About Rules

May 22, 2005 23:08

An interesting excerpt from "... Essays on Graduate Education" by David Merrill ( Read more... )

Leave a comment

Comments 4

nvivek May 23 2005, 10:54:32 UTC
You could always make do with thumb rules, but the question is, do you trust the person responsible for interpreting them? If you need to standardize, you need a hard'n'fast rule.

Come up with a better idea and you'll revolutionize society!

Reply

Not thumb rules aramanath May 23 2005, 12:57:29 UTC
The intention is not to leave things open to interpretation. What I am saying is that no single criterion should by itself be powerful enough to exclude or include. The choice of parameters and the model used to combine parameter values to give a single score should be standardized and transparent.

This will help because deviation from a minimum expected score corresponding to a certain parameter (60% in HSC, say) will be penalized or rewarded in proportion to the deviation. So, a person can compensate for a shortfall in one parameter by being better in a combination of certain other parameters.

The worry of course is that if the model is complicated then we will have a situation similar to rain curtailed cricket matches: the D-L or some other model throws up a score that a team has to get, and since no one really understands the mathematics (except the creators of the model), the losing side always complains. This is absolutely normal human behaviour, and so long as the results of using the model support intuition, it should be

Reply

Re: Not thumb rules nvivek May 24 2005, 03:59:01 UTC
You pointed out the problem yourself. Does using the model support intuition? Remember the India-Australia match in the World Cup when they cut short three overs and reduced the target only by three runs or the even more embarassing South Africa-England semifinal?

Pretty tough to meet your criteria for a good model and still be simple, elegant and correct!

Reply

Re: Not thumb rules aramanath May 24 2005, 05:25:40 UTC
I had these matches (1992 World Cup) in mind. Both these examples are from the pre-D-L era, and D-L so far has been more or less satisfactory. So, with a little thought, it should be possible to develop a model that supports intuition (to a greater extent than is being supported now).

Again, the key issue is whether multiple factors are being combined or not. If D-L simply said that the team with more runs at the stage of match suspension wins, then we would have had a problem (200/9 chasing 250, and the other team had 190/3). For D-L, the parameters are (AFAIK) number of runs, overs, and wickets. Note that it does not look at the overs and wickets qualitatively (Tendulkar's vs. Kumble's wicket), and that is where it can be improved. This added complexity might introduce loopholes that teams might start exploiting, but the idea might be worth trying.

Reply


Leave a comment

Up