AION Veritas

Pre-open probability bands for the major Indian indices. A range to position around, never a call to act.

The problem

Before the market opens, every trader wants the same thing: some sense of where the index is likely to trade. What is available is mostly one of two useless extremes.

At one end, confident directional calls, "NIFTY to 25,400 today", which are advice dressed as analysis, are wrong often enough to be dangerous, and conveniently disappear when they miss. At the other end, silence, or commentary so hedged it carries no information at all.

Neither helps you size a position. A trader does not need to be told what to do. They need to know the plausible range, and they need to know how often the source of that range has been wrong before.

How it solves it

Veritas publishes a pre-open probability band for NIFTY 50, NIFTY BANK and SENSEX: a range, with the confidence attached to it, before the session begins.

It states a range rather than a direction on purpose. A range is something you can position around and size against. A direction is an instruction, and issuing instructions is not what we do.

Every prediction is scored afterwards and published, including the misses. The daily validation receipts are permanent and public, so the hit rate is something you can audit yourself rather than something we assert. A forecaster who publishes only their successes is not giving you information; they are giving you marketing.

How to use it

Daily bands: Published pre-open on the Veritas page.
The scorecard: Every dated receipt, hits and misses alike, at validation receipts.
Specification: Scope, inputs and known limits in the model specification.

How to read a band

The band is a range of plausible outcomes, not a target and not a prediction of a close.
A band that is wide is telling you something real: the session is genuinely uncertain. Narrowing it artificially would be a lie that makes the output look better.
Input quality on any given morning bounds output quality. Thin overnight data produces a weaker band, and the receipt says so.

Today's bands See the misses

Changes to the validation receipts

12 October 2026. Receipts dated 13 October 2026 onward change in four ways. They show the model's own centre for each band in place of the middle of the band edges. They report direction in two groups: asserted calls, where the model committed to UP or DOWN, and neutral calls, where the model declined to assert; a neutral call is correct if the move from the previous close stays inside the stated tolerance and a miss if the move breaks it. The tolerance is 0.20 percent scaled by the previous close of India VIX divided by 14, kept between 0.10 and 0.50 percent, and it is the same for all three indices on a given day (0.21 percent at a VIX of 14.4). They show the minutes the index spent inside the band beside the band-held count. They state that receipts are published the same morning, after the first-15-minute window is scored. Receipts dated 12 October 2026 and earlier are left exactly as published. From 13 October 2026 the published direction comes from a simpler overnight rule, with its measured hit rate as its confidence; the earlier direction model is scored beside it. (Wording shortened on 20 October 2026: model method details are not published.)

14 October 2026. The first 15 minutes of trade is 09:15 to 09:30, 15 one-minute candles. Sessions before 14 October 2026 were scored on 14 candles: the request ended at 09:29 and the last candle was not returned. The pass line is 10 of 15 (66.7 percent) from 14 October 2026 onward. The stricter 11 of 15 (73.3 percent) was not adopted, because it would fail sessions that passed under the earlier 10 of 14 (71.4 percent). Historical receipts keep the 10 of 14 they were published under; nothing is recalculated.

What this is not

Informational and educational only. Not investment advice, not a recommendation to buy or sell, and not a SEBI-registered Research Analyst or Investment Adviser product. AION is an infrastructure provider: we publish information and leave the decision entirely with you.

We publish what a system does and what it produces. We do not publish how it is built. Research partners and institutional collaborators can request a technical briefing under confidentiality: [email protected]