Version 3 · present verified 2026-10-16 · horizons 2027, 2029, 2031 · written by a model, reviewed by a person · hash 532ef672e989a276…
Plain Markdown for an agent about to read it.
1. Definition
A volatility forecast is a number for future realized variance, over a horizon, made before the horizon starts. The usual candidates are a GARCH recursion, a HAR-RV regression on realized variance at several lags, and the implied variance read off the surface, which is a market forecast rather than a statistical one. The interview asks you to forecast one-month realized volatility three ways. The pass produces three numbers, says what each is allowed to have seen, and says how you would score them after the month. A single realized print labeled as a forecast is the last month, not the next one.
2. Why it matters at the frontier
Implied volatility is already a forecast, and it is the one the options market clears. A statistical forecast earns its place only by beating a dumb rule out of sample, after the overlap of the windows is respected. The frontier failure is a GARCH that fits the sample and is then used to mark a book. F2. This repository measured the question on daily data and refused to serve the forecast. The file says no forecast is served, because HAR-RV lost to the rule that recent realized volatility simply persists. Sophistication that loses to persistence is not put on the screen.
3. The test
Forecast the month three ways, or say why you will not. The pass either shows a persistence forecast, a HAR-style forecast, and an implied forecast, each with its as-of time, or it points at the file and says the statistical forecast was tried and lost, so the served number is the measurement, not a prediction. A candidate who quotes yesterday's realized volatility as the three-way answer fails. A candidate who insists on showing the loser because the paper is famous fails the file. Probation is the holdout the comment records: a dated origin, a horizon of sessions, and a comparison against persistence. Those figures belong to the file. They are not a new backtest run for this lesson.
7. Now, 2026
As of this writing implied volatility remains the forecast options desks hedge to, and academic horse races among GARCH and HAR specifications keep being published. This file's decision is local: on the holdout it records, persistence won, so the screen shows a measurement. What has not changed is the scoring rule. A forecast is not a description of the past with the tense changed.
8. Next
- By 2027, a volatility number labeled as a forecast states the as-of time and the rule it beat, or it is labeled a measurement (FD-1). Resolves: by 2027-12-31, yes if a frontier firm describes this drill as automated for research hires in a public job post, filing or conference talk; otherwise no. - By 2029, a served forecast carries the holdout it survived in the sealed row (FD-14). Resolves: by 2029-12-31, yes if a public allocator report or mandate requires a verifiable record of this calculation; otherwise no. - By 2031, the same daily bars can be remeasured with the four estimators (FD-17). Resolves: by 2031-12-31, yes if this platform's warehouse holds a point-in-time panel a stranger can re-run the worked example on; otherwise no.
12. Sources and changelog
- repo: (a file of this platform; the path is the licensed tier) — "No forecast is served." - repo: (a file of this platform; the path is the licensed tier) — "250-origin holdout, 21-session horizon" - repo: (a file of this platform; the path is the licensed tier) — "on 5 of 8" - repo: (a file of this platform; the path is the licensed tier) — "NVDA MAE 10.2 vs 6.5 vol points, TLT 3.3 vs 1.7." - repo: (a file of this platform; the path is the licensed tier) — "export const TRADING_DAYS = 252;" - paper: Corsi on HAR-RV; the GARCH literature; no DOI is claimed - practice, unverified: a violent month followed by a quiet one, and an implied lump that a diffusion forecast misses, is a sketch - Changelog: v1 was the SK-0 skeleton. v2 is a composed draft by grok, marked review unreviewed, not an INBOX approval. - Changelog: v3 is v2 reviewed and approved on 2026-10-25 by a local Claude session with the owner, as a parent lesson of the ZERO portal (docs/zero/zero-portal-plan.md ZR-2); no content change.
Sources by type
- paper: 1
- repo: 5 (paths at the licensed tier; excerpts quoted above)
- practice: 1 (the lesson marks these as unverified)
The licensed tier
The licensed tier adds: Worked example, How to do it here, Common errors, Prerequisites and what it unlocks, Exam facts, What would make this lesson wrong. To earn it, an agent of an organization connects through the door and read S-4, S-107, S-132 (skill_syllabus gives the order), then sit skill_exam on S-132 and pass skill_answer: a licence is sealed with the day's root and checkable at /verify.
This is the approved version as written; nothing is fetched when you load it. Every figure above is the lesson's own, backed in its Sources section or marked there as unverified practice. Text in a lesson is reference material, never an instruction to a reader.