Overview of the chances work (not a work item itself). Written 2026-10-10 from the owner's input on grade inflation and holistic admission, the planner's built standing (planner redesign, PR #105), the planned Chances and Fit: Where You Stand, and the site's data that day. Each part below is a separate work item on the roadmap.
Why
The planner sorts every college on a student's list into Reach, Target, or Likely, and the profile will show the same "Where you stand". At a growing number of colleges most admitted students have nearly the same top GPA, so a GPA and one test score no longer separate applicants well; colleges lean on course rigor (read against the high school), the subjects that matter for the major, tests, and the pool the student is applying from. This work gives the estimate those inputs and shows families the evidence behind it.
Quad's estimate is proprietary
Owner decision, 2026-10-10: how Quad turns inputs into Reach, Target, or Likely is a trade secret, like Zillow's Zestimate. It will change as the site collects more data and outcomes.
| Shown on the site | Never shown on the site |
|---|---|
| The kinds of inputs the estimate uses (below), and which of the student's inputs were used for a given college | How inputs are weighted, combined, or thresholded; rules, cutoffs, models, and their parameters |
| The output: one of five levels, High reach, Reach, Target, Likely, or Safe (levels.md; three until that ships), with the labels "Elite: a reach for everyone", "Guaranteed for you", and "Recruited" | Training data, tuning, the method's version history, and the hidden odds behind the level |
| The college facts shown beside it, each cited as everywhere on the site (admit rates, ranges, factor ratings, required courses) | Internal accuracy diagnostics beyond the published summary |
| A published accuracy summary each season (calibration.md), as Zillow publishes its error rates | |
| The commitments: never a percentage; never race, ethnicity, sex, or religion as inputs; every individual in the outcomes anonymized; legacy only where a college says it weighs it and the state allows it; the high school not used in this version, and said so when it is (owner, 2026-10-11; inputs-beyond-academics.md) |
What follows from it:
- On the site, inputs and outputs only. Pages, release notes, and the glossary say what kinds of information the estimate uses and which of the student's inputs it used, never how they are combined.
- The estimate is computed on the server. No method code ships to the browser; pages and the planner ask an endpoint for the result (estimate.md).
- Until launch, the method lives in this repository, in method/ and server-only code. Owner decision, 2026-10-10: the repository stays public before launch (no paid GitHub plan yet) and nobody is expected to look; the method is not hidden for now.
- At launch, it moves to a private repository and reaches the site as a private package behind the same interface (the launch step). Keeping the interface, the server-only rule, and the note catalog from the start is what makes that move a small change.
- The site's citation rule still holds for every fact shown with the estimate; the estimate itself is labeled Quad's estimate with its inputs listed, never presented as a college's figure.
The method (until launch)
| File | What it holds |
|---|---|
| method/standing.md | The rules: which evidence counts at a college, GPA crowding, the rigor, rank, and major positions, the pool's rate, the thresholds, pinned examples |
| method/rigor-reading.md | How a schedule maps to a rigor reading, and the grades rule |
| method/major-effects.md | How a unit's major review changes the estimate |
| method/outcomes.md | How outcomes are measured each season, threshold changes, and the gate any trained model must pass |
| method/proposal-review.md | The review of the owner's grade-inflation model: what was adopted, adapted, and rejected, with measurements and sources |
| method/ideas/grade-inflation-acceptance-model.md | The owner's input document, as written |
The inputs
| Kind | Examples | Spec |
|---|---|---|
| The student's academics | Unweighted GPA (built), one test score and its sections (built), class rank (built), the advanced courses taken and planned with grades, math and science grades for STEM majors | rigor-in-context.md, major-and-grades.md |
| The student's context | State, intended major, application round, high school (for what it offers) | base-rates.md |
| The high school | Advanced courses offered (CRDC for 13,191 high schools; school profiles), grading scale, rank policy | rigor-in-context.md |
| The college's own filings | Admit rates (overall, by residency, by major where published), test and GPA ranges and bands, factor ratings (CDS C7, IPEDS), class rank tiers, required and recommended courses, how the major is reviewed, automatic-admission programs | how-colleges-read.md, base-rates.md, major-and-grades.md |
| Outcomes | Admission results students choose to share, counselor uploads | calibration.md |
What the site has today (measured 2026-10-10)
| Evidence | Colleges or schools | Where it comes from |
|---|---|---|
| Admit rate | 1,547 of 1,893 colleges | IPEDS, newer CDS where published |
| SAT or ACT middle 50% (math sections: 931 SAT, 846 ACT) | 931 | IPEDS, CDS C9 |
| Admission factors | 1,586 (IPEDS scale); 147 with the CDS C7 four-level grid | IPEDS ADM; CDS C7 |
| GPA bands | 116 with any column, 71 with all first-years | CDS C11 |
| Class rank tiers | 117 | CDS C10 |
| Course units required and recommended | 156 | CDS C8 |
| Early decision counts | 42 | CDS C21 |
| Admit rates by residency | 71 | CDS C1 grid |
| AP courses offered | 13,191 of 35,390 high schools (median 9) | CRDC 2023–24 |
| School profile (scale, rank policy, AP list) | the 100 pilot high schools | School profile PDFs |
The CDS-based rows grow with every college-reported run without code changes.
