Monday, March 22, 2004

A list of general principles for fairness in conducting quiz finals

Since we are discussing formats old & new, I thought it would be a good idea to first jot down what are the parameters of comparison for fairness. The disclaimers here are that (a) these are my personal opinions, please leave comments behind so that this list can be stabilised to a unanimous set (b) these only apply to the normal quiz final format, no buzzers or special rounds are addressed here.

I think the following three major aspects have a bearing on the fair conduct of the final:

  1. The Passing Format
  2. Distribution of Questions in the Question Set
  3. The Order in which the Teams are Seated

The Passing Format:

This determines who gets the question first, to whom it passes next, and then who gets the next question. The most popular formats are IR and D&P. Here are what I think should be the parameters of interest:

  1. Each team should ideally have had the same (with tolerance) number of attempts at the end of the quiz. Here, "attempts" include "chances on passing" as well as "direct attempts". Since we cannot have all teams attempting all questions, this is the next best thing we can do. I have stats to show that using IR definitely achieves this goal. Since I don't have enough stats on D&P quizzes, I can't claim the same about D&P.
  2. Another important goal is to try and ensure that even at the unit level, similar number of attempts at questions have been provided to all teams. This essentially means that, say if the quiz has a set of rounds, there must not be a great distortion in the number of attempts for teams. Say it is a Sports round and 2 out of the 6 are the best in that area and can potentially get all 6. If the format causes one of those two to get 4 attempts and the other only 2, then it gives them an unfair advantage.

    There is a likelihood of this goal being violated in D&P. We know of certain combinations in D&P in which even though we may get the same number of attempts at the end, at certain points, there can be a massive skew. I have computed this only for the VIT-'04 quiz in the IR scenario, and it didn't seem to have this problem.

    Perhaps, if the questions are so well distributed and have very few specialised rounds (the best example being "Seamless quizzing" of which the only instance I've seen so far has been Chakravyuuh-'01 and a close version in the Mensa quizzes), this goal may not be important since things even out at the end. But especially with specialised rounds and for the psychological parity, I think this is important.

    This is the principle of Local Parity in which things are fair even in a smaller window of attempts - the Centaurian format uses this as the central theme. IR does have this.

    People may remember the minor quibble over the variation I had mentioned here which the VIT guys wanted to use but couldn't - preventing two questions going to the same team twice in a row helps Local Parity. Perhaps we'll talk about it another day in detail.

  3. A sub-goal that has emerged out of the collected stats has been the number of Directs to a team. It is widely felt that this number should be equal for all teams, and the newer format(s) essentially want to tackle the "anomaly" in IR where the number of Directs can be variable.

    Ok now, personally speaking I still don't believe in differentiating between a Direct attempt and an attempt on a pass. Apart from a minor psychological effect on the team with first shot, it has no real bearing on the events, IMHO. The only time this makes a difference is when the number of so-called "sitters" (where it is extremely well-known to a majority, has been asked before several times, and usually never gets passed at all) is higher than usual, the Directs can make a difference, especially when interspersed with tough questions. In this case, the more directs I get, the more sitters I have a probability of gaining and hence the advantage. The VIT-04 quiz is often cited as an example of this. I think it was a freak case where it so happened that a team that was both good and experienced was between two teams that were themselves getting a lot of correct answers. I haven't posted the Shyam Bhatt stats yet, but that shows that team A (us) got the highest number of directs - we still didn't win.

    However, if this goal can be achieved to to everyone's satisfaction, it could be useful. Again, this arrives out of improper distribution, so I feel we shouldn't find too many drastic solutions in the format to cover all the flaws there.

  4. It goes without saying that ideally, each question must have the same value attached to it. Since we've seen Connects that are large in scope and sometimes are never solved completely - perhaps we need to ensure that the amount of deduction & effort required from quizzers is also roughly the same. Connects, especially Visual ones, tend to overdo it.

Distribution of Questions in the Question Set

The focus has been on the format, but this is equally important. Would someone like to venture in enlisting goals for these?

The Order in which the Teams are Seated

Niranjan has been working on this too, so I would like to leave it for him to fill in.

No comments: