4A.5 Breadth, participant data, delivery and deals

Motivation

The Nifty is a free-float-cap-weighted index (M01 · 1.9). A rally led by five heavyweights and a rally in which most stocks rise produce the same index return and very different markets. Breadth statistics measure participation. NSE also publishes, every evening, who holds the open futures and options positions and how much of each stock's volume ended in delivery. Each dataset answers a narrow question exactly and a broad question badly.

4A.5.1Breadth#

Definitions

Over a universe of $N$ stocks (Nifty 500 is the usual choice on NSE), on day $t$ let $A_t$ , $D_t$ be the numbers of advancing and declining stocks.

Advance–decline line: $\mathrm{AD}_t=\mathrm{AD}_{t-1}+(A_t-D_t)$ .
Percentage above the 200-day moving average: $\frac1N\#\{i: P_{i,t}>\mathrm{SMA}_{200}(P_i)_t\}$ .
New highs minus new lows: the number of stocks at a 52-week closing high minus the number at a 52-week low.

All three are equal-weighted counts of signs; the index is a cap-weighted mean of returns.

See it live: set the Nifty 50 beside the sectoral indices on https://oicharts.com/indices and https://oicharts.com/sectors; when the index is up but most sectors are down, the rally is narrow. Compute the breadth statistics themselves from the NSE bhavcopy.

Proposition 4A.5breadth in a one-factor model

Let $r_{i,t}=\beta_if_t+\varepsilon_{i,t}$ with $\varepsilon_{i,t}$ independent $N(0,s_i^2)$ given $f_t$ . Then, conditional on the factor return $f_t$ , the expected fraction of advancers is

\mathbb E\Big[\frac{A_t}{N}\,\Big|\,f_t\Big]=\frac1N\sum_{i=1}^N\Phi\!\Big(\frac{\beta_if_t}{s_i}\Big), \tag{4A.5}

and as $N\to\infty$ with bounded $s_i$ and independent idiosyncratic terms, $A_t/N$ converges to this conditional mean.

Proof

$P(r_{i,t}>0\mid f_t)=P(\varepsilon_{i,t}>-\beta_if_t)=\Phi(\beta_if_t/s_i)$ ; average over $i$ . Conditional independence and Chebyshev give the law of large numbers.∎

So in a pure factor market breadth is a deterministic, increasing function of the factor return: it carries no information beyond the index. Breadth diverges from the index when the index return is not the factor return: when a few large-cap names (with idiosyncratic moves) dominate the cap-weighted index. A narrow rally means the cap-weighted return exceeds what (4A.5) implies for the median stock. That is a statement about concentration, and its usefulness for forecasting is an empirical question. We are not aware of a robust, out-of-sample study showing that breadth divergences predict Nifty returns after costs; treat them as context about how widely a move is shared, and as a risk flag (a narrow index is less diversified), not as a timing signal.

4A.5.2FII/DII cash flows#

NSE publishes provisional daily net purchases by foreign and domestic institutional investors in the cash market after the close; NSDL publishes the FPI custodial data (M11 · 11.12 has the definitions, the currency arithmetic and the 2026 numbers). Three cautions for daily reading: the provisional figures are revised; same-day correlation with returns mixes price impact with feedback trading, so it says nothing about causality; and primary-market flows (IPOs, block placements) can make FPIs net buyers on a day they sold in the secondary market.

4A.5.3Participant-wise open interest#

NSE publishes daily, for each of four participant types (Client, DII, FII, Pro), the number of open contracts long and short in index futures, stock futures, and index and stock calls and puts (file fao_participant_oi_DDMMYYYY.csv on nsearchives.nseindia.com/content/nsccl/). "Client" is broadly retail and non-institutional customers of brokers; "Pro" is brokers' proprietary trading.

Accounting identity Every open contract has one long and one short, so in each column group the longs across the four participants sum to the shorts. In the 27 Oct 2026 file: index futures long $251{,}789+47{,}790+33{,}588+25{,}297=358{,}464$ , equal to the shorts $57{,}363+8{,}410+254{,}645+38{,}046=358{,}464$ . The same holds for calls (3,060,791) and puts (3,206,757). Positioning is zero-sum: one group's net long is the others' net short.

