The US residency match let 20,000 doctors a year rank first, and be ranked never

The NRMP matches about 20,000 new physicians to US residencies yearly. By the 1990s applicants believed the hospital-proposing algorithm worked against them, and couples needing two nearby jobs broke the simple matching model the original 1952 design assumed.

In 1995 the NRMP board commissioned Alvin Roth, who with Elliott Peranson designed a replacement in 1997, adopted for the 1998 match. Because complementarities (couples, linked positions) make standard theory inapplicable, they tested variants computationally — and found the set of stable matchings, and the room for strategic manipulation, surprisingly small.

They then made the safe directional choice: propose from the applicant side. Applicants get their best achievable stable outcome; the algorithm is strategy-proof for them, ending the incentive to game rank lists. The work later anchored Roth's 2012 Nobel in economic sciences.

Flip who proposes: run deferred acceptance from the applicants' side

Theory said the market's complications could break stability; computational experiments showed the damage was tiny.

Applicant-proposing deferred acceptance is provably optimal and strategy-proof for the distrusting side.

Couple matching was engineered into the algorithm instead of being handled by side deals.

The redesign was adopted wholesale by a live market in 1998 — no pilot, and it worked.

When a market's rulebook is distrusted, the fix can be directional: run the same stable-matching machinery from the side you want to protect, and measure before believing theory's warnings.

The applicant-proposing algorithm has run every Main Residency Match since 1998, now placing over 40,000 applicants annually. Roth and Peranson's 1999 American Economic Review paper became a founding text of market design, cited in Roth's 2012 Nobel Prize.