AION Veritas - NIFTY, BANK NIFTY & SENSEX Pre-Open Probability Bands

Live bands Validation receipts Model spec Research

A guiding light, not a guiding hand.

AION Veritas prepares pre-open probability bands for NIFTY 50, BANK NIFTY, and SENSEX. According to the 8 November 2026 scheduler audit, the forecast is generated at 08:46 IST, relayed at 08:47, and checked for readiness at 08:50; the opening-window validator runs at 09:31 IST. These are scheduled steps, not proof of successful delivery on every trading date.

The forecast concerns the 09:15 - 09:30 IST opening window, not the closing price. Public receipts document available post-session outcomes, including misses. The historical archive has publication gaps and scoring-method revisions; a single unqualified historical hit rate would be misleading.

Current access: a free 7-day NIFTY 50 trial (verified email and passkey; no card), or a paid single-/all-index subscription. Live pre-open bands require an eligible key; the validation archive is free to read. The official GitHub publisher is AION-Analytics-India.

Veritas informs; it does not instruct. It describes a range to reason around, not a call to act.

Public reporting changes - 14 November 2026

The updated Veritas model page presents a rolling display of the last 30 completed trading sessions, filtered by index. It retrieves validation records dynamically. The static model-page HTML is not itself evidence of any current numeric hit rate or a verified post-calibration cohort.

  • Reference-price hit: A specified opening-window reference falls inside the published band. It does not establish that the complete 09:15-09:30 range remained within the band.
  • Full-range containment: Current rolling receipts do not provide the full-window low/high required for this test, so the website marks it Not measured. The historical containment study is a separate dated evaluation.
  • Other outcomes: Asserted BULLISH/BEARISH direction, NEUTRAL outcomes and minute-test results have separate denominators. Missing and unverified receipts remain visible and are not silently counted as misses.
  • Historical correction: A 13 November 2026 eligibility correction preserves superseded and corrected results in the historical audit.
  • October change: A band-serving configuration change was promoted for NIFTY 50 and SENSEX on 11 October following a short shadow comparison. BANK NIFTY was not promoted. The website does not describe this as model retraining or fine-tuning, and no common verified post-calibration cutoff has been established. Small historical comparisons do not establish future performance.
  • Publication target: Before 08:55 IST. A target is not proof that each session was published on time.
  • Advertised access on this page: Free seven-day NIFTY 50 trial; INR 499/month for one index; INR 1,999/month for all three, subject to the current plan catalogue.

The rolling reference-price test, historical full-range containment audit and post-change results must remain distinct. This repository describes public capabilities and evidence limitations, not proprietary forecasting implementation.

What Veritas is

  • A pre-open probability band for index-level opening-window uncertainty, subject to successful daily production and entitlement.
  • An 80% design coverage target, not an achieved coverage rate; observed performance requires dated, reproducible numerators and denominators.
  • A directional assessment and reliability indication, interpreted according to the method version in use.
  • Public post-session receipts for recorded dates, including misses and documented publication gaps.

What Veritas is not

  • Not a trade signal. It does not say buy or sell.
  • Not a price prediction. It gives a range, not a point.
  • Not investment advice. AION is not a SEBI-registered RA or IA.
  • Not infallible. Misses are published publicly - that is the design, not a disclaimer.

What a Veritas output conveys

A dated band may include a lower bound, upper bound, reported centre, directional assessment, direction reliability, and range confidence. The reported model centre must not automatically be treated as the arithmetic midpoint: field definitions and scoring rules changed during October 2026.

Conceptual field Meaning Boundary
Lower / upper band Bounds of the forecast opening-window interval Not buy/sell levels or a price target
Reported centre Reference point defined by that method version May differ from the arithmetic midpoint
Direction and reliability Separate directional assessment that can signal uncertainty Not a trade instruction or guaranteed result
Validation result Outcome under the dated receipt's scoring contract Compare only properly matched methodology cohorts

This is a conceptual field guide, not an exact machine-readable API schema or a live subscriber band. For actual fields, consult the dated receipt or authorized service output.

How it is scored

Range and direction are evaluated separately. The current validation job is scheduled for 09:31 IST, after the opening window. It is not the forecast-publication time.

Historical receipt contracts are not uniform. A 13 October 2026 revision affected the reported centre and direction interpretation, and a 14 October revision changed first-window scoring. Do not pool legacy and newer outcomes into a headline hit rate without reproducible cohort definitions and denominators.

