Methodology
How we rank colleges, calculate acceptance likelihood, and build the “Straight Talk” hybrid sort.
1. Which Colleges We Include
Before we rank or score anything, we decide which colleges belong in the database. A college qualifies if it appears in any of:
- Top 150 in US News Best National Universities
- Top 50 in US News Best National Liberal Arts Colleges
- Top 100 in Forbes America's Top Colleges
- Top 100 in Niche Best Colleges in America
The bar is intentionally inclusive. We want every college a meaningful number of students would plausibly apply to — not just the elites. Four lists with different methodologies means a school strong in any one dimension (national reputation, liberal arts depth, value, or student experience) makes the cut.
QS World Rankings and ARWU (Shanghai) are not inclusion gates — they're global research-focused rankings that don't reflect where a US high school student would actually apply. They still contribute to the composite ranking score (Section 3) once a school is included.
Military service academies (West Point, Naval, Air Force, Coast Guard) are excluded despite qualifying ranks. Their congressional-nomination admissions process is a fundamentally different experience from traditional college admissions and would distort the acceptance-likelihood model.
2. Ranking Sources
Our college rankings are compiled from multiple authoritative sources, each contributing a different perspective on college quality:
3. Composite Ranking Algorithm
Each source ranks colleges on different scales. We normalize all ranks to a 0–100 scale, apply the weights listed above, and combine them into a single composite score:
Regionally ranked colleges are scored in their own band, and that score is not comparable with a national one. US News ranks regional universities in four separate lists — South, North, Midwest and West — each a different pool from National Universities. Being #22 among regional universities and being #22 in the country are not the same achievement, and mapping one onto the other would invent a precision nobody has. So a college ranked only regionally gets a score between 20 and 50, placed by where it sits in its own list, and we label it as regional. Read it as “where this college stands among its regional peers” — never as a national position.
Colleges close together in the same regional list can end up with the same score. That is deliberate. Four colleges inside ten places of a 140-college list really are near-equivalent, and separating them would be inventing a difference we cannot see.
1. Normalize each source: score = (1 - rank/totalRanked) × 100 2. Apply weights: composite = Σ(weight × normalizedScore) 3. Program-specific boost: +5–15 points for top departmental rankings 4. Final score clamped to 1–100
Colleges not ranked by a particular source receive the median score for that source, preventing unranked schools from being unfairly penalized or boosted.
4. Program Strength Scoring
When you select a specific major, colleges are ranked by how strong their program is in that area. Program strength scores (0–100) are derived from:
- US News departmental rankings (where available)
- Niche program-specific grades
- Research output in the field (publications, grants)
- Faculty-to-student ratio in the department
- Graduate employment outcomes for that major
A score of 90–100 represents a top-10 nationally ranked program, 80–89 is top-25, 70–79 is top-50, and so on. Colleges without a particular program are excluded from results when that major is selected.
5. Acceptance Likelihood Model
The acceptance likelihood calculator estimates your chances of admission in three stages:
Your test score and GPA are compared to the college's admitted student profile using sigmoid functions, weighted equally (50/50). The score is measured against the school's 25th/75th percentile midpoint; GPA is measured against their average admitted GPA.
Either test works. Enter an SAT or an ACT and we compare it to that college's range for the same test. We never convert one into the other: most colleges are test-optional, and the students who submit each test are different groups, so a converted score would be a guess dressed up as your result. If a college publishes no range for the test you sat, your GPA carries the whole estimate there.
Rigor only helps if your GPA backs it up. Taking all APs with a 3.8 GPA shows you challenged yourself and succeeded — a meaningful boost. Taking all APs with a 2.0 GPA means you struggled in hard classes — an actual penalty. This reflects how admissions officers actually read transcripts: rigor without performance is a red flag, not a strength.
Academic fitness is converted to a likelihood percentage through a sigmoid whose threshold and steepness vary continuously with acceptance rate. Ultra-selective schools (~4% acceptance) require very high fitness (~75+) for even a coin-flip chance, while open-admission schools have low thresholds. The parameters are interpolated smoothly between anchor points, avoiding cliff effects at arbitrary tier boundaries.
At highly selective schools, your academic fitness is the dominant signal. But at schools with high acceptance rates, the rate itself is more predictive — a 90% school admits students well below its average profile. For schools above 40% acceptance, the model blends the acceptance rate into the likelihood with increasing weight (up to 60%), so below-average stats at a 90% school still yield a realistic Safety-level result rather than an artificially low score.
