I checked all five of my scripts. Physics had ten half marks, Chemistry eleven, English four, Maths two. Computer Science had none at all — every chip on the page is a whole number.

That changes the calculus of a half-finished answer. In Physics a partial derivation still banks fragments. Here a function that does not work is more likely to score nothing, and a function that works scores the mark even if it is inelegant.

So the priority is different: get something that runs, then make it neat. On my four-mark file-handling answer there are five separate ticks against five separate lines — the import, the open, the loop, the condition, the close — and each one is a whole mark being confirmed.

SQL is twenty marks of the seventy, and it is the most predictable content in the subject. The same handful of query shapes come round every year — SELECT with WHERE, GROUP BY with aggregate, ORDER BY, and a join — and once you have written thirty of them you are not going to meet a new one.

On my script every SQL question and its output table is correct. It is not because I am good at SQL, it is because there is not very much of it to be good at.

On seven either/or questions I wrote out both options. The summary on page one scores both and then discards one — the footnote says asterisked marks are omitted from totalling, and the unasterisked rows add up to exactly 68, which is the printed total.

On my paper both alternatives happened to score the same, so the script cannot tell you which one CBSE keeps. What it does tell you is that the second attempt cost me time I could have spent checking the rest of the paper.