Sunday, March 21, 2004

A Proposal For new Centaurian system of Scoring/Passing in quizzes

Life just got more exciting: Abhishek Nagaraj from F.C. sent in this proposal for a new format. I haven't read it, but promptly posting it here for instant dissemination. Abhishek: hope to send you some comments soon. Interestingly, you've pipped me to the post here - I've been thinking about a new format myself (like you, an amalgamation of D&P and IR) that I've been informally discussing with Niranjan and others. I'll post that here too for you to take a list. From what I've seen so far, your system is very close to what I came up myself (I'm not sure which of us is Newton and which one Leibnitz ;-) ) but I think that's because the motivations and the policies used are similar.

Abhishek, A latest copy of Niranjan's code can be made available by The Great One himself - Niranjan, could you please post it on inquizitive?

Will mull over it soon and we'll have lots of comments, I am sure. This blog is becoming highly inadequate for the amount of stuff that gets discussed here. We need something like twiki or Drupal or even an MT based blog.

A Proposal For new Centaurian system of Scoring/Passing in quizzes

There has been plenty of discussion on notes and stones blog and indeed at the various quizzes and meets that i have been a part of about infinite rebounds and D&P. However the major argument in the favour of IR is that
1. The final ranking is flatly based on the number of questions each team gets right.
2. The scoring for a correct answer should be the same
3. The number of questions each team gets to answer should be the same.

It has been seen by the data compiled till now that in IR 'mostly' the number of questions answered per team is same. However for eg in the VIT quiz the number of questions attempted varied very less from 23 to 25. That is really acceptable. However there is a large deviation in the total no of questions each team answered on the direct. This number varies from 7 to 15 which is huge considering the total no of questions answered is in 20s. Therefore there is more than 30% deviation of the total questions answered direct. This is the main argument in the case of D&P. The number of questions each team gets to ans direct is same. Many would say that does not matter whether a team answers it right or wrongly. There are many a questions to which a very obvious answer is wrong. The Frasier question is an example. Another example is the "god of construction and building" question in AFMC(though that was in the elims if in the finals it would have given a very obvious advantage to the passing teams). Many such examples can be given and i would say that there is a certain charm in these questions . They might be a good form of knowledge and entertainment(and indeed clearing basic fundas). We could call them PM(popular misconception) questions. These PMQs are the major bone of contention and this is the only point where D&P is better form than IR. While discussing this point with some ppl from FC there was a suggestion that such questions should not be set by the quizmaster in the finals. I feel this is escapism and too easy a way to solve the problem. There must be some method to cope with the problem of PMQ. Thus the major faults (which i percieve) in IR can be summed up as

1. Inequality in the no of direct questions
2. Give undue advantage in the case of PMQs.

My proposal intends to solve the both the problems.

What can be done is that the total no of direct attempts by each team be noted down. Once a question is answered or attempted by each and every person and there comes a time for the next question then the team with the least no of direct questions should be posed the next question. Following is some sample data


TEAM     A     B     C     D

     .     .
(where the dots represent no of direct attempts)
Consider following case. Direct Q to A - no one gets it. Therefore every team gets an attempt.Now B,C,D all have min no of direct Q answered. Then we go by alphabetical or round order(clockwise/anticlockwise)Then B gets direct Q. C answers correctly. Therefore as C & D both have min no of direct questions the order is decided by no of attempted questions(ie D gets next question). It is worth remarking that in this case the order is exactly like IR.


TEAM      A     B     C     D
     .     .     .     .     
          .     .
(where the dots represent no of direct & indirect attempts)
Now when D gets direct question suppose A answers it. Then according to above rule as C is the team with lowest no of direct attempts it should get the next question instead of B as in the case of IR. We must notice that IR penalizes team C for answering correctly by not giving it a direct question. This can be totally avoided by this method . It ensures that total no of direct questions remains the same. What we do is that we fill up the gaps in the table so that each team gets the same no of direct questions.
Now of both the no of direct attempts and total attempts show the same no, the the next question should go to the team which would have been asked the question if IR was followed(ie next team than the team to which the last question had been asked)
Thus one major drawback in the IR method is avoided and a major plus point of the D&P is added to the IR format. (see table)Therefore this is a kind of marriage between the two methods. I would like to call it the Centaurian method of scoring.

The second problem i feel gets automatically eliminated once the no of direct questions each team gets is the same. By the law of averages the no of such questions going direct or passed to each team will remain generally equal. Thus in a limited sense this problem is solved.

This system needs to be tested both in simulation as well as in real time. I am also working out a few details whether it would be better to see the indirect attempts table or the total attempts table(as i have done here) if the no of direct attempts is the same. I am trying out a comp. simulation for the same(but it is going to be in very very outdated QBASIC) I want comments from people on improving this system, will it really work, is it feasible to use it in real time etc etc. Also comments solicited on what the second table should be and what should be done if all the table are equal. Please do tell me if there are any flaws. There are a few which i would list out here :

1. Method too cumbersome - I don't subscribe to this point as it is not cumbersome for the participants or the audience but only to the scorer(who i feel have too easy a job with IR). The only problem will be the constant need of the scorer to communicate to the QM which team to ask the question next.Also if a comp is at disposal then the question of whom to answer the next question to becomes very easy with a simple program.

2. The problem of PMQs not completely and satisfactorily solved as this method quintessentially relies upon IR.

3. I personally do not have my simulator working and prob. wont have it working by day after. Therefore i yet have to test if the total no of attempts per team remains the same. If there is huge variation that will be a major fallacy in this method and probably will render it as good or as bad as the others. I will post my results again on sunday (provided my sim works). Lets keep our fingers crossed.

( PS : another idea is instead of using the direct questions table as given above we can directly give the next question to the team with the least no of questions attempted)

(PS: Just discussed this with a friend. He feels that for a quiz to be enjoyable the audience, and also the quizzers, scorers etc. need to understand the format. I don't agree)

:: Abhishek Nagaraj (cheerfulguy{at}rediffmail{point}com

No comments: