How RankMatrix predicts JoSAA cutoffs

Why this page exists

A predictor that says “you will get IIT Bombay CSE” without showing its work is just confidence wrapped in a UI. RankMatrix tries to be the opposite of that: a predictor that tells you exactly what data it is using, what assumption it is making, and where its answer is least trustworthy. This page is the long version of that contract.

The data the predictor stands on

Every projection on RankMatrix derives from three official JoSAA datasets, refreshed each counselling cycle:

• Historical opening and closing ranks — for each (institute, branch, category, gender pool, quota, round) tuple, JoSAA publishes the rank of the first and last candidate to whom that seat was offered. RankMatrix ingests these for the most recent counselling years.
• Seat matrix — for the current cycle, JoSAA publishes how many seats each institute has per branch, broken down by category and quota (HS / OS / AI / GN / OBC-NCL / SC / ST / EWS, with PwD and gender-neutral / female-supernumerary sub-pools).
• Participating institutes — the canonical list of IITs, NITs, IIITs, and GFTIs running through JoSAA in the current cycle.

You can browse the underlying data directly on Opening & Closing Ranks, Seat Matrix, and Participating Colleges.

How the predictor turns your rank into a list

You give the predictor a small set of inputs: your rank (JEE Main or JEE Advanced depending on which institutes you are targeting), category, gender pool, quota (home state vs other state for NITs), and any branch or institute filters you want to apply.

The predictor then does five things, in order:

1. Loads the relevant cutoff slice. For your category and gender pool, it pulls every (institute, branch, round) row from the historical cutoff dataset that matches your eligibility.
2. Compares your rank against each closing rank. The closing rank is the rank of the last candidate who got that seat in the round under consideration; if your rank is at or better than the closing rank, that seat was historically reachable for someone in your slice.
3. Buckets each seat by reachability. Every (institute, branch) pair ends up in one of three buckets relative to your rank — comfortably reachable, on the boundary, or out of reach. The boundary bucket is where most of your real decisions live, and that is where the predictor focuses the UI.
4. Cross-checks the current seat matrix. Some institute-branch pairs from past cycles may have changed seat counts or no longer exist; the predictor drops options that have zero seats in the current matrix.
5. Ranks the result. The default ordering is by closing rank, so the most competitive (institute, branch) you are likely to reach surfaces first. You can override the ordering and filter by institute, branch family, or location.

Which round's closing rank counts

JoSAA runs five to six rounds. The closing rank at a given (institute, branch, category) slides round to round as candidates accept, freeze, float, or slide their seats. Three conventions you will see across predictors:

• Final-round closing rank — the most permissive number. This is the rank of the last candidate who got the seat after every round of churn. Predictors using this make pages look more optimistic; in practice, very few people actually wait for the last round to lock in a seat.
• First-round closing rank — the most conservative. This is what would have been allocated if everyone froze on day one.
• A specific round — for example, round 4. This is what RankMatrix uses as the default. Round 4 historically reflects roughly where most candidates have either committed or moved out, and it tracks much closer to the rank distribution that actually matters when you are deciding to lock or float.

You can switch the round the predictor uses for its threshold from the predictor inputs. For a deeper read on what these numbers mean, see the opening and closing ranks guide.

What the predictor is not modelling

A historical-cutoff projection is a useful starting point, but it is structurally incapable of capturing things that move year to year. The predictor does not know:

• This year's applicant pool. If a particular branch becomes disproportionately popular in a given year — say, AI/ML across the system — its closing ranks will tighten beyond what last year suggests.
• New branches or new institutes.First-time offerings have no history to extrapolate from. The predictor will mark those as “no historical data” rather than guess.
• Mid-cycle seat-matrix changes. Seat counts can shift between rounds (supernumerary additions, withdrawals). The predictor uses the matrix snapshot it has ingested.
• Personal-eligibility nuances. Domicile rules for NIT home-state quotas, PwD certification, EWS recency — the predictor uses the category you tell it you are claiming and assumes the documentation is in order.

How to read predictor output without overcommitting

• Treat boundary-bucket seats as a discussion list, not a forecast. Anything within roughly five to ten percent of your rank on either side of last year's closing rank is genuinely on the bubble.
• Cross-check trends across rounds. A seat where the closing rank has been steadily tightening for three years is more likely to keep tightening than a seat with a one-off bad year.
• Use the seat matrix. A branch with one seat per category at an institute is statistically much more volatile than a branch with twenty seats. Volatility matters.
• Verify on the official portal before locking. RankMatrix exists to help you build a sensible preference list; the actual choice locking, document verification, and seat acceptance must happen on josaa.nic.in.

How the data is refreshed

When JoSAA publishes a new cutoff release or seat-matrix update, the underlying datasets are re-ingested and the relevant pages on RankMatrix are revalidated within a day. The “last updated” line on each detail page reflects the most recent regeneration. If you spot a discrepancy with the official source, the contact page is the fastest way to flag it.

Disclaimers, briefly

RankMatrix is not affiliated with JoSAA, JEE, NTA, the IITs, NITs, IIITs, or GFTIs. Predictions are projections from public data, not guarantees of admission. See the terms of use for the full version, and the privacy policy for what the site does and does not collect from you while you use it.