A Field Guide in Seven Chapters

How Elite Colleges Actually Decide

The SFFA trial pried open Harvard's reading-room machinery and entered every formula, rating, and committee vote into the public record. Pair that with NACAC, NBER, and CDS yield numbers and a system that looks mysterious from the outside resolves into something closer to a slow, deliberate, very human pipeline.

1–6 Rating scale across six components
4 Reads before a typical decision
~86% Admit rate for recruited athletes at Harvard
9:1 2017 admit-rate ratio, ALDC vs. non-hooked
Scroll to begin ↓
Chapter I

The 1–6 Scale Nobody Sees

Strip away the mystique and an Ivy application is graded on a sheet that looks like a high-school report card. The Harvard trial revealed a 1–6 rubric — one is best, six is worst, plus or minus for finer cuts — applied across six components: academic, extracurricular, personal, athletic, recommendations, and the alumni interview.

A 1 in academic is reserved for “potential summa” cases — near-perfect scores, intellectual genius, fewer than one percent of admits. A 2, the magna track, is what most admitted students actually score: a 750-plus on each SAT section, a 33-plus ACT, superb grades. Below that, academic ratings drop fast.

Personal and extracurricular ratings are softer but matter more at the margin. The trial transcript shows readers writing “humor, grit, leadership, integrity, courage, kindness” against a 1, and “bland,” “immature,” or “questionable” against a 4 or worse.

~90% of Harvard admits land at academic rating 2 — not 1. The system rewards excellence with separation, but almost nobody clears the genius bar.

A grid of ratings is just bookkeeping. The verdict comes from the overall score.

Chapter II

From a Number to a Verdict

The six component scores collapse into one overall rating. That single digit is what actually predicts admission, and the function is brutally non-linear.

An overall 1 is admitted essentially every time — a 100% rate. An overall 2, even a 2-, gets in roughly 70% of the time. Drop to a 3+ and the rate plummets to about 20%. A 3 is around 3%. A 4+ is below one percent. The cliff between 2- and 3+ is where most of the Ivy applicant pool lives, and where the actual decisions are made.

This is what “holistic review” means in practice. Hard factors (GPA, scores, course rigor) put you in contention; soft factors (essays, recommendations, personal rating) decide it among the academically qualified. The cliff is where personal narrative and institutional fit get their leverage.

3% → 20% → 70% Admission rates for overall ratings of 3, 3+, and 2. A single bracket on one composite score moves your chances by an order of magnitude.

One number, but four humans usually have to agree on it.

Chapter III

Four Reads, One Decision

Most selective colleges run applications through a four-step pipeline. The territory manager — the regional officer who knows your high school's grading curve and your counselor's reputation — reads first, in eight to fifteen minutes, and assigns preliminary ratings.

A second reader independently confirms or challenges. Where they agree, the decision often stands. Where they disagree by two points or more, the file gets flagged for a subcommittee of five to eight officers organized by region. Your territory manager presents and advocates.

The full committee — up to forty people at Harvard — reviews the subcommittee's recommendations under the Dean of Admissions. This is where class shaping happens: the geographic balance, the unfilled engineering slots, the waitlist triage. Stanford skips the committee and lets single readers decide more often, which removes a layer of advocacy. Big publics flip the order entirely — an algorithm sorts first, then humans read borderline cases.

20–40 Applications a single reader reviews per day at peak. Reader fatigue, halo effects, and territorial advocacy aren't bugs in the system — they're documented, durable features of it.

Once the file is decided, the math problem is no longer about who.
It's about how many.

Chapter IV

The Yield Math

No college admits exactly the size of its class. The ratio of admits to enrollees is the yield rate, and the entire over-admission strategy — how many extra students to let in — is calibrated against it.

Harvard yields ~84% of admits. Stanford, ~82%. Princeton drops to 72%. By the time you reach Yale, Penn, and Brown, yield sits in the high sixties — meaning these schools admit roughly 1.5 students for every seat. At schools ranked 50 and below, the multiplier swells to three-to-five-times target.

The single most powerful yield tool is Early Decision. Because ED is binding, its yield is 100% by construction. Schools fill 30–50% of their class in the binding round, then play the rest of the math against the gap. ED acceptance rates run two to three times higher than RD — Dartmouth's 19.1% ED versus 5.4% RD is a representative split.

2,576 Admits for a 1,700-seat target at 66% expected yield, after a 2% safety margin. The over-admit number is roughly 50% larger than the class itself.

Yield is a forecast. The waitlist is what runs when the forecast misses.

Chapter V

The Buffer

On May 1 the deposits land and a college finds out whether it guessed right. The waitlist is what activates when it didn't — a buffer that lets the dean fill specific gaps in the class without admitting from rejection.

Most colleges admit roughly 20% of waitlisted students — the NACAC average. At the most selective end, that drops to ~7%. The volume can be staggering: Michigan offered 26,898 waitlist spots for the Class of 2029 and admitted 973 of them, a 5.2% pull rate.

