a variety of music genre. The finals were formatted as a basketball game with 4 quarters with a round reversal after each (with no change in positions). The scoring was 2 points for a correct answer and 1 for half, doing away with the usual multiple of 10s scoring pattern. This made the final scores rather close: 20 for the 1st upto 9 for the 6th. A new 'pole position' system was introduced where the 3 toppers top of the elims were rewarded with 3,2,1 points to start.

The top 2 teams that won led the quiz neck-to-neck upto the 1st half of the quiz. Amit & Leslie maintained their lead till the 3rd quarter capitalising on all the cricket questions. Finally in the last quarter, Niranjan & Sudarshan managed a stunning breakthrough by cracking the visual connects on Andrew Lloyd Webber and Savarkar. The unique 'Kekule' connect (with APJ Abdul Kalam at the root node and 5 audio pieces) was particularly the highlight of the quiz. And yes, Gaurav did ask a football related question.

The Pune Quizzing Season of 2004-05 has ended dramatically with one one of the best quizzes we have seen so far. We hope to have a similar great year coming up.

Defending quizzing

Samrat's quizzing partner at IIM-L, Tadatmya Vaishnav, has written a couple of posts on the usual arguments against quizzers & quizzing. Here they are:

* In defence of quizzing - I
* In defence of quizzing - II