Methodology · plain English

How we predict — in plain English.

This page is for the student (or parent) who wants to understand what's under the hood before trusting our rank prediction. We'll walk through the inputs, the models, the uncertainty, and — honestly — what we don't know.

Try the rank predictor — Free See pricing

TL;DR

The short version.

We observe how you perform on our practice and mocks over time.
We compare your performance to historical cohorts whose eventual rank we know.
We produce a probability distribution of where you'll land.
We show you three points on that distribution: best-case, likely, worst-case.
We tell you how uncertain we are. Always.

Inputs

Four streams of data.

Your practice

Every question you attempt — DPP, mock, PYQ — contributes. We track correctness, time taken, difficulty, topic.

Your mocks

Full-length mocks are the highest-signal data. Each full mock gets weighted more heavily than daily practice.

Exam metadata

Target exam, category, home state, paper preferences. Rank depends on these because the exam does.

Historical cohorts

Trained on 15 years of rank-vs-performance data from students who eventually wrote the real exam.

P10, P50, P90 — what they actually mean.

When we show you "your rank window is 8,120 → 12,400 → 17,600", we mean:

P10 = 8,120 — 10% of students with your current profile end up at or above this rank. This is your best case.
P50 = 12,400 — the median. If we ran 100 simulations of your exam day, 50 would be at or better than this.
P90 = 17,600 — only 10% of students with your profile end up worse than this. Your reasonable worst case.

The ±800 figure next to your P50 is the 95% confidence interval of the median itself. Even P50 has uncertainty in it.

Honest limits

What this model can't predict.

✗Your exam-day performance. Everyone has variance.

✗Sudden syllabus shifts or pattern changes by NTA.

✗Your future effort — the model adapts after it sees the change.

✗Unusual biographical factors — family crisis, health issues.

✗Cohort noise — some years are genuinely harder than others.

Try it free.

One free rank prediction per season. Enough to see how the model handles your data.

Try free rank predictor See pricing

Loading…

The short version.

We observe how you perform on our practice and mocks over time.
We compare your performance to historical cohorts whose eventual rank we know.
We produce a probability distribution of where you'll land.
We show you three points on that distribution: best-case, likely, worst-case.
We tell you how uncertain we are. Always.

Four streams of data.