StartSprint Learning Intelligence
Calibration edition
Monthly analysis of learning patterns, calibration data, and misconceptions across UK classrooms.
This month, students across StartSprint completed 762 quiz sessions. The headline number is not the score. It is this: 34% of all wrong answers were confident — answered fast, without hesitation, without any attempt to change.
A student who answers in under 2 seconds and gets it wrong every time is a different kind of problem than a student who gets it wrong after 8 seconds of thought. This report is about the first kind.
That distinction is what StartSprint measures — not just whether a student got the answer right, but how they arrived there. Response time, answer changes, IDK selection, and hesitation patterns together form a calibration picture that raw scores cannot show. What follows is a snapshot of March: where students were confidently wrong, where uncertainty told the truth, and where reteach is most urgently needed.
34%
Wrong answers submitted in under 2s with no changes
22%
Wrong answers with hesitation or answer changes
11%
IDK selections — students know they don't know
+3.4s
Avg time difference between wrong and right answers
Why this matters
Why this split matters for teaching: overconfident wrong answers rarely get revised without intervention — the student does not feel the gap. Uncertain wrong answers self-correct with feedback. IDK selections reveal honest knowledge gaps. Three different signals, three different classroom responses. A score alone cannot tell you which is which.
Fraction equivalence — confident wrong dominates. Procedural fluency outpaces conceptual understanding.
Forces and motion — uncertainty spikes on Newton's third law. Students hesitate rather than guess.
Language technique identification — healthy signal split. Reteach load is moderate.
Tectonic plate boundaries — highest IDK rate of any subject. Students know they don't know.
Cause-and-effect sequencing — overconfidence on dates, uncertainty on consequences.
Average response time in seconds per answer category. Confident wrong answers are faster than correct ones — the pattern-match is instant.
In the classroom
What this means in the classroom: when a student answers a question incorrectly in under two seconds, they are not guessing — they are pattern-matching to a confident but wrong mental model. This is slower to detect than outright struggle, because the student shows no hesitation. Traditional marking sees only the wrong answer; it does not see the confidence behind it. The intervention here is not more practice. It is direct correction of the underlying model.
The invisible student
The invisible student: this is the student who answers quickly, appears engaged, and still fails the assessment. Their wrong answers look identical to a distracted student's wrong answers on paper. Only response time reveals the difference. StartSprint surfaces this signal before the test does.
67%
of teachers who received a March misconception signal reteached the topic within two weeks. The signal turned into action.
+19%
increase in IDK selections over the last three months. Students are increasingly willing to say they don't know — which is the first step to closing a gap.
Methodology: each misconception must appear in at least 3 classes and 30 responses in the reporting period. Overconfidence rate = percentage of wrong answers submitted in under 2 seconds with no answer changes. Correct rate = percentage of all attempts answered correctly.
How to read this list
How to read this list: rank is by overconfidence rate, not by error rate. A high-error question where students hesitate is learning. A medium-error question where students answer fast and wrong is a misconception. The second case is what this list surfaces.
Wrong answer most chosen
6/9
Wrong mental model
Students added 3 to both numerator and denominator rather than multiplying. The procedural error reflects a confusion between equivalent fractions and simple arithmetic progression. The fix is explicit instruction: "equivalent means both parts scale by the same factor."
Wrong answer most chosen
The weight of the book
Wrong mental model
Students identify the weight of the book as both the action and the reaction force — treating Newton's third law as describing balance on a single object. The correct mental model pairs forces across two objects: the book pushes the table, the table pushes the book back.
Wrong answer most chosen
Simile
Wrong mental model
Students default to "simile" as a catch-all for any figurative language. The absence of "like" or "as" is the diagnostic signal, but students treat any imagery as a comparison. Personification requires explicit contrast against simile and metaphor.
Wrong answer most chosen
40
Wrong mental model
Students compute area instead of perimeter — the two formulae are procedurally adjacent, and under time pressure the wrong one fires first. The mental model to reteach is "perimeter is around the outside, area is inside."
Wrong answer most chosen
Constructive
Wrong mental model
Students associate "constructive" with building up, and map that onto volcano formation. The boundary type that generates stratovolcanoes is destructive — subduction melts crust. The conflict is with the intuitive meaning of the word, not the geology.
Wrong answer most chosen
Liver
Wrong mental model
Students conflate the liver's role in blood sugar regulation (glycogen storage) with insulin production. Both relate to glucose, so the brain bundles them. Explicit separation of glucose storage (liver) versus hormone production (pancreas) is the intervention.
Wrong answer most chosen
1918
Wrong mental model
Students anchor on the most memorable date associated with WWI — the end date, 1918 — and return it when asked about the start. Temporal confusion between start and end dates is common under recall pressure. The fix is paired dating: "1914 to 1918."
Wrong answer most chosen
10
Wrong mental model
Students subtract 6 from 14, get 8, then skip the divide-by-2 step. The mental model treats the equation as a one-step operation. Emphasising "inverse operations in order" and rehearsing two-step problems is the correction path.
Wrong answer most chosen
To end a paragraph
Wrong mental model
Students confuse topic sentence with concluding sentence. Both are "important sentences" in a paragraph and both are taught together, which creates interference. Explicit placement instruction ("topic sentence opens, concluding sentence closes") helps separate them.
Wrong answer most chosen
Coral reef
Wrong mental model
Students choose coral reef because biodiversity is often taught alongside reef ecology. The correct answer is tropical rainforest, which has both higher species count and greater structural diversity. Competing correct-adjacent answers create the confusion.
Mathematics shows the highest overconfidence rate this month, driven almost entirely by fraction equivalence (41% confident wrong). 12 classes affected, response times under 1.5 seconds — students are pattern-matching to a wrong procedure without knowing it.
Recommendation
Deliver a targeted reteach using concrete visual models before returning to symbolic notation. Use a StartSprint quiz with three-option mini-quizzes focused on the additive-versus-multiplicative distinction. Remeasure in two weeks.
Science uncertainty rate rose from 24% in February to 31% in March. Students are increasingly hesitant on action-reaction pairs. This is a healthy signal — they know there is something they are missing — but if left alone it will harden into overconfidence.
Recommendation
Introduce paired-object diagrams and reinforce the cross-object nature of Newton's third law. The window is now — uncertainty is the moment before confident error.
English shows the most balanced calibration profile: 24% overconfident wrong, 22% uncertain wrong, 9% IDK. This is the healthiest distribution across all subjects. Signal-to-correction turnaround averaged 9 days — the fastest of any department.
Recommendation
Maintain current practice. Document the department approach (question design, feedback cadence, peer review) and share with the maths and science leads.