The 80% coverage figure is a design target, not a measured success rate. Earlier headline performance percentages were withdrawn pending reproducible, method-consistent cohort evidence; no measured accuracy rate is asserted here.

The 8 November 2026 reconciliation identified 104 linked historical receipts, including a previously unlinked 26 June file. It identified 56 missing calendar dates from 24 July through 17 September, which includes non-trading days and must not be described as 56 missing trading sessions. Of the linked receipts, 66 have Day labels and 38 do not.

Dated operational limitation (8 November audit): New historical API rows were not accumulating at that checkpoint pending an archival-only replacement. The current-session API used the 08:46 forecast snapshot. Existing public receipt files are separate. The uploaded 14 November model page does not establish whether historical API persistence has since been repaired. Authenticated API access, signup and checkout were not retested in the 8 November audit.

Dated public example: 5 November 2026 (not a live band)

The original public validation receipt for 5 November 2026 records 0 of 3 ranges held, under both its reported median-window test and its 10-of-15-minute test. It also reports 3 of 3 asserted directional assessments correct. Range and direction are distinct outcomes; a correct direction did not prevent all three range misses.

Index Published lower / upper band First-window reference reported in receipt Minutes inside band Asserted direction and receipt result
NIFTY 50 23,988.99 / 24,288.84 24,354.70 0/15 BULLISH, correct
BANK NIFTY 55,393.68 / 56,275.54 56,914.45 0/15 BULLISH, correct
SENSEX 77,314.07 / 78,272.66 78,722.76 0/15 BULLISH, correct

The reference values above are the receipt's 09:15/09:20/09:25 median opening-window measures, not closing prices or current quotes. The example predates the 5 November previous-close correction documented by the forecaster audit. Its input provenance and applicable scoring rules must be considered before comparisons with newer cohorts. A single day, whether a miss or a hit, does not establish calibrated accuracy, comparative model quality, or a trading recommendation.

Hugging Face Veritas documentation card is a discovery surface; it does not host the proprietary forecasting model, live band endpoint, or downloadable validation dataset. The dated website receipt remains the source for the figures in this example.

Coverage

Three indices, at the index level only:

  • NIFTY 50
  • BANK NIFTY
  • SENSEX

Veritas is index-level. It does not publish sector-level or single-stock bands.

Access

Veritas offers a free 7-day NIFTY 50 trial key with verified email and passkey, no card required. The trial lasts 168 hours from activation, including weekends and market holidays. After the trial, or for BANK NIFTY and SENSEX live bands, a paid subscription is required. The public validation receipts remain available to everyone.

For current plan prices, eligible indices and API conditions, consult the live Veritas access page and API catalogue. Billing terms are intentionally not duplicated in this documentation; the verified product catalogue is authoritative.

Learn more

Collaboration

AION builds research-led probabilistic information systems for Indian markets. For research or engineering collaboration, contact AION Analytics (India): https://aionanalytics.in/

Identity

Published by AION Analytics Private Limited (India) through the official GitHub organization, founded by Lokesh Gupta. Official website | Product registry. Not affiliated with Aion Analytics LLC (United States) or AION Labs (Israel).

Website documentation comparison (14 November 2026): The supplied model-page HTML carries revision veritas-model-20261114-october-shadow, including a rolling 30-session scorecard, distinct performance tests and a historical eligibility correction. This supersedes the 8 November staged-copy warning for this reviewed page. Dynamic rates and receipt completeness require separate data readback.

Disclaimer

Content is for educational and research purposes only. It is not investment advice. AION Analytics is not a SEBI-registered Research Analyst or Investment Adviser. Trading involves substantial risk. SEBI regulations apply. Past performance is not indicative of future results.

Documentation mirror and configuration history

This is a documentation and discovery repository for AION Analytics Private Limited. No AION model weights are distributed, no live AION model inference is hosted here, and no downloadable private datasets are provided. Public validation receipts are maintained on the official website.

The 14 November model page dates the active artifact-set activation to 24 June 2026. The band-serving configuration was promoted on 11 October 2026 for NIFTY 50 and SENSEX; the first forecast under that configuration was generated on 12 October 2026. BANK NIFTY retained its previous configuration. This was a configuration promotion, not model retraining or fine-tuning. The five-session shadow comparison is a small historical sample, not established future performance, statistical superiority or a verified general calibration improvement.

Downloads last month

-

Downloads are not tracked for this model. How to track
Inference Providers NEW
This model isn't deployed by any Inference Provider. ๐Ÿ™‹ Ask for provider support

Collection including aion-analytics-india/aion-veritas