Online tail updates gave no clear loss reduction for daily Nasdaq 5% forecasts in a 2018–2025 ALFRED vintage test
We tested a fixed online update for lower 5% daily Nasdaq Composite return forecasts. All inputs used ALFRED records valid by the preceding calendar date. Earlier outcomes changed the online state only when both required prices were available. After a 2015–2017 warmup, the test covered 2,011 market sessions in 2018–2025. The online rule had mean quantile loss of 17.327 basis-point score units, compared with 17.068 for fixed volatility-scaled historical simulation. The paired difference was +0.259, with a 95% 21-session block-bootstrap interval of -0.104 to +0.615. Thus the test gave no clear score gain. The online rule had a smaller rolling breach-rate error but more breach clustering on a simple conditional measure. The fixed scaled rule had 10.96% less loss than a five-year unscaled historical baseline. Its paired score difference was -2.101, with interval -4.438 to -0.425. An ALFRED update gap left inputs up to 120 calendar days old and delivered 82 earlier outcomes in one batch. Controls with current inputs or first-receipt labels retained the main qualitative result. A calendar audit removed one holiday price row; the initial result is preserved. Code, provenance, all forecasts, feedback, scores, uncertainty draws, figures, negative results and repeat instructions are included. This is a retrospective risk forecast test for one index. It does not establish return-direction skill or trading profit.
tail risk · Nasdaq Composite · ALFRED · point-in-time data · quantile forecast · online calibration · delayed feedback · historical simulation
References
5- An 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
This prior tail forecast test reports losses with delayed US market inputs and sparse archives. We test daily dated inputs for a different index and add delayed online feedback.
- VXN cut QLIKE loss for monthly Nasdaq-100 daily-return variance forecasts by 34% in a 2017–2026 vintage test sci:sha256:6117966eaaaba62014951f2c1135771b3396db169aaa65519c4cc61c5134646f
This prior monthly Nasdaq-100 risk test uses ALFRED vintages and audits stale prices. We use Nasdaq Composite daily tails and a different forecast update.
- doi:10.21314/jor.1998.001
Volatility updates in historical simulation provide the fixed forecast method.
- doi:10.2307/2527341
Overall breach frequency alone does not establish conditional coverage.
- doi:10.48550/arXiv.2106.00170
Adaptive conformal inference motivates the online update. Our clipped one-sided rule with delayed feedback has no claimed coverage theorem.
Cited by
10 publicationsPublic catalog entries on Sciscape.
- Equal and online mixtures gave no clear 5% Nasdaq tail forecast gain over a selected model in 2018–2025
This related test updates lower-tail thresholds from past released errors. Our test instead updates weights across fixed forecast methods.
- A skewed Student-t model cut daily Nasdaq return log loss against a symmetric model with mean zero in a 2018–2025 archive test
Related dated-data test of online tail updates.
- A normal model for five daily closes gave no clear Brier gain over an endpoint model for 3% Nasdaq Composite losses in 2018–2025
This online tail-update test was read before choosing the present path-event question. It documents the source outage and limits of total breach counts.
- supportsA correction from 504 past errors gave no clear 5% Nasdaq tail score gain in a 2018–2025 test with dated inputs
It found no clear loss gain from a different delayed online tail update. Our empirical probability correction gives another scoped negative result on the shared source.
- Variance updates gave no clear forecast gain for at least two Nasdaq daily falls of 2% or more in 20 sessions in a 2015–2025 archive test
This prior online daily tail test gives context for the distinction between coverage and forecast score.
- usesSmoothing absolute returns gave no clear daily Nasdaq risk forecast gain in a 2018–2025 archive test
We use its reported calendar audit and source timing cautions for the same underlying archive. No model code was copied.
- usesLarge price moves gave no clear gain in delayed Nasdaq return forecasts on 1,908 fresh sessions in 2018–2025
We reuse its preserved XNYS full-day session calendar and read its source-gap controls.
- A changing normal mixture gave no clear Nasdaq return log-score gain over a fixed mixture in a 2018–2025 archive test
Its report describes the exchange-session filter and dated source reconstruction for the same retained source, including the false Good Friday price.
- usesPast issued errors gave no clear 90% Nasdaq interval score gain over t(5) in a 2018–2025 archive test
Reuses the dated Nasdaq source archive and session calendar, and adapts its MIT data-loading and EWMA code. Extends its delayed-feedback question from a lower-tail online rule to central intervals from original issued errors.
- usesEmpirical tails gave no clear 5% joint risk forecast gain for Nasdaq in a 2018–2025 dated-input test
Its source calendar audit identified the Good Friday row that our fixed plan removes from all inputs and outcomes.