USCIS made the H-1B lottery count people, not entries, and the duplicate bubble popped

Every March, USCIS holds a lottery for the 85,000 H-1B cap slots. Since electronic registration began, entering costs only a $10 fee per registration, and each registration was a separate ticket. As odds worsened, intermediaries and employers stacked registrations for the same desirable workers — through related entities or duplicate identity documents. In fiscal year 2024 the system received 780,884 registrations, 758,994 of them eligible; 408,891 — about 54% — were for beneficiaries who had other eligible registrations filed for them too. Only 188,400 registrations were selected.

DHS answered with the final rule 'Improving the H-1B Registration Selection Process and Program Integrity' (89 FR 7456, in force for the FY2025 season). The core move is disarmingly simple: when random selection is needed, USCIS draws among unique beneficiaries, and each beneficiary counts once toward cap projections no matter how many registrations, or how many petitioning entities, stand behind him. Filing five registrations for one star candidate now buys exactly the same chance as filing one.

The rule also attacked the supply of paper tickets at the source. Registrations must be bona fide — backed by a real position consistent with the proffered wage level, with the registrar attesting that no agreement exists to file duplicates to inflate odds — and USCIS warned it would deny or revoke registrations that were not. USCIS noted that attorneys or accredited representatives alone submitted 48% of registrations, evidence of how industrialized ticket-stacking had become.

The effect was visible immediately. By the FY2026 registration period USCIS reported a significant decrease in total registrations and eligible beneficiaries versus FY2025, including registrations for beneficiaries with multiple filings, and the average number of registrations per beneficiary fell to roughly 1.01 (from 1.06 in FY2025) — a system where, on average, each worker holds about one ticket. USCIS attributes the collapse of unfair-advantage attempts to the beneficiary-centric selection process itself.

Select people once, however many tickets they hold

Changing what the draw counts — people instead of paper — removes the payoff from duplicates outright, so gaming dies without needing to catch and punish anyone.

Under per-entry selection the bubble was self-reinforcing: as honest odds fell, more employers stacked tickets, driving odds lower still; one-count-per-person breaks the loop.

Requiring each registration to be bona fide (real job, matching wage level, no collusion) raises the cost of the residual strategies one-counting does not cover.

The $10 fee made marginal tickets nearly free, so only changing the counting unit — not more penalties — could beat zero-cost entry.

Counting a beneficiary once across all registrants closes the multi-shell route of related entities, not just the single-filer route.

A lottery is only as clean as its unit of counting. If extra tickets are cheap, the prize goes to whoever files the most paperwork. Select the person, not the paper.

The beneficiary-centric draw has run every season since FY2025. USCIS reports far fewer attempts to gain an unfair advantage and an average of 1.01 registrations per beneficiary in the FY2026 season, with no second lottery needed to fill the FY2027 cap. The mechanism is now evolving further: from FY2027, USCIS weights the draw by wage level — unique beneficiaries enter the selection pool one to four times depending on the OEWS wage level of the proffered wage — while keeping one-beneficiary-one-count at its core.