Volatility year-over-year is enormous. UC Berkeley pulled 1,191 students from its list for the Class of 2028 and just 26 the following year. Wesleyan went from 201 to five. The waitlist is not a queue; it is a last-mile sorting tool, used to fix specific demographic, geographic, or revenue holes.

0% / 35.7% Two extremes. Swarthmore and USC don't pull from a waitlist at all in a typical year. Tufts pulled 35.7% of its accepting waitlist in 2025 — a deliberate yield-protection strategy.

Inside the read, before it ever reaches a waitlist, hooks are already at work.

Chapter VI

The Hook Hierarchy

Of all the trial revelations, ALDC — Athletes, Legacies, Dean's-list applicants, Children of faculty — is the one most students never see coming. These four categories don't get a thumb on the scale. They get a forklift.

Recruited athletes at Harvard are admitted at roughly 86%. The non-hooked baseline is about 5–6%. That's a 14–17x lift — the equivalent of ~200 SAT points worth of advantage. Children of faculty admit at ~47%. The Dean's interest list, which tracks donor relationships, runs ~42%. Legacies sit at ~34% — a 5.7x multiplier on baseline.

First-generation status is the only ALDC-adjacent factor where the lift is small — a modest 1.3–1.5x. The hook hierarchy is steep, durable, and almost completely invisible from the outside of an application reading room.

43% of Harvard's white admits are ALDC. The same categories make up under 16% of Black, Asian, and Hispanic admits. Three-quarters of white ALDC admits, the trial concluded, would have been rejected without their hook.

The most surprising trial finding wasn't the rates. It was the trajectory.

Chapter VII

The Trend Underneath

From 2000 to 2017, Arcidiacono and colleagues found the ratio of ALDC to non-ALDC admit rates at Harvard more than doubled — from roughly 4-to-1 to 9-to-1. Hooks didn't fade as the applicant pool got more competitive. They became more decisive.

That trajectory reframes the rest of the system. As pools grow, holistic review's discretionary middle widens, and discretionary middles are where institutional priorities — athletic recruiting, donor cultivation, alumni continuity — do their work. The post-SFFA decision in 2023, which eliminated explicit racial consideration, sharpened that pressure on every other category.

Look further down the funnel and the same logic recurs. Summer melt — deposit attrition before September — is near-zero at HYPSM (1.5% on average), tolerable at the Ivy+ (3%), and runs into double digits at the selective publics. The system is engineered for stability where the endowment is biggest.

4:1 → 9:1 Harvard's ALDC-to-non-ALDC admit-rate ratio, 2000 vs 2017. The leverage on hooks didn't decay with the applicant boom — it intensified with it.
The Six-Component Reading Sheet
What each rating means at Harvard, drawn from the SFFA trial record
Source: Holistic Review Process, college_decision_model.md (SFFA v. Harvard trial findings, 2019).
Admit Rate by Overall Rating
A single composite score is the strongest predictor of admission
Source: Admission Rates by Overall Score, college_decision_model.md.
The Four-Step Reading Pipeline
From territory manager to dean — how a file moves through committee
Source: Committee vs. Reader Structure, college_decision_model.md.
Yield Rates and the Over-Admit Multiplier
Yield falls down the tier ladder; the over-admit number rises with it
Source: Yield Rates by Tier (Class of 2029 data), college_decision_model.md.
Waitlist Pull Rates by College
A buffer that activates only when yield misses — and varies wildly
Source: research/waitlist_data.json (Class of 2029 figures, IvyWise / individual CDS).
ALDC Admit Rates vs. the Non-Hooked Baseline
Each bar is a category at Harvard; dashed line marks the ~5.5% baseline
Source: Hook Weighting and Balance — ALDC Categories at Harvard, college_decision_model.md.
ALDC vs. Non-ALDC Admit-Rate Ratio, 2000–2017
As applicant pools grew, the hook advantage at Harvard intensified
Source: Trend Over Time, college_decision_model.md (Arcidiacono et al., NBER w26316).
Coda

What the Trial Record Actually Tells You

The most useful framing for an applicant is not holistic versus formulaic. It is two filters in series. The first is academic: hit the magna-track 2 and you are in the room. Miss it, outside of the hook categories, and almost no committee theatre rescues you. The second filter is institutional: under what set of priorities — class shape, donor relations, athletic recruiting, geographic spread — does your file look useful?

The yield math, the waitlist buffer, the ALDC ladder, and the four-read pipeline are not separate phenomena. They are the same machine, viewed from different angles. The committee meeting in March is, in a real sense, just the visible end of decisions about endowment, athletic recruiting, and yield modeling that began the previous summer.

None of it is unknowable. The trial record made the rubric public; the CDS forms publish the yield numbers; NACAC and the news services tally the waitlist. What looks opaque from the outside of an admissions reading room is, on the inside, almost obsessively documented — which is the real argument for studying the process before applying to it.