Even a perfect applicant can't guarantee admission at a 4% acceptance school — essays, extracurriculars, legacy, and luck dominate at that level. A smooth, monotonically increasing ceiling caps likelihood: ~53% for ultra-selectives, ~77% for mid-range schools, approaching 95% for high-acceptance schools. This ensures a school with a higher acceptance rate always has a higher ceiling than a more selective one. The floor for very poor stats at selective schools is ~1%.
fitness = testFit × 0.5 + gpaFit × 0.5 + rigorAdjustment // testFit = SAT or ACT
rigorAdjust = (rigor × max(0, gpa-2.5) × 4) // benefit
- (rigor × max(0, 3.0-gpa) × 6) // penalty
academic = sigmoid((fitness - threshold) × steepness)
threshold & steepness interpolated between anchor points:
5% acc → threshold 75 | 17% → 60 | 37% → 45 | 75% → 30
accBlend = clamp((accRate - 40) / 80, 0, 0.6) // 0 at ≤40%, 0.6 at ≥88%
likelihood = (1 - accBlend) × academic + accBlend × accRate
ceiling = min(95, 48 + 50 × (1 - e^(-accRate/35)))
4% → ~53% | 20% → ~62% | 50% → ~80% | 90% → ~95%5b. Transfer Student Likelihood
Transfer admissions are fundamentally different from freshman admissions. College GPA is the dominant factor — it's a more reliable signal than high school GPA, and admissions offices weight it heavily. Course rigor (AP/IB) is irrelevant, and SAT scores matter much less (many schools waive testing requirements for transfers entirely).
- GPA is compared against each college's average admitted transfer GPA (not the freshman GPA), with a tighter standard deviation (0.20 vs 0.25) because college grades are more reliable
- A test score — SAT or ACT — is weighted at 15% (vs 50% for freshmen) when submitted, and defaults to not submitted
- Course rigor has no effect — it's not part of the transfer fitness calculation
- Transfer acceptance rates are used instead of freshman rates (sourced from Common Data Set Section D and IPEDS)
- Selectivity curves are slightly steeper — transfer admissions are more GPA-deterministic and less holistic
fitness = gpaFit // GPA-only (no test score)
fitness = gpaFit × 0.85 + testFit × 0.15 // if SAT or ACT submitted
// No rigor adjustment
academic = sigmoid((fitness - threshold) × steepness)
threshold & steepness use transfer-specific curves:
5% acc → threshold 72 | 17% → 57 | 37% → 42 | 75% → 28
// Same ceiling/floor/blending logic, using transfer acceptance rateTransfer acceptance rates can vary significantly by major, timing (fall vs spring), and how many credits you've completed. Some programs have no transfer spots at all in a given year. Our model uses overall transfer acceptance rates as a baseline, but your actual chances for a specific program may differ. Community college students may have additional pathways (articulation agreements, guaranteed transfer programs) that our model cannot capture.
6. Likelihood Labels
At schools like MIT, Stanford, and the Ivies, nearly every applicant already has a near-perfect GPA and top SAT scores. Academics are table stakes, not a differentiator. The actual selection at these schools is driven by factors our model cannot measure: essays, extracurriculars, recommendations, leadership, demonstrated passion, legacy status, recruited athletes, and institutional priorities. A student with a 3.95 GPA and 1550 SAT is competitive on paper but is still more likely to be rejected than accepted at a 4% school.
This means our likelihood labels are least reliable for the most selective schools. A “Possible” at Harvard is very different from a “Possible” at a 40% acceptance school. At ultra-selective schools, treat our labels as measuring academic competitiveness only — the holistic factors that actually decide admissions are beyond what any stats-based model can predict.
7. "Straight Talk" Sorting
Default sorting ranks colleges by program strength for your selected major (or overall ranking when no major is selected). When you activate “Straight Talk,” a hybrid score adjusts rankings based on your acceptance likelihood:
qualityScore = programStrength (with major) or rankingScore (no major) if likelihood ≥ 60%: hybridScore = qualityScore // Pure quality elif likelihood ≥ 20%: hybridScore = qualityScore × (0.5 + 0.5 × (l-20)/40) // Gradual penalty else: hybridScore = qualityScore × (0.1 + 0.4 × l/20) // Heavy penalty
With a major selected, schools are ranked by program quality tempered by your chances. Without a major, the best overall colleges you can realistically get into rise to the top. Reach schools receive a moderate penalty and long shots are pushed down but never hidden.
8. Data Currency
Admissions and cost figures were re-audited in September 2026 against each college's own Common Data Set and published price pages, and reflect the 2026–2027 year where the college has published it. Every figure was checked against a primary source; where a college publishes nothing, we say so rather than guess quietly. Rankings are updated when the major ranking organizations publish new editions.
9. GPA Data Methodology
The acceptance likelihood model compares your GPA to each college's average admitted GPA. All GPA values in our system are unweighted (4.0 scale). Sourcing accurate unweighted GPAs is surprisingly difficult — colleges report GPA data inconsistently, and many don't report it at all.
We categorize colleges into three groups based on what they publicly disclose, and use a different methodology for each:
These colleges publish unweighted GPA data in their Common Data Set (CDS), usually in section C11 or C12. We use these figures directly. Sources include official CDS filings and institutional research pages.
