Scale规模

Outside the United States, who sends the most美国之外，谁送得最多

Bar length is five-year finalist projects; the darker segment is projects that won under the current lens. The United States is 25× the runner-up, so it sits on its own row above with its own scale.

横条为五年入围项目数，深色段是当前口径下获奖的项目。美国体量是第二名的 25 倍，单列在上方，用自己的标尺。

The United States alone is 67.5% of all finalists. That is not a distribution of talent — it is a distribution of quota: 278 ISEF-affiliated fairs, each holding its own allocation.

美国一国占全球 67.5%。这不是实力分布，是赛区配额分布：278 个 ISEF 附属赛区，每个持有独立名额。

Rates & Uncertainty比率与不确定度

Grand-award rate, with confidence intervals大奖率排行与置信区间

The dot is the rate; the line is the Wilson 95% interval. Only countries with 40 or more projects in the award-eligible window are shown — below that the interval is too wide to rank.

点是比率，横线是 Wilson 95% 置信区间。只列评奖口径项目数 ≥40 的国家与地区——低于这个数，区间宽到无法排名。

Countries that send few finalists post high rates because of the funnel: Singapore sent 30 projects in five years — its strongest 30. The United States sent 4,558, covering everything from magnet schools to rural county fairs. Puerto Rico and Qatar are the exceptions that the funnel does not explain: both hold quota and convert very little of it.

送得少的国家比率高，是选拔漏斗造成的：新加坡五年只送 30 个项目，是全国最强的 30 个；美国送 4,558 个，覆盖从顶尖磁校到偏远县赛区的全谱系。波多黎各与卡塔尔是漏斗解释不了的例外——有配额，转化极低。

Eligibility资格

How much of a special award is US-bound专项奖里有多少是美国身份限定的

Bar length is total special-award entries; the darker segment is entries whose name contains University, Scholarship, Air Force and similar markers of US-only eligibility.

横条为专项奖条目总数，深色段是名称含 University／Scholarship／Air Force 等美国限定标记的条目。

This keyword test is an approximation, not an official classification — but it is enough to show why a combined rate flatters American finalists.

这个关键词判定是近似口径，不是官方分类——但足以说明合并口径为什么会偏向美国选手。

Stability稳定性

Headcounts barely move year to year各国人数逐年几乎不变

What changes is quota, not form. Vertical axis is zoomed to the range of variation.

变的是配额，不是状态。纵轴已放大到波动区间。

Mainland China is the most volatile of the major entrants: 35 projects in 2024, 5 in 2025, 27 in 2026. That year checks out as real — the 2025 corpus is complete and the country field has no gaps. The record cannot tell us why.

中国大陆是主要参赛方中波动最剧烈的：2024 年 35 项，2025 年 5 项，2026 年 27 项。该年数据经核为真实——当年语料完整，国家字段无缺失。原因语料无法回答。

Four Regions四个地区

Mainland, Taiwan, Macau, Hong Kong大陆、台湾、澳门、香港

The dot is the rate; the line is the Wilson 95% interval. Hong Kong and Macau each have only around twenty projects, so their intervals run from single digits to the forties — wide enough that ranking them would be meaningless.

点是比率，横线是 Wilson 95% 置信区间。香港与澳门各只有二十余个项目，区间从个位数横跨到四十几——宽到排名毫无意义。

Quota配额

Only the mainland moves只有大陆在波动

Taiwan has sent exactly twelve projects every year for five years. Macau and Hong Kong are equally flat. These are stable affiliated-fair allocations.

台湾五年每年整整 12 项。澳门与香港同样平直。这些是稳定的附属赛区配额。

Inside the Mainland大陆内部

Shanghai and Beijing differ threefold上海与北京差三倍

City assigned from the school name. Bar length is projects; the darker segment is grand awards.

按学校名归属城市。横条为项目数，深色段为大奖项目。

Shanghai versus Beijing: Fisher exact test p = 0.0012. Of the mainland's eight first-place-or-above awards in five years, seven come from Shanghai schools; Beijing's 38 projects produced none. The record cannot explain why, and we will not guess — selection mechanism, mentoring resources and project type are all plausible and none is measurable here.

上海对北京，Fisher 精确检验 p = 0.0012。大陆五年 8 个一等及以上奖项，7 个来自上海的学校；北京 38 个项目零一等奖。原因语料无法解释，我们也不猜——赛区选拔机制、指导资源、项目类型都说得通，但在这份记录里都不可测。

Tracks赛道

Where China crowds in, and where it is absent中国扎堆在哪，缺席在哪

China's share of each track against the American share, sorted by ratio. Technology Enhances the Arts was only created in 2024, so its share uses a 2024–2026 denominator; Systems Software was renamed Software Design in 2026 and the two are merged.

中国合计与美国的赛道占比对照，按倍数排序。技术增强艺术 2024 年才设立，占比用 2024–2026 做分母；系统软件 2026 年改名软件设计，两者已合并。

Schools & Projects学校与项目

111 schools, 254 projects111 所学校，254 个项目

The Structural Fact结构性事实

Quota goes to fairs, not to states名额发给赛区，不发给州

The rulebook is explicit: each affiliated fair may send the number of projects written into its own affiliation agreement. Society for Science sets that number from four factors — students at the fair, schools represented, high-school population of the territory served, and estimated participation in feeder fairs. The agreements are not public.

规则手册写得很明确：每个附属赛区可送出的项目数，写在它与 Society for Science 单独签的协议里。该数字由四项因子决定——赛区参赛学生数、参赛学校数、所辖区域高中生人口、下游 feeder 赛区参与人数估算。协议不公开。

Median allocation is two projects. Sixty fairs hold exactly one. The two largest — Long Island and Westchester, both in New York — hold twenty each, more than most states hold in total.

名额中位数是 2 个。60 个赛区只有 1 个。最大的两个是纽约长岛与威彻斯特，各 20 个，比大多数州的全州名额还多。

By State按州

Finalists and awards by state各州入围与获奖

Bar length is five-year finalist projects; the darker segment is projects that won under the current lens.

横条为五年入围项目数，深色段是当前口径下获奖的项目。

Switch the lens and Arizona moves: first in the country on the combined rate at 66.3%, but 27.6% on grand awards alone — below Florida. Seventy-two of its eighty-three special-award entries are scholarships from the two big Arizona universities. On grand awards the real leader is California.

切换口径，亚利桑那会动：合并口径 66.3% 全美第一，只算大奖 27.6%，低于佛罗里达。它 83 条专项奖里 72 条是本州两所大学的奖学金。大奖口径下真正的第一是加利福尼亚。

Fair Directory赛区名录

283 affiliated fairs, 2023 season283 个附属赛区，2023 赛季

Allocation per fair, cross-checked against the official finalist record: 38 of 49 states match exactly, 45 within one, and the totals differ by 0.56%.

各赛区名额，与官方入围记录交叉验证：49 个州中 38 个完全一致，45 个差值不超过 1，合计相差 0.56%。