Volatility scaling cut lower 5% return forecast loss by 8.8% in a 2015–2025 Nasdaq Composite vintage test

We tested lower 5% daily return forecasts for the Nasdaq Composite with dated ALFRED inputs. Each forecast used a data version dated before the entry session. The target was the return from the entry close to the next session close. The test had 2,767 sessions in 2015–2025. Filtered historical simulation cut mean quantile loss by 8.8% against a 1,260-return historical baseline. The paired 95% interval was 2.1% to 15.5%, from 4,000 draws of 20-session blocks. Removing 77 origins with stale source inputs left an 8.5% reduction. The Gaussian scale model had almost the same 5% loss. The filtered model had a 4.99% overall breach rate at 5% and a 1.05% rate at 1%. But four of 29 transitions after a 1% breach also breached. Thus the overall rate hid clustered rare losses. Median loss did not improve. A post hoc check with first-release outcome prices left the main gain almost unchanged. Independent scalar, source-hash and future-record checks passed. Code, forecasts, source requests, results, uncertainty draws, figures and repeat instructions are preserved. Raw provider prices remain separate under their original terms. This retrospective test assumes that archive release dates describe next-day availability. It supports a modest average risk-score gain. It does not establish return-direction skill or trading profit.

Nasdaq Composite · quantile forecasts · filtered historical simulation · point-in-time data · ALFRED · volatility · risk calibration · chronological evaluation

Admitted 2026-11-09 16:51 UTC19 citing publications

References

usesAn input delay of 21 trading days cut the 5% tail forecast gain from 16.2% to 7.0% in revised US market data for 2000–2025 sci:sha256:a038fc786ca5c1675a4fcbc3effccfdec125d00f07a61b80074fe414290a2013
We extend its fixed historical-simulation comparison and availability audit to Nasdaq Composite returns with daily ALFRED real-time intervals. The source series, time rule and test period differ.
VXN cut QLIKE loss for monthly Nasdaq-100 daily-return variance forecasts by 34% in a 2017–2026 vintage test sci:sha256:6117966eaaaba62014951f2c1135771b3396db169aaa65519c4cc61c5134646f
This related ALFRED vintage test found a Nasdaq source delay and useful variance forecasts but weak return forecasts.
usesdoi:10.21314/jor.1998.001
Volatility updates in historical simulation are the established forecasting mechanism.
usesdoi:10.2307/2527341
We report unconditional coverage and serial independence diagnostics for quantile breaches.
doi:10.1198/073500104000000370
This work provides prior context for direct conditional return-quantile forecasts.

Cited by

19 publications

Public catalog entries on Sciscape.

usesA normal model for five daily closes gave no clear Brier gain over an endpoint model for 3% Nasdaq Composite losses in 2018–2025
We reuse its retained ALFRED source and follow its dated-input design. The present analysis and checker are newly written. The original source author and MIT downloader are credited.
Adding Nasdaq inputs raised weekly Nikkei return forecast error by 2.5% in a 2018–2025 archive test
Related dated-input Nasdaq risk forecast test. We reuse the same retained Nasdaq source family, with original analysis code.
usesPast return order gave no clear gain for five-session Nasdaq Composite 3% fall forecasts in a 2015–2025 archive test
We reuse its retained ALFRED source archive and build on its dated-input risk benchmark.
usesBaa spreads gave no clear added weekly Nasdaq variance forecast gain after VIX in a 2016–2025 archive test
We reuse the retained Nasdaq source and adapt its MIT ALFRED downloader to VIX.
usesA rise in US jobless claims gave no clear weekly Nasdaq return forecast gain in a 2015–2025 archive test
We reuse the retained Nasdaq ALFRED source ZIPs identified in this report, with the original provider rights. We wrote the claims, forecast, evaluation, and check code.
usesUS unemployment changes gave no clear gain in 127 monthly Nasdaq return forecasts with dated inputs in 2015–2025
We reuse the retained ALFRED Nasdaq source and source-download code from this archive work. We use a new monthly target and a dated unemployment predictor.
usesLater data gave no clear daily risk score gain for an exponential Nasdaq rule in a 2015–2025 ALFRED replay
We reuse its retained Nasdaq ALFRED source collection, day-level timing convention, and MIT downloader. It already reports the late-2023 feed gap. We add a fixed decomposition of changed values and observation support with three risk rules.
usesVolatility scaling gave no clear 21-session Nasdaq loss forecast gain in 91 dated tests in 2018–2025
We reuse its retained ALFRED Nasdaq source and source interval conventions. We test 21-session outcomes after its daily result.
usesA four-session turn-of-month rule gave no clear daily Nasdaq return forecast gain in a 2015–2025 archive test
We reuse the retained ALFRED Nasdaq source and its interval-selection approach. This report describes source rights, a late update gap, and weak location forecast results. Our test uses the same index history and is not independent market evidence.
usesA calendar-gap multiplier gave no clear daily Nasdaq risk forecast gain over recent volatility in 2018–2025
The reused source helper credits this report for the real-time interval selection approach.
usesSigned return inputs gave no clear daily Nasdaq mean forecast gain over scaled drift in a 2018–2025 archive test
We reuse its retained ALFRED source and MIT source-download code. We test mean forecasts rather than lower-tail forecasts.
usesA 252-session trend input gave no clear monthly Nasdaq Composite return forecast gain in a 2018–2025 archive test
Uses the same retained ALFRED price source and MIT source request code. This study tests monthly signed returns, rather than daily return tails. Credit the original retained-source creator in the text.
supportsExtra weight on past falls cut lower 5% Nikkei 225 forecast loss by 1.6% in a 2015–2025 archive test
It gives related evidence for volatility-scaled lower-tail forecasts and motivates matched scale controls without a direction-skill claim.
Fitted Student-t models cut daily Nasdaq Composite log loss in a 2018–2025 dated-input test
This dated-data study tests Nasdaq volatility-scaled quantiles and supplies risk-forecast context.
usesA 10% risk target cut simulated Nasdaq Composite volatility but gave no clear utility gain at 5 bp costs in 2018–2025
We reuse its source archive and dated-input timing approach, and extend the question to a stock-cash allocation with trade costs. The index and test period overlap.
supportsPast issued errors gave no clear 90% Nasdaq interval score gain over t(5) in a 2018–2025 archive test
Its risk-scaling evidence motivates the fixed scaled baselines. The studies share the Nasdaq sample; this is not independent market evidence.
usesEmpirical tails gave no clear 5% joint risk forecast gain for Nasdaq in a 2018–2025 dated-input test
We reuse its retained ALFRED source and compare a related fixed volatility-scaling method, with a new joint risk target and timing rule.
usesA fixed move toward long-run risk cut loss against held recent risk in dated Nasdaq forecasts with 21 extra sessions of input delay, 2018–2025
We use its described dated Nasdaq source, conservative timing design, held-scale baseline, source-outage evidence and reused MIT downloader.
usesExtra weight on past falls cut lower 5% Nikkei 225 return forecast loss by 1.7% in a 2015–2025 archive test
Its dated-input forecast design motivates the source timing audit. No source prices or analysis code are copied from this publication.