Where the numbers come from

A closing cutoff is a record of what happened, not a threshold set in advance. It is the rank and percentile of the last candidate who accepted a seat in that course, under that seat type, in that round. Nobody decided it. It fell out of how many people applied, what they preferred, and how many seats there were.

Each row carries two figures and they are not interchangeable:

• Closing percentile — your MHT-CET score expressed as the percentage of candidates you finished at or above. Higher is better. 99.42 means roughly the top 0.58 per cent.
• Closing rank — the merit number in the CAP list. Lower is better. This is the merit rank from your CAP allotment documents, not your position in a coaching class test.

The predictor never converts one into the other. The official lists publish both, and deriving a rank from a percentile would produce a number that looks precise and is a guess. If you enter a percentile, you are compared against closing percentiles only; if you enter a merit rank, against closing ranks only. Rows that have no value in the mode you chose are left out of the result rather than estimated.

Each row also carries a seat type (your category, gender and whether the seat is reserved for a defence, PWD, TFWS, orphan or minority candidate), a seat level (Home University, Other Than Home University, State Level, All India, Minority) and a stage. Those three decide which rows are yours. A GOPENS figure at a Pune college is irrelevant to an OBC candidate from Nagpur, which is why the form asks for category and university scope before it shows you anything.

The cutoff documents are wide tables in PDF, and reading them mechanically goes wrong in specific ways. Three of them mattered:

• Continuation pages. Tables too wide for a page put the remaining columns on the next page with no headers. Fourteen pages did this. Missing them would have silently dropped the most competitive colleges’ figures.
• Columns match by position, not by order. A stage often publishes values for only some of the categories in its header, so the third number is not the third category. Values are matched by their horizontal position on the page.
• Stage labels wrap. “I-Non” on one line and “PWD” on the next is a single label, I-Non PWD. About 6,700 rows in Rounds II and III are affected.

Extracted row counts were then checked against an independent count of percentile tokens taken from the same PDFs with different software: 36,059, 34,391 and 19,839 for the three rounds, matching exactly. Three deliberately awkward records were also read off the PDF by hand and compared. The extractor reports anything it cannot align rather than guessing; that report came back empty, with 0 anomalies and no blocks dropped.

There is no model and no machine learning. The calculation is a comparison, and it is deliberately simple enough to explain in a paragraph.

• Rows are first filtered to the seat types that apply to you — category, gender, and Home University or otherwise — and to the rounds and branch groups you selected.
• For each remaining row, your figure is compared with that row’s closing figure. In percentile mode the margin is your percentile minus the closing percentile. In rank mode it is the closing rank minus your rank. Either way, a positive margin means you finished ahead of the last candidate admitted.
• The margin decides the band. Comfortably ahead is good chance; close to the line, on either side, is possible; behind it is reach. In rank mode the thresholds are proportions of the closing rank rather than fixed numbers, so the same setting behaves sensibly at a cutoff of 900 and at a cutoff of 190,000.
• Results are sorted by band, and within a band by how competitive the course is — not by how easily you clear it. The most selective college you have a real shot at appears first, because that is the one you are trying to find.

The comparison is against a previous year. It is evidence, not a forecast, and there are real reasons it will be wrong for some students:

• Cutoffs move every year. Seat intake changes, new institutes are approved, a branch becomes fashionable, the paper is harder. A course that closed at 96.2 last year can close two points either side of that this year.
• Allotment depends on your preference list. CAP allots the highest preference it can reach, so a course you had a good chance at but ranked twelfth may never be offered to you.
• Eligibility rules are not modelled. Domicile, subject combinations, minimum marks, documentation and category certificate validity are decided by the CET Cell, and this site checks none of them.
• Institute-level location is derived from each institute’s registered name, which is all the cutoff documents carry. It is filled in for most colleges and blank for the rest, and is not a confirmed address.
• Fees, placements, hostels and infrastructure are not in these documents. Where a college page shows them, a person checked them against a primary source; where it does not, the page says so rather than filling the space.

Use the result as a shortlist to research and to order your preference form sensibly. Do not use it as a reason to leave a college off that form.

If a figure here does not match the official document, that is a bug and we want to know. Tell us the institute code, the course, the round and the seat type, and it will be checked against the source PDF and corrected. Contact details are on the about page.