Worked example 4A.4FII index-futures positioning, Oct–Nov 2026

From the NSE participant-wise OI files (counts of contracts):

Date	FII index futures long	FII short	FII long share	Client long share
27 Oct 2026	33,588	254,645	11.7%	81.4%
5 Nov 2026	52,037	189,475	21.5%	69.4%

Between the two dates the Nifty rose about 7.9% (from 22,749.75 on 27 Oct, Yahoo, to 24,536.6 on 5 Nov, M11 · 11.12). FIIs cut their net index-futures short from 221,057 to 137,438 contracts. A broker note would call this "FII short covering". What the file cannot tell you:

Why. An FII short in index futures can be a hedge against a long cash portfolio (FPIs held a large stock of Indian equities throughout), an arbitrage leg, or a directional bet. The file does not link segments.
Size in rupees. "Index futures" pools Nifty, Bank Nifty and other index contracts with different lot values. A contract count is not a notional.
Delta. Option columns count contracts regardless of strike; a deep OTM call and an ATM call count the same. M09A develops the options side and the delta problem; SEBI moved index position limits to delta-based futures-equivalent OI in 2025 for this reason (M07 · 7.5).
Direction of causality. A ratio at an extreme after a large move is mostly a description of the move. Whether it predicts the next move is a hypothesis to test with the M04 protocol.

4A.5.4Delivery percentage#

For each stock and day, NSE's full bhavcopy (sec_bhavdata_full_DDMMYYYY.csv) reports traded quantity, deliverable quantity and delivery percentage = deliverable quantity / traded quantity. Shares bought and sold by the same client on the same day net out and are not delivered; the deliverable quantity is what settles (T+1).

On 5 Nov 2026 (NSE full bhavcopy, EQ series): median delivery percentage across 2,742 stocks 55.2%; among the 100 highest-turnover stocks, 48.9%. Selected names: Reliance 46.5%, HDFC Bank 63.2%, ICICI Bank 70.8%, TCS 61.8%, Titan 40.5%.

The folk reading is "high delivery means genuine buying". The caveats:

Every delivered share has a buyer and a seller. High delivery means positions changed hands for longer than a day, not that buying dominated.
Delivery percentage is mechanically higher where intraday trading is light: illiquid stocks, stocks in the trade-for-trade segment, and days when intraday traders are absent. Compare a stock with its own history, not with other stocks.
Large institutional trades, block deals and index-rebalance trades are delivery trades. A spike can be a single fund's trade.
Use it as an input to a tested model (for example, relative delivery volume as a variant of the Gervais–Kaniel–Mingelgrin volume shock), not as a reading.

4A.5.5Bulk and block deals#

A bulk deal is a trade or trades in a day by one client in a stock whose total quantity exceeds 0.5% of the company's listed equity shares; the broker must report it and the exchange publishes it the same day (SEBI's long-standing disclosure rule; check NSE's bulk-deal page for current wording).
A block deal is a large single trade executed in a separate block-deal window. SEBI's circular of 8 Oct 2025 (SEBI/HO/MRD/POD-III/CIR/P/2025/134), effective about 7 Dec 2025, raised the minimum order size to ₹25 crore and revised the windows and price band (reported by SCC Online, Taxmann and others; secondary reports give a ±3% price band around a reference price and a VWAP published around 14:00–14:05 for the afternoon window; read the circular for the exact times).

Bulk and block deals name the counterparties, which makes them popular reading. The signal problems: a seller in a block deal is often a promoter, private-equity fund or strategic holder selling for reasons unrelated to valuation (fund life, pledges, regulatory minimum public shareholding), and the buyer list is usually funds already in the stock. Any study must control for the discount at which blocks are placed (a mechanical next-day effect) and for event clustering. These are event studies in the sense of M12 · 12.1, not readings.