Some colleges (e.g. Harvard, UNC, Georgia Tech, Maryland) report only weighted GPA, which can exceed 4.0 and isn't directly comparable. For these schools, we estimate unweighted GPA using peer matching: we find the 3 most similar colleges that report both weighted and unweighted GPAs (matched by acceptance rate and weighted GPA), compute their weighted-to-unweighted gap, and apply an inverse-distance-weighted average of those gaps.
For each weighted-only school: 1. Find 3 nearest peers from Group 1 (by acceptance rate + weighted GPA) 2. Compute each peer's gap = weightedGPA - unweightedGPA 3. Weighted average of gaps (closer peers count more) 4. estimatedUW = weightedGPA - weightedAvgGap 5. Cap result at 4.0, floor at 3.0
Elite schools like MIT, Yale, Columbia, and many top liberal arts colleges leave the GPA sections of the CDS blank entirely. For these, we cross-reference multiple third-party sources (CollegeSimply, Clastify, CollegeVine, PrepScholar, Admissionado, CampusReel) and use the consensus estimate. Values above 4.0 from any source are flagged as weighted and excluded. Where available, official class-rank data (e.g. “93% in top decile”) is used as a sanity check.
All GPA values were re-checked in September 2026 against the newest Common Data Set each college has filed. Group 2 and Group 3 figures are estimates, so we label them: any GPA we could not confirm against the college's own reporting is marked est. wherever it appears on the site. That is 94 of our 170 colleges — more than half, which is a fact about how little colleges disclose, not a gap in the research. We err conservative, rounding down rather than up, so we do not overstate how hard a school is to get into.
9b. Test Scores and Test-Blind Schools
SAT ranges are the 25th and 75th percentiles of admitted students, taken from each college's Common Data Set (section C9). We note whether a school requires scores, considers them if you send them, or ignores them entirely.
For the 15 colleges that are test-blind — eight University of California campuses, Washington, Reed, Washington State, Cal Poly San Luis Obispo, San Diego State, San Diego, and Cal State Long Beach — we publish no SAT or ACT range at all. These schools do not read scores, so a range would only tell you to work on something that cannot help you get in. For the same reason, neither test raises nor lowers your chances at these schools in our model; your GPA and course rigor carry the whole estimate.
Separately, some colleges do read scores but publish no range for one of the two tests. That is a decision not to report, not a decision to ignore — so we show what they publish and say nothing about what they do not. Where an ACT range rests on only a handful of students, we print how many submitted alongside it, because a range drawn from 45 people and one drawn from 3,800 are not the same kind of claim.
Test policies move around, and a school that piloted a policy is not always still running it. We re-verify each one against the college's own admissions pages rather than trusting a prior year's entry.
9c. What a Year Actually Costs
We show three numbers, because a college bill has three parts and lumping them together hides real money. Tuition is tuition alone. Required fees are what every undergraduate pays on top of it, whatever they study. Room and board is a standard double room plus a standard meal plan.
Fees are not a rounding error. Penn charges $8,308 of them and SMU $8,080, while MIT charges $420 and Yale charges none at all. At the University of Oklahoma, fees are $4,953 against $5,532 of tuition — 47% of the bill. A site that publishes only tuition tells a student Penn is cheaper than it is, and ranks schools in an order nobody actually pays.
Two rules keep the numbers honest. We count a fee only if every undergraduate pays it regardless of major, so major-specific program differentials and waivable health insurance are excluded; where a fee varies by college within a university, we use the floor everyone pays. And for room and board we use a standard double with a standard meal plan, not the Common Data Set figure, which is defined as the maximum meal plan and overstates the bill — at Temple by $790, at Oklahoma by $3,321.
A few colleges never publish tuition apart from fees. Florida State bills a per-credit rate with every fee already inside it; Bentley and Colorado State publish a single combined line. Those pages say “Tuition & Fees” rather than inventing a split the college does not report. One college is a deliberate exception: Cooper Union's $44,550 sticker is a number nobody is billed, because every student in good standing receives a half-tuition scholarship, so we show the $22,275 they actually pay.
All of it is sticker price. It is not what most families pay after aid — use each college's net price calculator for that.
10. Limitations & Disclaimers
This tool is for exploration, not official admissions advice.
- Acceptance likelihood is a statistical estimate, not a prediction. Holistic admissions factors (essays, extracurriculars, recommendations, legacy, athletic recruitment, demographics) are not captured.
- Program strength scores are approximations based on publicly available data and may not reflect recent changes in faculty, funding, or curriculum.
- Tuition and financial aid vary significantly by individual circumstances. Published tuition is the sticker price; actual cost after aid is often much lower.
- Rankings inherently simplify complex institutions into single numbers. A college ranked #50 may be a better fit than one ranked #10 depending on your specific goals.
- Always consult official college admissions offices and financial aid calculators for the most current and personalized information.