tejhq/indian-markets

End-of-day data for every NSE and BSE listed equity, built straight from the exchanges' official bhavcopy. Six parquet trees: raw prices, corporate actions, back-adjusted prices, symbol history, derived metrics, and a survivorship-bias-free liquidity universe. Refreshed every trading day by an open pipeline.

No broker, no auth, no scraping. The same data is served at api.tejhq.dev and mirrored on Cloudflare R2.

Coverage

Exchange	Series	Instruments	Coverage
NSE	EQ, BE, BZ	~2,300 / day	2010-01-04 to today, ~4,100 trading days
BSE	A, B, T	~2,200 / day	2024-07-08 to today

NSE and BSE both moved to the SEBI CMTS bhavcopy format on 2024-07-08. The pipeline parses both the modern CMTS layout and the legacy NSE layout, which is how NSE reaches back to 2010. Legacy BSE files identify instruments by numeric scrip code with no clean bridge to modern tickers, so BSE starts at the cutover.

Derived trees are computed per instrument chain, not per raw ISIN: when a face value split or re-listing issues a new ISIN under the same symbol within 30 days, adjustment factors, rolling windows and universe eligibility carry across the change. The isin column always holds the value the exchange reported on that day.

Pre-2012 NSE rows carry no isin and no trades because the source never had them. They are null in the raw tree rather than invented. The derived trees backfill the ISIN for symbols that traded continuously across the 2012 cutover so corporate actions and rolling windows apply back to 2010; symbols that did not survive the cutover stay ISIN-less and are windowed by symbol.

The six trees

Pick a config in the viewer above to browse each one.

1. nse/, bse/: raw bhavcopy

One rollup per exchange per year: <ex>/year=YYYY/<ex>_YYYY.parquet. The current year's file is rewritten every trading day.

Field	Type	Notes
date	date	Trading date
symbol	string	Ticker, e.g. RELIANCE
series	string	Exchange series code
isin	string	ISIN; null before 2012 on NSE
name	string	Full instrument name
open, high, low, close	float64	Prices in INR
last	float64	Last traded price
prev_close	float64	Previous close
volume	int64	Shares traded
turnover	float64	Value traded, INR
trades	int64	Trade count; null before 2012 on NSE

2. actions/: corporate actions

One file per exchange per year: actions/<ex>_YYYY.parquet. NSE from 2010, BSE from 2024.

Field	Type	Notes
exchange	string	NSE or BSE
symbol	string	Ticker on ex_date
isin	string	As reported by the source
company	string	Issuer name
ex_date	date	First day the price trades ex the action
record_date	date	Null when not reported
type	string	dividend, split, bonus, rights, buyback, demerger, merger, agm, other
ratio_num, ratio_den	int64	Bonus and rights ratio
cash_amount	float64	Per-share cash for dividends
face_value_from, face_value_to	float64	For splits
raw_subject	string	Verbatim source text, kept for audit

3. prices_adjusted/: back-adjusted prices

One file per exchange per year. Every raw bhavcopy column, plus:

Field	Type	Notes
adj_factor_cumulative	float64	Product of factors of all actions with ex_date after this row, per instrument chain. 1.0 when none.
adj_close	float64	close * adj_factor_cumulative. Continuous through splits, bonuses and dividends.

Dividend factor follows the NSE convention, (prev_close - cash) / prev_close, with prev_close taken from the session before the ex date across the full history; several payouts on one ex date are summed. Demergers, rights, buybacks and mergers are not scaled, which is where this series and Yahoo Finance differ by design. Reconciled at 96.88% within 1% of Yahoo Finance adjusted close across 25,329 daily comparisons on 48 symbols; the two outliers, TRENT and ITC, are Yahoo-side (a bonus applied on the wrong date, a demerger scaled into history).

4. symbol_history/: which symbol an ISIN traded under, and when

One file per exchange. Each row is one contiguous interval.

