Flows & positioning

Every day is measured against the median day-size of the previous 250 sessions before it is ranked, because the rupee amounts are not comparable across the archive — the typical day grew about 58× between 1999 and today, so an unscaled percentile would report what year it is rather than whether the day was unusual. This describes the settled series on 2026-10-01, not the provisional tape below it — those are different measurements, and ranking one inside the other's history is wrong by a median 22 percentile points.

Every number on this page is read from primary exchange data · download this view as CSV →

Measured over 72 stored sessions, 2026-06-19 → 2026-10-01. This feed is collected forward from the exchange's daily release and has no backfill, so windows longer than the store are not offered rather than shown empty. With this many sessions a σ reading is a coarse “unusual for the history we have”, not a distribution.

Foreign and domestic institutions publish their cash-market buying and selling after the close. Only the two aggregate categories exist in the exchange feed — there is no cash-versus-derivatives split to show, so none is invented here. The bars and the table beneath them are the provisional T-day tape — the same-day print the market reacted to. The panel above it is the settled depository-confirmed number, which is a different measurement and lands later. One day of institutional flow is close to noise, a twenty-day run is a stance.

We can name 4 fund houses here, across 167 month-ends of published portfolios (2012-09-30 to 2026-07-31) covering 1,599 schemes and 1,259 stocks. 🔴 That month count is every month ANY house published, not a depth every house has — each one starts at a different point, listed below, and a stock is only compared where the same house held it in both months. That is mutual-fund money only — not foreign investors, insurers or pension funds — published about ten days after each month ends. Buying and selling inside a single month cancels out and cannot be seen here at all, so a stock showing no movement is a stock we have no evidence about, not a quiet one.

How far back each house goes: Dsp 162 months, from 2013-01Axis 141 months, from 2012-10Nippon 141 months, from 2012-09Sbi 43 months, from 2020-09

🔴 Sbi published no portfolio for Sep 2025, so nothing below reflects what Sbi did in that month. A house with no published portfolio for a month is a house we have no evidence about in that month — not a house that sat still, and we never fill one in. Months with nothing published, inside each house's own stretch: Sbi 28 months (latest Sep 2025) · Nippon 26 months (latest May 2024) · Axis 25 months (latest Apr 2023) · Dsp 1 month (latest Feb 2013).

Shares bought or sold is the figure after adjusting for splits and bonuses — the plain month-to-month change in share count is not a trade, and on a split it can carry the opposite sign. A stock can appear on both lists if different houses moved opposite ways.

Zero is the drawn reference: above it the group is net long index futures, below it net short. The two sides sat on opposite sides of zero on 3041 of 3653 stored sessions — the mirror is the normal state of this market, not today's news.

Index option lean = (call longs + put shorts) − (call shorts + put longs), in lakhs of contracts. Positive leans bullish, negative bearish; it is a positioning descriptor, not a forecast.

observed: published aggregatesunknowable: per-stock participant split

FII = foreign investors · DII = Indian mutual funds and insurers · Pro = professional trading desks · Client = everyday and wealthy retail. "Long" is a bet up, "short" a bet down. Market-aggregate positioning, published daily after the close — a sentiment descriptor, never a timing signal.

F&O honesty fence: the Phase-0 edge gate (2026-07-24) tested these reads for cross-sectional edge. Only PCR selected at all, weakly (δ≈0.06) and FORWARD-TEST-ONLY; max-pain distance, basis and OI-change failed. This board ranks and flags — it never picks.

Ran and found nothing on this tape: Fresh long on a thin book. A check with no names is a reading, not a gap.

0 = this name's own low · 50 = its middle · 100 = its own high. Percentiles need at least 60 sessions of the stock's own F&O history before they are shown at all, so a blank cell means too little history — never zero.

Every number on this page is read from primary exchange data · download this view as CSV →

Every free tool shows today's F&O snapshot. The context nobody adds is whether today is unusual FOR THIS NAME. A short-covering day on a book that still sits in its 90th percentile of open interest means the shorts merely rolled — they did not leave. Percentiles need at least 60 sessions of a stock's own history before they are shown at all.