How to Rank Residency Programs: NRMP Algorithm Guide

Abstract

This article gives a scientific framework for ranking residency programs in the NRMP Match. The central conclusion is simple but often emotionally difficult: once you have interviewed, the correct ordering strategy is to rank acceptable programs by true preference, not by predicted likelihood of matching.

That conclusion follows from the structure of the NRMP algorithm. The algorithm is applicant-proposing. It first attempts to place an applicant at the applicant's most preferred ranked program. If that cannot become a tentative match, the algorithm tries the next program, and continues until the applicant is tentatively matched or the applicant's rank order list is exhausted. Therefore, a lower-ranked program is not harmed by being lower on the list unless a higher-ranked program successfully matches with the applicant, which is exactly the outcome the applicant said they preferred.

The part applicants can improve is not gaming rank order. It is the quality and completeness of the list: ranking every acceptable interviewed program, excluding programs where training would be impossible or unacceptable, verifying program codes and tracks, using supplemental rank order lists correctly for advanced programs, and applying a structured program-evaluation model before certifying.

Main conclusion

The rank list has two jobs. First, it tells the algorithm your preferences. Second, it creates the set of acceptable outcomes where you are willing and able to train. Those jobs should not be confused.

Ordering affects where you match among programs that can match with you. Length affects how many acceptable opportunities the algorithm can consider. The best strategy is therefore: remove unacceptable programs, rank all acceptable interviewed programs, and order them by true preference.

A rank list cannot make a program rank you. It cannot convert a non-interview into a match. It cannot overcome visa, licensing, contract, or eligibility barriers after Match Day. It can, however, prevent self-harm. The most common self-harm is moving a preferred program down because of fear, failing to rank an acceptable program, ranking the wrong track code, or ranking a program that would be unsafe or impossible to attend.

How the algorithm works

The NRMP algorithm can be understood as a version of deferred acceptance adapted for residency markets. Programs have quotas, applicants have preferences, programs have preferences, and the algorithm looks for a stable matching between the two sides.

A simplified version works like this. The algorithm takes an applicant and attempts to place that applicant into the first program on the applicant's rank order list. If the program ranked the applicant and has an open position, the applicant is tentatively matched. If the program is full but prefers that applicant to another tentatively matched applicant, the lower-preferred applicant is displaced and the new applicant becomes tentatively matched. The displaced applicant then tries the next program on their own list. The process continues until no further tentative match can be improved under the submitted lists.

A tentative match is not final until the algorithm is complete. At the end, tentative matches become final and binding for training. The binding nature of the result is why the acceptability filter matters as much as the preference order.

The math

Let A be the set of applicants and P be the set of programs. Each applicant a has a preference list over acceptable programs, and each program p has a preference list over interviewed applicants plus a quota q(p). A match is stable if there is no applicant-program pair where both sides would prefer each other over their current outcome and the program has capacity or prefers that applicant to one of its current matched applicants.

For a single applicant in an applicant-proposing deferred-acceptance system, truthful ranking is the dominant strategy under the standard model. In practical language: if you prefer Program X over Program Y, ranking Y above X cannot improve your outcome. It can only make you match at Y in a scenario where X might have been possible.

The probability of matching somewhere is not improved by moving a lower-preference program higher. If you do not match at the higher-preference programs, the algorithm will still consider the lower-preference program when it reaches that rank. Therefore, 'I should rank the safe program first' is usually a misunderstanding of the mechanism.

The applicant's optimization problem is better written as: maximize expected utility among acceptable outcomes, subject to eligibility and feasibility constraints. A useful pre-ranking formula is U(program) = training quality + supervision + resident culture + career fit + logistics + personal constraints + IMG feasibility - red flags. The formula helps estimate preference; it should not replace true preference once the applicant has made a careful judgment.

What increases match probability

Rank-list ordering and match probability are related but not the same thing. Ordering determines which program you get if multiple outcomes are possible. The number of acceptable ranks determines how many chances the algorithm can evaluate after interviews.

NRMP Charting Outcomes reports repeatedly show that applicants who match tend to have more contiguous ranks in their preferred specialty than applicants who do not match. For IMGs in the 2024 report, matched U.S. IMGs averaged 8.8 contiguous ranks compared with 2.4 among unmatched U.S. IMGs. Matched non-U.S. IMGs averaged 6.2 contiguous ranks compared with 2.5 among unmatched non-U.S. IMGs.

