All documentation

Exam Score Prediction

A predicted raw score band for JEE / NEET / CAT / UPSC / CUET — and the exact topics that would add the most marks if you fix them.

Students5 min read
Why predict raw score, not rank?
Rank depends on the whole exam-taking population. Raw score depends only on you. The score predictor isolates your readiness from external noise.

Overview

The predictor outputs a min-max score band against the maximum possible marks for that exam. It blends two signals:

  • Test performance (60% weight) — your average percentage across the last 20 graded test attempts.
  • Subject mastery (40% weight) — average topic mastery scores from your Learning Profile.

How the score is computed

avgPct       = mean of (totalScore / maxScore × 100) across last 20 attempts
avgMastery   = mean of every topic score in your masteryMap
blendedPct   = avgPct × 0.6 + avgMastery × 0.4
predictedMid = round(blendedPct / 100 × maxPossible)
band         = round(maxPossible × 0.08)
score range  = [predictedMid − band, predictedMid + band]

The ±8% band tightens implicitly as you collect more attempts (theconfidence field rises from 0.50 toward 0.90).

Requirements

  • At least one completed diagnostic test (your mastery map must exist).
  • At least one graded test attempt — DPP, batch test, or mock test all count.

If either is missing, the predictor returns an explanatory message instead of a prediction.

Per-exam max scores

ExamMax marks
JEE Main300
JEE Advanced360
NEET720
UPSC (CSE prelims paper-1)200, but predictor uses 2025 for full GS+CSAT band
CAT300
CUET800

Unknown exam types default to a 300-mark scale. The mapping lives in the predictor as EXAM_MAX_SCORES — add a new entry when you add a new exam type.

Weak areas & lift

Alongside the score band, the predictor surfaces up to 5 weak-area topics: topics in your mastery map below 60. For each, it shows the additional marks you'd plausibly gain by bringing that topic to mastery — currently a flat 3% of max marks per topic.

Click any weak topic to jump to relevant flashcards or batches that cover it.

Refresh & caching

Predictions are cached for 24 hours per (user, exam) pair. The dashboard shows the generated timestamp. Click Refresh to force a fresh prediction — useful right after a new mock attempt.

Refresh costs nothing
Score predictions are computed locally, not via an external AI. Refresh as often as you like — but cached values are perfectly accurate if no new attempt has been graded.

Header Logo