Your argument seems to imply that the only thing that an exam is possibly allowed to measure is ability to answer a question at some arbitrary point in time. It also separates the people with the wit to understand a bunch of topics in a span of 3 or 5 months, the ability to plan accordingly, enough self control, and some motivation to see the plan through from the ones that don't. That's not a bad thing, as diplomas are often used as a proxy not only for intelligence, but also for suitability to certain jobs.

▲

viraptor 5 hours ago | parent [-]

The only thing I'm saying is that measuring once maximises the luck component. Sure, measure other things by proxy as well. But to borrow your phrase, understanding and taking into account the variability of real life behaviours, sampling noise and environment biases shows "not only intelligence, but also for suitability to certain jobs."

▲

tgv 3 hours ago | parent [-]

> maximises the luck component

No, it doesn't. Repeating it ad infinitum does that.

But therefore it's a good thing grades aren't just pass or fail, but have a range, from 1 to 10, or to 20 or whatever. Too bad there's so much external pressure on the system.

	▲	viraptor 2 hours ago \| parent [-]
		> Repeating it ad infinitum does that. Take a moment to think about it. Why does the confidence interval go smaller with more samples? What does it say about the impact of luck in repeated tests?