That evidence should be interpreted correctly. Contiguous ranks are not magic. They are a final proxy for interview yield, program fit, competitiveness, documentation, communication, and programs being willing to rank the applicant. You cannot create the benefit of twelve interviews by typing twelve random programs into the list. But after interviews are complete, you can avoid losing opportunities by ranking every acceptable program.

Build the list in two passes

A scientific rank list starts with an exclusion pass, not a prestige pass. Before asking which program is best, ask which programs are actually acceptable outcomes.

The first pass is binary: eligible or not, acceptable or not. The second pass is comparative: among acceptable programs, which would you prefer if the Match could place you there? Keeping those decisions separate prevents anxiety from contaminating the list.

Do not create the final order during an emotional post-interview high or after a frightening conversation with another applicant. Write structured notes within 24 hours of each interview, then return to those notes when building the list.

Program evaluation scorecard

A scorecard is useful because it makes vague feelings comparable. It is not meant to turn a human decision into a fake exact science. It is meant to prevent the loudest factor, usually prestige, geography, or fear, from overpowering the whole decision.

The weights below are a starting model. Applicants should modify them before scoring. A fellowship-focused applicant may increase career fit. A parent may increase geography and schedule predictability. A non-U.S. IMG may make visa feasibility a non-negotiable filter rather than a weighted factor.

How applicants actually choose

The NRMP 2025 Applicant Survey helps explain what applicants value when ranking programs. The most frequently endorsed ranking considerations included interview day experience, overall goodness of fit, desired geography, quality of residents, program reputation, and work-life balance.

That list is useful because it matches how applicants actually experience programs. It also needs interpretation. Interview day experience can be informative, but it is also curated. Reputation can matter for career mobility, but it can hide poor supervision. Geography can protect wellbeing, but it can narrow opportunity. The goal is not to reject these factors. The goal is to evaluate them with evidence.

A good ranking process turns subjective impressions into testable notes. Instead of 'the residents seemed happy,' write what you observed: residents answered questions without faculty present, described realistic challenges, gave specific examples of teaching, and could name changes made after feedback.

Program quality evidence

Program evaluation should use multiple sources because each source has a different bias. Official databases may be accurate but incomplete. Program websites may be current but promotional. Resident conversations are rich but anecdotal. Online forums can reveal concerns but are often unverified and skewed toward extreme experiences.

The strongest decisions triangulate: NRMP data for specialty competitiveness, ACGME for accreditation status, Residency Explorer or FREIDA for structured program information, program websites for curriculum and policies, and direct resident conversations for day-to-day reality.

IMG-specific ranking

IMG applicants should use the same core ranking rule: true preference among acceptable programs. The difference is that acceptability has more hidden constraints. A program is not acceptable if the applicant cannot meet visa, ECFMG, state training license, onboarding, or start-date requirements.

For non-U.S. IMGs, visa sponsorship should usually be a pre-ranking filter, not a low-weight scorecard item. For older graduates, clinical recency and support for nontraditional trainees matter. For applicants with attempts or other red flags, resident support and remediation culture are not cosmetic; they are training-safety variables.

IMGs should also distinguish program friendliness from program fit. A program may have many IMGs but still be a poor fit because of geography, schedule, supervision, visa type, fellowship goals, or resident wellbeing. Another program may have fewer IMGs but a clear, supportive, legally feasible pathway for your profile.

Advanced and preliminary programs

Applicants applying to advanced PGY-2 programs must think in pairs: the advanced position and the required PGY-1 year. NRMP allows applicants who rank advanced positions on the primary rank order list to create supplemental rank order lists for preliminary PGY-1 positions.

The scientific principle is the same, but the object being ranked changes. You are not only asking, 'Which advanced program do I prefer?' You are also asking, 'Which PGY-1 combinations make this path acceptable?' A dream PGY-2 program with no feasible preliminary year may not create a complete training pathway.

Applicants should verify track type, program code, and linkage carefully. Some programs have categorical, preliminary, primary care, research, rural, physician-scientist, or other tracks with similar names but different NRMP codes. Ranking the wrong code is a preventable error.

• Categorical program: usually includes the full required training pathway beginning at PGY-1.
• Advanced program: typically begins at PGY-2 and requires a separate compatible PGY-1 year.
• Preliminary program: a one-year PGY-1 position often used before advanced specialty training.
• Supplemental rank order list: a list of PGY-1 options attached to an advanced program rank.

