CAP Priority List

A practical helper for Maharashtra CAP engineering admissions. Filling the option form means ranking dozens of college–branch pairs under time pressure. This project turns published CAP cutoff history into a short, ordered preference list tailored to one candidate.

You provide a CET Final Merit Status page (recommended) or an MHT-CET scorecard. The tool reads category, gender, percentile, and — when available — PCM State General Merit No. It then compares those against three years of CAP Round 1–3 closings already parsed into data/cutoffs_db.csv.

The result is a CSV of about 100 college–branch options split into aspirational, moderate, and safe buckets. Government, government-aided, and university institutes are preferred when they qualify. Use it as a starting preference list — not a guarantee of allotment. See Assumptions below for every modeling choice.

CLI · generate_recommendations.py · open-source on GitHub

Inputs

Merit .mht (preferred) or CET scorecard PDF + raw CAP cutoff PDFs

Output

output/recommendations.csv — priority-ordered choices

Who it’s for

Candidates (and families) preparing CAP preference forms for FE engineering

01→04

Pipeline

Parse DB → read candidate → filter → score & shortlist

Assumptions

What this tool assumes · click to expand · 16 modeling choices · planning aid, not a guarantee

These are deliberate modeling choices. The shortlist is a planning aid, not an allotment prediction.

Past closings ≈ chance. Historical CAP Stage-I closings are a useful signal for next year. Applicant pools, seat matrices, and rules can still change.
Stage-I only. Later stages / special allotments in the PDFs are ignored for scoring.
Easiest eligible seat per year×round. Among seats the candidate can claim in that round, the tool keeps the easiest closing (best for the candidate) and scores against that.
Category + OPEN. Eligible codes = the candidate’s reservation category seats (e.g. DT/VJ → GVJ*) plus OPEN (GOPEN*). Separate EWS / TFWS / PWD / Defence paths are not modeled beyond what those standard category codes already cover.
Gender. Male → G* seats only. Female → G* and Ladies L* seats.
Home University is not applied. HU is read from the merit page when present, but H / O / S cutoffs are pooled. Regional home-seat advantage is not enforced yet.
Merit mode uses PCM State General Merit No. When a Final Merit Status file is provided, matching uses that state general merit number — not category merit, university merit, or All India merit.
Cutoff merit_rank = State General Merit scale. CAP PDF closing ranks are treated as comparable to the candidate’s State General Merit No (lower rank is better; cleared if candidate ≤ closing).
Percentile fallback. Scorecard-only runs match on percentile (cleared if candidate ≥ closing %). Difficulty scores always use percentiles so ordering stays comparable.
Fixed time weights. Years: 2024–25 = 0.50, 2023–24 = 0.30, 2022–23 = 0.20. Rounds: CAP1 = 1.00, CAP2 = 0.75, CAP3 = 0.55.
Bucket thresholds are heuristics. Safe / moderate / aspirational cutoffs (clear rates, merit gaps, percentile margins, reachability slack) are tuned rules of thumb, not CET Cell policy.
Selected streams only. Branches outside the five chosen streams (CS/IT, Electronics, Electrical, Mechanical, Civil) are excluded.
Ownership preference. Within a bucket, fill Government → Government-Aided → University before Private / Un-Aided when those options qualify.
No reputation ranking. Order is cutoff competitiveness + ownership bias — not NIRF, placements, fees, or location preference (except what cutoffs imply).
Multiple branches from one college are allowed. The shortlist does not force one row per institute.
Ultra-easy “filler” safes are dropped. Options far weaker than the candidate are removed so the safe bucket stays useful as real preferences.

read these before treating the CSV as final preferences

Recommendation logic

How matching works

Per choice code (college–branch), each year×round keeps the easiest eligible historical closing. Then: would this candidate have cleared it?

Merit mode: State General Merit No vs CAP closing ranks (lower rank = better; cleared if candidate ≤ closing)
Percentile mode: scorecard fallback (cleared if candidate ≥ closing %)

primary signal = merit when .mht provided

Eligible seats

Category + OPEN
e.g. DT/VJ → GVJ* + GOPEN*
Gender: male G* only; female also L*
HU: not applied (H/O/S pooled)

filters before scoring

Difficulty

year:  .50 / .30 / .20
round: 1.0 / .75 / .55

difficulty =
  Σ(pct × yw × rw)
  / Σ(yw × rw)

orders options inside a bucket

Aspirational

Stretch — near history, low clear rate

×30

Moderate

Realistic — clears a fair share of past rounds

×40

Safe

Comfortable CAP1 clearance + margin

×30

Shortlist

Govt → Aided → Univ → Private

~100 rows

Bucket rules

Merit-mode thresholds

Safe: CAP1 clear ≥ 67% and median gap ≥ +3,000
Moderate: clear ≥ 35%, or CAP1 ≥ 33% with gap ≥ −2,000
Aspirational: reachable ≥ 25% and gap ≥ −8,000
Gap = closing_merit − candidate_merit (positive = stronger candidate)

Percentile-mode thresholds

Safe: CAP1 clear ≥ 67% and median ≤ candidate − 3
Moderate: clear ≥ 35%, or CAP1 ≥ 33% with median ≤ candidate + 2
Aspirational: reachable ≥ 25% and median ≤ candidate + 8
Hopeless / ultra-easy filler options are dropped

Run

Command line

pip install -r requirements.txt
python scripts/parse_cutoffs.py
python generate_recommendations.py "candidate.mht" --streams all
python generate_recommendations.py scorecards/scorecard.pdf --streams 1,2

flags: --streams · --aspirational/--moderate/--safe · --female/--male · -o

Not a guarantee

Past closings only — pools change yearly

No HU filter

Home University not applied yet

No reputation rank

Cutoff strength + ownership bias

Merit first

.mht State Merit drives clear/bucket