3 ms·

So the funny thing, looking say these from EU, Polish perspective, they are supremely easy and when applied here would have low statistical discriminatory power

So the funny thing, looking say these from EU, Polish perspective, they are supremely easy and when applied here would have low statistical discriminatory power. Our tests are new and statistically validated. That still does not mean they are good. Even our more advanced tests have issues there, so you get university admission packed with people in the top percentile, with single digit sometimes deciding whether you're in or out. It helps that pupils have a choice of their test for wherever they want to apply, so they generally pick what they're best at. Basics (language and maths) regularly get >80% except for special cases. Extended similarly. Special cases tend to completely flunk it, that usually being less than 15%. Latest COVID bumped it up somewhat. The problem is not really the testing, it's what the tests are testing for. The maths basics test some rote memorization of basic trigonometry and application thereof, including linear equations. Language test reading comprehension and whether you read a book in the syllabus, plus can write mostly coherently and logically, plus some analytical techincalities. The reading comprehension on basic is literally on the level of newspaper column. Extended does not test an unfamiliar technical text either. Physics and chemistry likewise test mostly rote pattern application. History, geography, even more so. And the problem is that university requires more than that. You start getting more open ended maths tasks, unfamiliar language in unfamiliar subjects in books... None of them tests basics of being able to learn, as that is not taught. (As to test results, there's a clear gradation between academic middle school, technical middle school and finally faulty system of job middle schools. It's mostly about filtering and focus of the pupils there. They tend to get worse grades too. Usually people get failed on maths.)

WalterBright 4y ago
I know that the tests are all considered "not good". But I've found a consistent pattern that people who fail those tests don't understand the material. The people who do understand it do well on the tests. The tests work. At Caltech, the tests were open book open note. That means that memorizing rote rules did not work at answering the test questions. > with single digit sometimes deciding whether you're in or out That's inevitable when you've got 100 people applying for 10 positions. The same thing happens in sports, where one thousandth of a second distinguishes the winner from the forgotten loser.
- eru 4y ago
  > That's inevitable when you've got 100 people applying for 10 positions. The same thing happens in sports, where one thousandth of a second distinguishes the winner from the forgotten loser. This is not universally true. Eg the world's richest and second richest (or 10th and 11th richest) persons differ in their wealth (or income etc) by more than a fraction of a penny. Details depend on the underlying distribution.

WalterBright 4y ago

I know that the tests are all considered "not good". But I've found a consistent pattern that people who fail those tests don't understand the material. The people who do understand it do well on the tests. The tests work. At Caltech, the tests were open book open note. That means that memorizing rote rules did not work at answering the test questions. > with single digit sometimes deciding whether you're in or out That's inevitable when you've got 100 people applying for 10 positions. The same thing happens in sports, where one thousandth of a second distinguishes the winner from the forgotten loser.

eru 4y ago
> That's inevitable when you've got 100 people applying for 10 positions. The same thing happens in sports, where one thousandth of a second distinguishes the winner from the forgotten loser. This is not universally true. Eg the world's richest and second richest (or 10th and 11th richest) persons differ in their wealth (or income etc) by more than a fraction of a penny. Details depend on the underlying distribution.