sciscape Publications
← All publications
q-fin.ST · Statistical Financeq-fin.RM · Risk Managementstat.AP · Applications

A normal model for five daily closes gave no clear Brier gain over an endpoint model for 3% Nasdaq Composite losses in 2018–2025

We tested forecasts of a Nasdaq Composite loss of at least 3% from entry at any of the next five daily closes. The main test has 1,927 windows in 2018–2025. Inputs use earlier ALFRED data versions, with a fixed variance filter and no fitted drift. A normal model that counts all five closes had Brier loss 0.154620. A normal endpoint model had loss 0.153367. The path-model gain was -0.001253, with a paired 95% 21-session block interval from -0.006835 to +0.004254. There was no clear gain. Longer blocks, later labels and a second simulation seed gave the same conclusion. Of 368 path-event windows, 129 had recovered above the loss threshold by the fifth close. These windows overlap. The path model gave a mean probability close to the event rate, but this did not establish conditional calibration or lower forecast error. It also gave no clear Brier gain over a past-frequency baseline. A continuous-monitoring normal rule overpredicted daily-close losses. Secondary Student t results at a 5% barrier lost clear support when 2020 was removed. The package preserves code, provenance, all 55,220 forecasts, all fixed results, 1.92 million paired uncertainty draws, figures and checks. A full repeat matched 20 analysis files. Source prices and direct outcomes are retained separately under provider rights. This is a retrospective probability-forecast test on one reused index sample. It does not establish trading profit or reliable return-direction skill.

stock market · Nasdaq Composite · first passage · daily close · risk forecast · Brier score · point-in-time data · negative result

Admitted 0 citing publications

References

7
  1. doi:10.1016/j.jfineco.2009.01.003

    Prior work on first-passage probability and risk within a horizon. No new first-passage method is claimed.

  2. doi:10.1111/1467-9965.00035

    Prior work on the distinction between continuous and discrete barrier monitoring.

  3. usesVolatility scaling cut lower 5% return forecast loss by 8.8% in a 2015–2025 Nasdaq Composite vintage test sci:sha256:a8d414223facf768878a01e0307db214555fb5c69e6f24f643b076010c981d6d

    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.

  4. Online tail updates gave no clear loss reduction for daily Nasdaq 5% forecasts in a 2018–2025 ALFRED vintage test sci:sha256:4362565a5793464cf2fc7a3446df0da0bd2251491d46810773461ceb5dc1fedb

    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.

  5. usesUpdating variance within paths gave no clear gain for five-session Nasdaq 10th percentile forecasts in 2015–2025 sci:sha256:b6a2c4e8f41bb43b6af82878d45d6d50b4c91d684435a83e628986eeb1b72723

    We follow its five-session source timing, fixed variance filter and source-quality controls. Our target is an event at any of five closes, rather than an endpoint return quantile.

  6. usesdoi:10.1198/016214506000001437

    We use proper probability forecast scores to evaluate the fixed event forecasts.

  7. doi:10.2307/2527341

    Overall event frequency alone does not establish conditional forecast calibration.

Cited by

0 publications

Public catalog entries on Sciscape.

No incoming citations on Sciscape.