DEV Community

AION Analytics for AION Analytics (India)

Posted on Originally published at aionanalytics.in

Do not trust our market forecast. Audit it with 15 lines of Python.

Every trading morning before 08:55 IST, we publish a band for where NIFTY 50 will trade in the first 15 minutes of the session. Every afternoon, we publish whether it held. The receipts are public, dated, permanent, and include the misses.

That is a claim anyone can make. So instead of asking you to trust it, this post shows you how to check it yourself with a few lines of Python, and what you will find when you do.

The product is Veritas, from AION Analytics (India). The NIFTY 50 band is free every trading day.

What gets published, and when

Before the open (by 08:55 IST): a lower and upper bound for NIFTY 50 in the first 15 minutes of the session, the midpoint, a directional lean, and how confident the model is in that lean. It appears on aionanalytics.in/veritas and, word for word, on our Telegram and Discord channels.

After the session: a receipt at aionanalytics.in/veritas/validation for NIFTY 50, BANK NIFTY and SENSEX, with the predicted band, the actual first-15-minute price (the median of 09:15, 09:20 and 09:25), whether the band held, whether the direction was right, and the confidence published that morning. Receipts are never edited.

Here is the NIFTY 50 part of the receipt for 30 September 2026:

Predicted band 22,575.58 to 22,818.92
Actual first-15-minute price 22,702.85
Band Held
Directional lean Bullish, 44.7% confidence
Direction Wrong

The band held and the direction was wrong. The receipt says both, and it says the model was close to a coin flip on direction that morning, which it had also said before the open. We tell people to rely on the range and ignore a weak lean, and this is why.

Audit it yourself

The receipt index is a plain public page. This reads it and counts:

import re
import requests

URL = "https://aionanalytics.in/veritas/validation"
page = requests.get(URL, timeout=30).text
text = re.sub(r"<[^>]+>", " ", page)

# Each receipt on the index reads like: "2026-09-30 Day 50 3/3 bands held"
rows = re.findall(r"(20\d\d-\d\d-\d\d)\s+Day\s+(\d+)\s+(\d)/(\d) bands held", text)

sessions = len(rows)
held = sum(int(h) for _, _, h, _ in rows)
total = sum(int(t) for _, _, _, t in rows)
all_missed = [d for d, _, h, _ in rows if h == "0"]

print(f"{sessions} sessions, {held}/{total} index-bands held ({held / total:.1%})")
print("sessions where every band missed:", all_missed)
Enter fullscreen mode Exit fullscreen mode

When we ran it on 1 October 2026 it printed:

42 sessions, 91/126 index-bands held (72.2%)
sessions where every band missed: ['2026-09-24', '2026-07-10', '2026-07-02']
Enter fullscreen mode Exit fullscreen mode

What that number means, honestly

The band is designed to hold about 8 times in 10. On the receipts currently in the index, it has held a little over 7 times in 10. That is below the design target, and we would rather you find that out from our own receipts than from somebody's screenshot.

Three more things you should know before reading too much into it:

  • There is a gap. Veritas was not running from 24 July to 17 September 2026, because its market data feed stopped when a broker access token expired. There are no receipts for those dates, and none were back-filled. The validation page says so at the top.
  • The index does not yet list everything. Public tracking started on 7 April 2026, which is why the latest receipt is Day 50. The index currently lists receipts from June onward; the April and May receipts exist but are not in the index yet. The script above only counts what the index lists.
  • 42 sessions is a small sample. It is enough to see whether something is badly wrong. It is not enough to claim a precise rate in either direction.

Why bother publishing it

Because a forecast that only shows its hits cannot be evaluated. The ordering is what makes this a test: the band is committed before the market opens, and the score is written after. You can run the script above next month and see whether the number moved, and so can anyone else.

Using the free band

If you want the band in your own morning routine:

  1. Read it on aionanalytics.in/veritas after 08:55 IST, or follow the Telegram or Discord channel linked from that page.
  2. Treat the range as the main output. Look at the confidence before giving the directional lean any weight; below 50% it is close to a coin flip, and the post says so.
  3. Check the receipt after the session.

BANK NIFTY and SENSEX bands, JSON and CSV export, and API access are for subscribers: ₹499 a month for one index of your choice, or ₹1,999 a month for all three. The receipts for all three indices stay free either way.

What it is not

Veritas is a probability range for the opening minutes, not a forecast of where the day ends. It is not a buy or sell call and not investment advice, and AION is not a SEBI-registered Research Analyst or Investment Adviser. The band can and will break; the receipts are there so you can see how often.

If you find a receipt that looks wrong, or the script stops working because we changed the page, the feedback box at the bottom of the Veritas page comes straight to us.

Top comments (0)