HuggingFace Space: NSE India Security Master — Interactive Exploration Guide
#2
by tickertruth - opened
HuggingFace Space: NSE India Security Master — Interactive Exploration Guide
This guide shows you how to use the TickerTruth NSE India Security Master dataset via the HuggingFace Space. Learn to query real corporate actions, explore symbol history, and apply results to your backtests and analyses.
What You Can Explore
The Space gives you interactive access to four core datasets:
- Security Master — Current and historical NSE equity symbols with metadata
- Corporate Actions — Dividends, splits, bonuses, mergers, delistings with effective dates
- Symbol Lineage — Ticker renames, mergers, status changes over time
- Adjustment Factors — Pre-calculated ratios for price normalization
All data is versioned monthly. You can query specific dates or ranges, download results, and export to CSV/Parquet.
Use Case 1: Fix a Broken Backtest with Split Adjustments
Problem: Your backtest shows an impossible price jump on a specific date. This usually means a stock split wasn't adjusted.
Query:
SELECT
symbol,
ex_date,
action_type,
value_ratio,
adjustment_factor
FROM fact_corporate_action_event
WHERE symbol = 'RELIANCE'
AND action_type IN ('SPLIT', 'BONUS')
AND ex_date BETWEEN '2020-01-01' AND '2026-06-07'
ORDER BY ex_date DESC;
Example Result:
symbol | ex_date | action_type | value_ratio | adjustment_factor
---------|------------|-------------|-------------|-------------------
RELIANCE | 2023-05-31 | SPLIT | 1:5 | 0.2
RELIANCE | 2020-11-17 | BONUS | 1:2 | 0.5
How to Use It:
- Find the split on 2023-05-31 with ratio 1:5 (means 1 new share for every 5 old shares)
- Adjustment factor 0.2 = multiply all prices BEFORE 2023-05-31 by 0.2 to make them comparable to post-split prices
- Example: If RELIANCE closed at ₹2000 on 2023-05-30, the adjusted price is
2000 × 0.2 = ₹400(comparable to post-split prices)
Apply to Backtest:
# Pseudo-code
historical_prices = load_prices(symbol='RELIANCE', start='2020-01-01', end='2026-06-07')
splits = query_splits(symbol='RELIANCE')
for split in splits:
ex_date = split['ex_date']
factor = split['adjustment_factor']
# Adjust all prices BEFORE ex_date
historical_prices[historical_prices.index < ex_date] *= factor
# Now your price series is continuous across splits
Use Case 2: Reconcile a Portfolio NAV After Dividends
Problem: Your portfolio manager says the NAV dropped ₹50/share on a specific date, but prices didn't fall that much. What happened?
Query:
SELECT
symbol,
ex_date,
dividend_amount,
COALESCE(frequency, 'INTERIM') as dividend_type
FROM fact_corporate_action_event
WHERE symbol IN ('INFY', 'TCS', 'WIPRO')
AND action_type = 'DIVIDEND'
AND ex_date = '2026-05-15'
ORDER BY symbol;
Example Result:
symbol | ex_date | dividend_amount | dividend_type
-------|------------|-----------------|---------------
INFY | 2026-05-15 | 21.0 | FINAL
TCS | 2026-05-15 | 13.5 | INTERIM
WIPRO | 2026-05-15 | 6.0 | FINAL
How to Use It:
- INFY went ex-dividend on 2026-05-15 paying ₹21/share
- If a fund held 1000 INFY shares, it lost ₹21k in NAV due to the dividend (shareholders get cash, stock price drops by ~dividend amount)
- The portfolio manager would see a ₹21/share drop in INFY on that date
- This is expected — it's the dividend payout, not a price decline
Apply to Portfolio Analytics:
# Pseudo-code
portfolio = {'INFY': 1000_shares, 'TCS': 500_shares, 'WIPRO': 250_shares}
ex_dates = query_dividends(symbols=list(portfolio.keys()), date='2026-05-15')
dividend_impact = {}
for dividend in ex_dates:
symbol = dividend['symbol']
amount = dividend['dividend_amount']
shares = portfolio[symbol]
dividend_impact[symbol] = amount * shares
total_dividend = sum(dividend_impact.values())
# Portfolio NAV will drop by total_dividend, but it's cash outflow to shareholders (expected)
Use Case 3: Track Symbol Renames for Data Pipeline Updates
Problem: Your production data pipeline has hardcoded symbol lists. When NSE renames a ticker, your pipeline breaks because it can't find price data under the old symbol.