Couples ranking

Couples ranking is mathematically different from individual ranking because the unit of preference is a pair of outcomes. The couple ranks combinations, such as Applicant 1 at Program A and Applicant 2 at Program B. This can create many more possible rank lines and more complex constraints.

The guiding principle remains preference ordering among acceptable pairs, but couples should be especially careful about feasibility. Geography, commuting, visa needs, preliminary years, family constraints, and backup combinations need to be explicit. A pair that looks acceptable on paper may be impossible in daily life.

Couples should build their list from joint utility, not from one applicant's prestige preference alone. A stable, livable pair of programs may be better than one applicant's ideal program combined with the other's unworkable outcome.

Ranking mistakes

Most ranking mistakes come from treating the Match like a negotiation after interviews are over. It is not. Once interviews are complete, the rank order list is a preference document plus a binding-commitment document.

The algorithm does not reward flattery, fear, or attempts to read minds. It rewards accurate preferences and complete acceptable lists.

A practical ranking protocol

A strong process reduces cognitive bias. Use the same workflow for every program, especially if interviews were spread across weeks or months.

The protocol below is designed for applicants who want a defensible, evidence-based list rather than a list built from panic during the final week.

Example

Suppose an applicant interviews at six internal medicine programs. Program A is prestigious but residents describe weak feedback and poor schedule transparency. Program B is less famous but has strong clinical training, supportive residents, a visa pathway, and graduates entering the applicant's desired fellowship. Program C is in the applicant's favorite city but has unclear visa answers. Program D is community-based, supportive, and acceptable. Program E is acceptable but geographically hard. Program F has a contract or visa barrier that the applicant cannot meet.

A scientific list might remove Program F entirely. Program C stays only if the visa issue is resolved. The applicant then ranks A, B, D, and E by actual preference, not by perceived odds. If B is truly preferred over A after weighing supervision, fit, and career outcomes, B should be ranked above A even if A has the bigger name. If A is still the true preference despite its weaknesses, A can be ranked first without harming the applicant's chance at B if A cannot match with them.

This is the important psychological shift: the rank list is not a place to prove you are realistic. It is a place to state your preferences among realistic, acceptable outcomes.

Final audit

Before certifying, do one final audit with a rested mind. The audit should be boring. That is the point. You are trying to catch preventable errors, not reinvent the list under stress.

Ask: Is every ranked program one where I can train? Is every acceptable interviewed program included? Are the programs ordered by true preference? Are track names and NRMP codes correct? Are supplemental lists attached where needed? Did I save and certify? If I changed anything after certification, did I re-certify?

• Confirm the list is certified in R3.
• Check for confirmation email from NRMP.
• Verify every program code and track name.
• Verify categorical, preliminary, advanced, and supplemental list logic.
• Remove programs you would not attend.
• Keep acceptable lower-preference programs on the list.
• Do not rely on memory; compare against your interview notes and program documents.

Limitations

This article explains the general Main Residency Match strategy. It is not individualized legal, immigration, financial, or medical career advice. Applicants with visa constraints, couples lists, advanced-specialty pathways, military match obligations, ophthalmology or urology early matches, prior residency training, NRMP waiver issues, or unusual state licensing concerns should consult official sources and qualified advisors.

The matching-theory conclusion about true-preference ranking is strongest for individual applicants under the standard applicant-proposing model. Couples matching is more complex because preferences are joint outcomes. However, couples should still avoid manipulative guesswork and should rank acceptable pairs in genuine joint preference order.

Rank-list length is associated with matching, but it is not purely causal. More ranks usually mean more interviews and stronger program fit. The actionable lesson is to rank every acceptable interviewed program, not to add unacceptable programs or programs where no interview occurred.

Conclusion

The most accurate scientific advice is also the cleanest: rank acceptable programs in true preference order. Do not rank based on where you think you will match. Do not punish yourself for liking a reach program. Do not create false safety by moving a lower-preference program upward.

To increase the chance of a successful Match, focus on what the rank list can actually control. Include every acceptable interviewed program. Remove programs where you cannot or would not train. Verify codes, tracks, visas, licensing, contracts, and supplemental lists. Use a structured scorecard to clarify preference, then let the applicant-proposing algorithm do what it was designed to do.

A successful rank list is not the list that looks most strategic to strangers. It is the list that is complete, technically correct, ethically honest, and aligned with where you can become the physician you are trying to become.