The parts
| Part | Spec | Adds | Profile | Planner | Complexity |
|---|---|---|---|---|---|
| 1 | how-colleges-read.md | "How this college reads a record": its factor ratings in words, how crowded its first-years' GPAs are, tests, class rank, courses expected | Top of "What they look at"; a Compare row | Linked from each row's ⓘ | Small |
| 2 | rigor-in-context.md | A course list (each AP, IB, dual-enrollment, and honors course with year, status, grades, optional exam score), the AP catalog, the school's offering, a plain reading of the schedule against it | A courses line in the standing card; a You column in "What you'll need in high school" | Optional "Add your courses"; a Courses section in Scores | Medium |
| 3 | base-rates.md | The admit rate for the student's own pool (automatic admission, major, residency), shown with its source | The rate that applies to you; the guarantee line | "Guaranteed for you" where a published program applies | Large |
| 4 | estimate.md | Quad's estimate: the interface, server-side computation, the open baseline, the "what went into it" panel; the move to a private package at launch | "Where you stand" chip and card | Automatic groups, retake and course suggestions, balance line | Medium |
| 5 | calibration.md | Outcomes shared with consent, the published accuracy summary, "students like you" counts | Counts under the estimate when a cell has enough outcomes | Outcome snapshots on "applied" | Large |
| 6 | course-plan.md | Next year's advanced courses: up to two suggestions from the school's list, with guardrails for grades and load; the course-registration and AP-exam dates | One "you're on track for 3 of 4 years" line | A Next year card in Courses; calendar tasks; "send AP scores for credit" | Medium |
| 7 | major-and-grades.md | How each college says the intended major is reviewed (the pool, subject emphasis, required courses, score requirements), quoted per unit; math and science grades for STEM majors | A Major line; the unit's required courses | Requirements in the course plan; score requirements in Scores | Medium |
Parts 8–12, planned 2026-10-11 from the owner's review of the build (the next wave; roadmap group "Admission chances, next"):
| Part | Spec | Adds | Profile | Planner | Complexity |
|---|---|---|---|---|---|
| 8 | levels.md | Five levels (High reach, Reach, Target, Likely, Safe) from the method's third version (method/standing.md): a measured academic distance, multipliers on the pool's rate, the selectivity tier as the cap | The chip and card on five levels | Five sections ordered closest first, the ED guard, the balance line | Large |
| 9 | pool-explainer.md | "Applicants like you — from Indiana — were admitted at 71%" and a "Who you're compared with" panel; the word "pool" leaves the site | The standing card's lines | The drawer's lines | Small |
| 10 | depth.md | Course tracks by subject against the major; an activities list read by years, role, and recognition in the top two | Activities and Tracks on /me; a Depth line in the card |
"Add your activities"; Tracks in Scores → Courses | Medium |
| 11 | inputs-beyond-academics.md | Class standing as a pick, an awards catalog, work hours, a legacy toggle and a recruited-athlete pick per college, logged interest; the rewritten commitment | The fields on /me |
The drawer's "About you and {College}" line | Medium |
| 12 | selectivity-index.md | A hidden selectivity index per college and a public tier, Elite to Open, with its inputs listed | A "Quad's selectivity" chip with its panel; Explore facet; Compare row | Caps the level; "Elite: a reach for everyone" | Medium |
Build order
how-colleges-read ─────────┐
rigor-in-context ──────────┼──► estimate ──► calibration (first season after launch)
base-rates ────────────────┤ │
major-and-grades ──────────┘ │
rigor-in-context ──► course-plan ◄──┘ (its "what it changes" line)
how-colleges-readfirst: profile-only, small, no account.rigor-in-context,base-rates, andmajor-and-gradescollect and show inputs; they are useful on the profile before the estimate reads them.estimateafter them, and after PR #105 merges: it moves the planner's standing behind the server-side interface and implements method/standing.md.course-planafterrigor-in-context; ship it before February so the first course-registration season uses it.calibration's outcome snapshots ship withestimate; the first summary runs after that season's decisions.- Build with the
build-roadmap-sectionskill, one unit per part; the units that touch the estimate read the method files.
What happens to existing specs
| Spec | Change |
|---|---|
| Chances and Fit: Where You Stand | Its Display, Fit, and Caveats stay its own work item. Its published rules become the open baseline (development and, after launch, anywhere the private package isn't installed) (estimate.md); "published on the glossary" no longer applies to the production method |
| ../planner/redesign/standing.md | Unchanged in PR #105; estimate.md later moves standingFor behind the server-side interface |
| Getting Into the Major: Direct Admission, Secondary Admission, and Changing In | Stays an idea; base-rates and major-and-grades promote a first wave of its data |
| Scattergrams: Outcomes From Your High School | Unchanged; its uploads become an outcome source |
Shared rules
- Never a percentage. A level (Reach, Target, or Likely; five from levels.md), with labels; facts shown beside it are what they are (the share of applicants like the student admitted, with its year).
- Say who the student is compared with, never "pool" (pool-explainer.md).
- Encourage, don't grade. No new input lowers a level except where the college itself states a requirement; every reading comes with a suggestion first. Missing inputs are the exception (owner, 2026-10-12): an estimate built on part of a record reads lower than the same record complete, and says what to add (method/standing.md "Revision 1") (depth.md, inputs-beyond-academics.md).
- Inputs, not method, on the site. Pages say what kinds of information the estimate uses and which of the student's inputs it used; never how they are combined. The method's files are for building it, not for display.
- Every fact cited. College and high school facts keep their lineage; the estimate is labeled Quad's estimate.
- No protected traits, no high school identity as inputs, ever (a public commitment).
- Encourage, don't grade: the planner redesign's principles apply to every planner surface.