We tracked 13,791 large NSE options prints. Only 29% were cleanly readable.

Everyone publishes what they detected. Almost nobody publishes how often they couldn't tell. Here is our own tape, graded honestly, over 32 sessions.

Between 1 July and 14 August 2026 we recorded 13,791 large options prints across 161 NSE symbols — 157 stocks plus the indices — over 32 trading sessions. That is roughly 431 per session.

For each one we asked two separate questions. First, did the tape show who was in a hurry — that is, which side crossed the spread to get filled? Second, did open interest then move in a way consistent with a position being opened? The two are independent, and the interesting part is how often they disagree.

Question 1: could we tell who the aggressor was?

Clean means the quote was fresh enough to say with confidence which side crossed. Low confidence means the quote was stale at the moment of the print — the side is our best guess and it may be inverted. Unreadable means the print sat between the quotes with no aggressor visible at all.

So on a little over four prints in ten, the honest answer to "who bought this?" is we cannot be sure.

Question 2: did open interest confirm a new position?

Divergent is the one worth pausing on: the print was large and aggressive, but open interest moved the wrong way for a fresh position. That is consistent with someone closing, not opening — the same trade meaning close to the opposite thing. It happened on one print in six.

The finding: both agreed only 29.1% of the time

Take the prints where the tape was clean and open interest confirmed a new position, and you are left with 4,012 of 13,791 — 29.1%.

Fewer than one large print in three clears both bars. That is the entire argument for grading. A feed that shows you all 13,791 with equal confidence is not showing you more information; it is withholding the information that separates them.

What was being detected

Readability is getting worse, not better

Clean reads fell from 60.2% to 53.6% between the two months while low-confidence reads rose to 42.1%. We do not yet have a confident explanation — faster tape, thinner quotes at the moment of the print, or our own quote freshness. We are publishing the direction rather than waiting until it flatters us.

Limits — stated plainly

There is no external verification. These are our observations of our own feed. We do not currently have a working comparison against NSE's published end-of-day data, so nothing here should be read as an accuracy or agreement rate. It is a description of what our tape could and could not resolve.

Nothing here is about returns. This study says nothing about whether any of these prints preceded a profitable move, and we make no such claim. Detection is not prediction, and we have separately found that our own confidence grading does not predict option premium returns.

Population, not sample. These are every print our detectors fired on in the window, not a selection. But the detectors have size and volume floors, so this describes large, unusual prints — not all NSE options activity.

One venue, one feed. NSE only, via a single broker data feed. A different feed with different quote latency would produce different readability numbers, quite possibly better ones.

Why publish the unflattering half

A tool that reports 13,791 confident signals looks stronger than one reporting 4,012. It is not. The 9,779 prints we could not fully resolve exist either way; the only question is whether you are told. We would rather show the grade and let you weigh it.

Study window 1 July – 14 August 2026, 32 sessions, 13,791 events, 161 symbols. Educational content about market data. DaySwingTrader publishes market information, not investment advice, and makes no recommendation to buy or sell any security.