N5 scaled score vs actual score?

I originally thought for the N5 one just needed to score about 50% correct overall and more than 1/3 correct in any one section, but then someone pointed out to me that may not be the case, since JLPT uses scaled scores.
https://www.jlpt.jp/e/guideline/results.html

I am wondering if anyone has any data on this, since there is no data on the % difference in scaled score given to the most difficult question and the easiest question.

I mean, could one get a majority of questions right and still fail because some questions are worth 3-4x others? Or is the variation not that great like 0.9-1.1?

1 Like

This explanation of how the scaling works can be found two links away from the page you posted. I’m… not sure if it actually makes much sense, or if it’s just too late at night for me, but it’s something.

2 Likes

I’ve read a few discussions about this online and nobody seems to know how big the variation is exactly (I couldn’t find any specific numbers at least). I’m not sure if I’ve understood the whole scaled scoring stuff correctly either…
Theoretically, it might be possible to answer the majority of questions correctly but still fail, if your answering pattern suggests a low proficiency. I assume, that would have to include lots of lucky guessing or lots of ridiculously easy questions though. Both scenarios feel kinda unlikely. But idk. :sweat_smile:

2 Likes

The JLPT folks don’t release raw data, so nobody knows other than them. The chances are that the cross question variation between different questions of the same type won’t be too massive because that would imply that the question setters were asking inappropriately easy or difficult questions, and by this point they should have a big enough bank of questions and experience not to do that. But it seems plausible that answering a long reading comprehension question correctly is likely to be worth more than a single short grammar question. (JLPT prep books often have some kind of “points for different question types” scoring scheme, which isn’t how the real exam works but presumably isn’t massively mis-calibrated.)

The old 4-level JLPT used unscaled scores, and the pass mark on those was 60% for levels 2-4 and 70% for level 1. Since the new test is supposed to be at roughly the same difficulty, it might not be too far off the mark to guess that you should be getting over 60% right to end up at the scaled-down pass mark. But it’s all very hard to tell…

7 Likes

Online, people seem to be using the 60% figure around a lot, for everything except N1 which is more like 70%+. However from what you have written it seems possible that getting 65%+ of the questions right but getting derailed by a few select question that has more emphasis can still cost a candidate a pass.
Has anything official been posted, or even good guesses, about sections that one “must not get wrong”?

1 Like

Yeah my guess is that, in addition to the published minimum requirements for passing, there may also an unpublished scoring system that assigns different point values to different question types.

If true, that could mean that the minimum number of correct answers to guarantee a pass may be closer to 70% on N5-N2 and 75% on N1, assuming you answered them correctly in ascending order from lowest point value to highest.

1 Like

This topic was automatically closed 365 days after the last reply. New replies are no longer allowed.