A student who can do the algebra still loses the question if they answered the one they thought was being asked. It looks like a content gap on the report card, so it gets treated with more content — more practice on the thing they could already do.
One question at a time, and it stops before the answer. Read the stem, name the command word, say in your own words what is actually being asked, and choose the opening move. That is the whole drill — no working out, no final answer, no marks for arithmetic.
Stripping the calculation out is the point. A student who does forty of these in the time it would take to grind through six full questions gets forty repetitions of the step they are actually losing marks at.
The right opening move guarantees at least partial credit even when the rest goes wrong. The wrong one guarantees zero no matter how clean the working is underneath it — a perfectly executed answer to a question nobody asked.
And it is the fastest thing on a paper to get better at. Content takes a term. Reading the question accurately takes a couple of weeks of short runs.
The stems come from the same question library the rest of the app uses, in the courses the student is actually enrolled in — so the command words, the phrasing and the topics are the ones on their exam, not generic reading practice.
Every run feeds the academic profile. If recognition is where a student is losing marks, that is what the profile says and what the practice keeps pointing at; if it is already solid, it stops asking and spends the time elsewhere.
Knowing the material and knowing what the question wants are two different skills. Exams test both. Only one of them gets taught.