How we check
A practice question with a wrong answer doesn’t just waste your evening — it quietly mis-calibrates what you think you know. So every StudyCove question runs a pipeline designed by people who assume they make mistakes. Here is the whole thing, including the parts that aren’t finished yet.
Every question starts from a blueprint tied to one of the 112 subtopics of the Edexcel specification, with a target difficulty and command words drawn from real Pearson house style. Nothing is free-styled: coverage, tariffs and calculator rules are set before a word is written.
Every answer a computer can compare symbolically is re-derived and pinned by a computer-algebra system (SymPy) — if the stated answer and the derivation disagree, the question cannot pass, full stop. The few answer forms a computer can't compare (a sketch, a worded argument) can never earn the checked badge at all; they stay at the in-verification tier below.
Three independent solvers — separately configured, unable to see the author's solution or each other's — sit the question cold, exactly like a student would. A question only earns its badge when the panel's answers agree with the verified ones.
A marker then takes each blind solution and marks it against the encoded mark scheme, node by node. If a legitimate method can't be awarded full marks — say the scheme hard-codes elimination and a solver used substitution — the question fails and goes back for revision. This is the check most question banks never run.
Every draft is screened against the rest of our bank so near-duplicates never ship. Screening against real past papers is the next layer: it runs against a licensed paper corpus before public launch, and until then we say so here rather than claim it.
A separate rater — shown only what a student would see — places each question on our difficulty ladder. If the rating disagrees with the author's intent by more than one tier, the question can't be badged as checked.
Our modelled residual-error estimate — the chance a wrong answer slips past every layer — is about 0.03% per accepted question. That number is an estimate built from published per-check catch rates, not a measurement; the honest track record lives on the errata ledger (currently 0 published).
The question is retired immediately and a permanent entry lands on the public errata page within 24 hours of confirmation — with credit to whoever caught it. Attempts pin the exact question version, so your history always shows what you actually saw.
Original questions written for the Pearson Edexcel A Level Mathematics (9MA0) specification. Not affiliated with or endorsed by Pearson.