Dataset
Every number, with its source
Most writing about the draft recycles four or five numbers whose origins nobody checks. This is the opposite: a browsable table where every row names the document it came from, the page inside that document, and an honest grade for how much weight the number can carry. Where the record is contested, both figures are printed. Where the record is silent, there is a row saying so. The parameters the simulator itself consumes — distributions, grades and sources, including the unsourced ones — are published separately on the ledger page.
How to read a grade. A a primary official document stating or measuring the thing directly. B an official planning assumption, or arithmetic over official documents. C a foreign case, a historical analog, or an analyst estimate — always with a transfer caveat attached. D weak: published, but thin, contested, or resting on a proxy. A high grade means the citation is solid. It does not mean the number is a forecast.
| What it is | Value | Basis | Grade |
|---|
Notes on how this table was built
The five findings a journalist should start with
Each one is in the table above with its full citation. These are the ones that change the shape of an argument rather than decorating it.
1. The 193-day deadline is 13 days plus a six-month guess
The Pentagon's standing requirement — first inductees on day 193 — is not a 193-day engineering estimate. It is the old 13-day delivery requirement with 180 days of assumed strategic warning bolted on the front, set by a memorandum in November 1994 and never revisited. The 13-day leg was never relaxed. And the regulation guarantees a registrant at least 10 days between his induction order and his reporting date, which consumes 10 of those 13 days and leaves roughly three for the lottery, the selection, order generation and the mail. Full derivation →
2. The agency's own answer is 660 days, not 193
In January 2026 Selective Service ran an internal tabletop exercise. Its after-action report's bottom line: approximately 660 days to return to full operations from the current state of readiness, with 44 identified actions that could cut up to a year off if done in advance. The same report states that the agency "currently does not have any planning COAs to guide a return to full operations" and asks, as an open question, "what is the true pacing function for mobilization?" In 2012 the agency had already told the Government Accountability Office it would need 285 days rather than 193. 193 vs 285 vs 660 →
3. On measured capacity, examination is the binding constraint
The plan's own milestones imply about 5,882 inductions per day. The Department of Defense's stated full physical capacity across all 65 military entrance processing stations is 660,000 examinations a year — about 2,542 a day, less than half the requirement. The department also carries an 18,000-a-day figure in mass-mobilization planning, which it labels an assumption, and which contradicts its own stated capacity by roughly seven times. For scale: the requirement is about twice the highest daily induction rate the United States has ever achieved (87,172 inductions in January 1951).
Caveats that travel with this, and are not optional: the 5,882/day figure is arithmetic on two published milestones (first inductee day 193, first 100,000 by day 210), not a published rate, and no ramp shape is published, so a flat rate is an assumption. The 2,542 figure is physicals-only. Surge behaviour above measured volumes is genuinely unknown, and USMEPCOM's per-station maximum daily capacity allocation appears in no public document at all — which makes the binding constraint in a capacity-limited draft unmeasurable from outside.
4. Inferring evasion from unfilled notices overstates it about a hundredfold
South Korea publishes a complete induction funnel. In 2025, 27.1% of induction notices produced no inductee — and measured non-appearance was 0.25%. The gap is almost entirely lawful deferrals, plus the fact that the agency deliberately over-issues notices at 1.27× its induction plan. Sweden's failure-to-appear rate is about 0.4%. The Vietnam-era American record points the same way: roughly 40% initially failed to comply with an induction order, but about 80% of those eventually reported, leaving terminal loss near 8% — non-compliance was follow-up workload, not attrition.
Foreign cases carry a transfer caveat: different legal regime, different population, and in Ukraine's case a wartime criminal mandate. They bound the plausible range; they are not US rates.
5. The registry is not a clean list, and the trend is down
Published registration compliance has fallen from over 97% in 1991, through an 18-year plateau at 91–93%, to 78.7% in 2025 — roughly two points a year, with a documented cause: Congress abolished a registration channel that carried about 20% of annual registrations. Meanwhile six states record more registrants than the Census counts eligible men, against Massachusetts at 38% — a 2.8× spread that no automatic-registration rule addresses, because no rule has been published. What December 2026 changes →
What this dataset deliberately does not contain
The gaps are findings, not omissions, and several of them are load-bearing. There is no published claim rate, appeal rate, per-board caseload or adjudication time anywhere in the record — the official timeline simply allocates a 50-day window to appeal boards and assumes it suffices. The authoritative medical-disqualification series ends with the FY2016–20 cohort, because a records-system migration broke the data pipeline that produced it. No foreign case selects by lottery, so the selection stage has no modern analog anywhere on earth. And there is no published total-volume requirement for an American draft at all: the standing requirement names three dates and no quantity.
Every one of those statements is a row in the table above, with the coverage document that establishes it.