Top-Seed Title Conversion, Draw Structure and Semifinal Load in Olympic and World Championship Badminton, 1992–2025

World rankings determine seeding at badminton’s major championships, but a seed summarises pre-tournament standing rather than the probability of winning a title. Seed-to-title conversion has been estimated for single Olympic Games in judo and wrestling, but a documented literature search found no comparable analysis of badminton’s major championships. This retrospective census compiled seeds and available knockout results for all 44 Olympic badminton events from 1992 to 2024 and all 65 World Championship events from 2009 to 2025. The primary comparison was top-seed title conversion in the rally-scoring era. The top seed won 10 of 25 Olympic titles from 2008 to 2024 (40%; 95% Wilson CI 23% to 59%) and 22 of 65 World Championship titles (34%; 24% to 46%), a difference of 6.2 percentage points (95% CI ?14.4 to 28.0; Fisher p = 0.63). Olympic champions came from the top two seeds more often (68% vs 43%; adjusted odds ratio 2.93, 95% CI 1.09 to 7.84), although none of the six event-level contrasts was below 0.05 after Holm adjustment. Among knockout matches from 2008 to 2025 between sides in different recoded seed levels, the better-seeded side won 231 of 314 quarterfinals and semifinals (74%) but 47 of 84 finals (56%; 45% to 66%); in a model clustered by event, its odds of winning a final were less than half those in a quarterfinal (odds ratio 0.45, 0.25 to 0.80). Chinese sides won more often against non-Chinese opponents than the broad seed measure accounted for (odds ratio 1.72, 1.29 to 2.28), but nationality could not identify a mechanism. In rally-scoring finals, the finalist with the lighter semifinal won 21 of the 39 in which semifinal game counts differed (54%; 39% to 68%), an imprecise estimate that neither demonstrates nor excludes a fatigue effect. A conditional model of the Rio 2016 men’s singles draw showed that a top-two seed increased the modelled probability of gold relative to a third or fourth seed by 8 to 23 percentage points across the evaluated range and could not reduce it. These results are historical base rates for title conversion, not personalised forecasts or causal estimates of national-system, fatigue or ranking-strategy effects. The dataset, source register, literature-search log and analysis code accompany the manuscript.