Field	Type	Notes
exchange	string	NSE or BSE
isin	string	Stable instrument ID
symbol	string	Symbol active during the interval
valid_from, valid_to	date	First and last trading day
trading_days	int64	Days in the interval

5. metrics/: returns and rolling stats

One file per exchange per year, plus <ex>_latest.parquet with only the newest trading day for quick cross-sectional reads. Computed on adj_close; volume and turnover on raw values.

Field	Type	Notes
date, symbol, isin, adj_close		Carried for joining
ret_1d, ret_5d, ret_21d, ret_63d, ret_126d, ret_252d	float64	Simple returns over N trading days
ret_ytd	float64	Since the first trading day of the year
high_52w, low_52w	float64	Max and min adj_close over 252 trading days
pct_off_52w_high, pct_off_52w_low	float64	adj_close / high_52w - 1, adj_close / low_52w - 1
avg_vol_20d, avg_vol_60d	float64	Rolling mean of raw volume
avg_turnover_20d	float64	Rolling mean of raw turnover

Rolling fields are null until a full window of history exists. No partial windows.

6. universe/: point-in-time liquidity universe

One file per exchange: universe/<ex>_liquid.parquet. On the first trading day of each month, instruments are ranked by trailing 63-trading-day mean turnover and the top 500 are kept. Membership uses only data available on the rebalance date, so a name that later delisted still appears in the months it qualified for.

Field	Type	Notes
exchange	string	NSE or BSE
rebalance_date	date	First trading day of the month
valid_to	date	Day before the next rebalance
rank	int64	1 = highest trailing turnover
symbol, isin, name	string	As in force on rebalance_date
avg_turnover_63d	float64	Trailing mean turnover, INR

liquid100, liquid250 and liquid500 are rank <= N filters over the same rows.

Quickstart

Polars, full history of one symbol

import polars as pl

df = (
    pl.scan_parquet("hf://datasets/tejhq/indian-markets/nse/year=*/nse_*.parquet")
    .filter(pl.col("symbol") == "RELIANCE")
    .collect()
)

DuckDB, adjusted close joined to the universe

SELECT p.date, p.symbol, p.adj_close
FROM read_parquet('hf://datasets/tejhq/indian-markets/prices_adjusted/nse_2024.parquet') p
JOIN read_parquet('hf://datasets/tejhq/indian-markets/universe/nse_liquid.parquet') u
  ON u.isin = p.isin
 AND p.date BETWEEN u.rebalance_date AND u.valid_to
WHERE u.rank <= 100
ORDER BY p.date, p.symbol;

datasets

from datasets import load_dataset

nse = load_dataset("tejhq/indian-markets", "nse")
actions = load_dataset("tejhq/indian-markets", "actions")

REST

curl "https://api.tejhq.dev/v1/ohlcv/nse/RELIANCE?from=2024-01-01&to=2024-12-31"

No key needed for prices, snapshots, actions and the pipeline status at /v1/status. Adjusted prices and symbol history need a free key from tejhq.dev/keys. Metrics, universe, batch, resolve and the screener are on the Pro tier.

Caveats

• Same-day rows usually land between 19:00 and 22:00 IST. Fetching earlier returns the previous session. /v1/status carries trading_date, the honest answer to what is current.
• Bhavcopy occasionally carries anomalies (close outside [low, high], zero-volume rows). The pipeline drops those before publishing.
• Corporate action feeds are messy at the source. type = other keeps events the classifier could not place, with the original text in raw_subject.
• actions.isin is as reported and can be a stale post-merger ISIN. The adjusted tree resolves it; join on symbol_history if you need the same.

Source and license

• Source: NSE and BSE official EOD bhavcopy and corporate action feeds. Public, free, redistributable.
• Pipeline: github.com/tejhq/tej-bazaar, MIT. Every tree here is reproducible from it.
• Data: Exchange-published. Redistribution as cleaned parquet is permitted. Verify exchange terms before commercial use.

Part of TejHQ: open, free, end-of-day market data for India. Build log on the blog.