What is charted here
Eight test lengths — 10, 20, 25, 30, 40, 50, 60, 100 — and every score on each of them, worked out to a percentage, a letter grade and a count of what was wrong. That is 335 pages, one per score, and each of them answers the follow-up question as well as the first one: what the next letter up would take, what a single question is worth, and what the same percentage looks like on a test of a different length.
These eight lengths are not an arbitrary range. They are the totals US tests are actually built on: ten- and twenty-item quizzes, the twenty-five and thirty question unit test, forty and fifty for a mid-term, sixty for a final, and the hundred-point scale most gradebooks normalise everything back to. Test lengths nobody sets — out of 37, out of 83 — are deliberately not here.
Why the test length changes the grade
A percentage looks like a continuous number and is not. On a test of 10 questions each answer is worth 10%, so only 10 grades exist above zero and most of the plus/minus bands cannot be reached at all. On a hundred-point test each answer is worth 1% and every band boundary falls exactly where the school says it does. The same understanding, tested twice at different lengths, can produce two different letters — not because the marking changed, but because the shorter test has fewer grades available to give.
| Test length | One question is worth | Reaches 93% | Reaches 60% |
|---|---|---|---|
| Out of 10 | 10% | 10 out of 10 | 6 out of 10 |
| Out of 20 | 5% | 19 out of 20 | 12 out of 20 |
| Out of 25 | 4% | 24 out of 25 | 15 out of 25 |
| Out of 30 | 3.33% | 28 out of 30 | 18 out of 30 |
| Out of 40 | 2.5% | 38 out of 40 | 24 out of 40 |
| Out of 50 | 2% | 47 out of 50 | 30 out of 50 |
| Out of 60 | 1.67% | 56 out of 60 | 36 out of 60 |
| Out of 100 | 1% | 93 out of 100 | 60 out of 100 |
Read across a row and the trade-off is visible: a ten-question quiz gives away a tenth of the grade per question, so one careless answer costs a whole letter, while a hundred-point paper lets a student be wrong seven times and still take an A. Neither is more accurate. They measure at different resolutions.