Query:
SELECT
old_symbol,
new_symbol,
effective_date,
reason
FROM lineage_symbol_event
WHERE event_type = 'RENAME'
AND effective_date >= DATE_TRUNC('month', CURRENT_DATE) - INTERVAL '24 months'
ORDER BY effective_date DESC;
Example Result:
old_symbol | new_symbol | effective_date | reason
-----------|------------|----------------|-------------------
POWERGRID | POWERGRDL | 2025-12-30 | Name change by issuer
AEGISLOG | AEGIS | 2025-10-15 | Name change by issuer
ORIENTCEM | ORIENTCEM | 2025-03-20 | Symbol optimization
How to Use It:
- POWERGRID became POWERGRDL on 2025-12-30
- Historical price data exists under POWERGRID until that date, then under POWERGRDL
- Your pipeline should map both symbols to the same company for continuous backtesting
Apply to Data Pipeline:
# Pseudo-code
symbol_map = build_symbol_map() # {old_symbol: new_symbol, ...}
def get_price_series(security_id, start_date, end_date):
# Get all symbol aliases for this security across the date range
aliases = query_symbol_lineage(security_id=security_id, date_range=(start_date, end_date))
prices = []
for alias in aliases:
symbol = alias['symbol']
valid_from = alias['valid_from']
valid_to = alias['valid_to']
# Fetch prices for this symbol during its valid period
data = fetch_prices(symbol=symbol, start=valid_from, end=valid_to)
prices.append(data)
# Concatenate and sort by date
return pd.concat(prices).sort_index()
Use Case 4: Understand Delisting Events and Merger Chains
Problem: A stock was delisted 5 years ago. Was it acquired? Renamed? Defaulted? You need the full history for research.
Query:
SELECT
symbol,
event_type,
effective_date,
old_symbol,
new_symbol,
reason
FROM lineage_symbol_event
WHERE symbol IN ('ZENITHEQUIP', 'ADANIPORTS') -- one active, one delisted/merged
ORDER BY effective_date DESC;
Example Result:
symbol | event_type | effective_date | old_symbol | new_symbol | reason
-------------|------------|----------------|---------------|------------|--------------------
ZENITHEQUIP | DELISTING | 2022-06-30 | ZENITHEQUIP | NULL | Merger with ICIL
ZENITHEQUIP | MERGER | 2022-06-30 | ZENITHEQUIP | ICIL | Reverse merger
ADANIPORTS | ACTIVE | NULL | NULL | NULL | Currently trading
How to Use It:
- ZENITHEQUIP was delisted on 2022-06-30 due to a merger with ICIL
- Any backtest using ZENITHEQUIP data after this date is invalid — the company no longer trades
- If you want to include merged companies, you need to track ICIL (the surviving company) from that point forward
- For research, this explains why ZENITHEQUIP price data stops abruptly
Apply to Research & Backtesting:
# Pseudo-code
def get_tradeable_symbols(as_of_date):
"""Get all symbols that were actively trading on this date"""
return query_lineage(
as_of_date=as_of_date,
status='ACTIVE' # excludes delisted, suspended, etc.
)
def backtest_valid_period(symbol):
"""Returns (start_date, end_date) when symbol was tradeable"""
history = query_symbol_lineage(symbol=symbol)
start = history[0]['listing_date']
end = history[-1]['delisting_date'] or CURRENT_DATE
return (start, end)
# ZENITHEQUIP valid period: [inception, 2022-06-30)
# ADANIPORTS valid period: [inception, current] (still trading)
Use Case 5: Validate Price Gaps Against Announced Events
Problem: You see a 15% price drop on a date. Is it a market crash or a dividend payment? Or a split adjustment?
Query:
SELECT
ca.symbol,
ca.ex_date,
ca.action_type,
ca.value_ratio,
bp.prev_close,
bp.close,
((bp.close - bp.prev_close) / bp.prev_close * 100) as price_change_pct
FROM fact_corporate_action_event ca
LEFT JOIN fact_equity_eod bp ON ca.symbol = bp.symbol AND ca.ex_date = bp.date
WHERE ca.symbol = 'HCLTECH'
AND ca.ex_date BETWEEN '2024-01-01' AND '2026-06-07'
ORDER BY ca.ex_date DESC;
Example Result:
symbol | ex_date | action_type | value_ratio | prev_close | close | price_change_pct
----------|------------|-------------|-------------|------------|-------|------------------
HCLTECH | 2026-03-20 | DIVIDEND | 15.0 | 1520.50 | 1505.15 | -1.0%
HCLTECH | 2025-06-15 | SPLIT | 1:5 | 1850.00 | 370.00 | -80.0%
HCLTECH | 2024-08-10 | BONUS | 1:1 | 1650.00 | 825.00 | -50.0%
How to Use It:
- 2025-06-15: 1:5 split → stock goes from ₹1850 to ₹370 (~80% drop). This is expected, not a crash.
- 2024-08-10: 1:1 bonus → stock goes from ₹1650 to ₹825 (~50% drop). This is expected, each share splits into 2.
- 2026-03-20: ₹15 dividend → stock drops ~1% (₹15 ÷ ₹1520 ≈ 1%). This is expected.
How to Interpret:
- If the price drop percentage matches the expected adjustment ratio, the event is priced in
- If the drop is much larger, there's additional market news (earnings miss, etc.)
- This helps you distinguish between company-driven events (corporate actions) and market-driven events (sentiment, macros)
Use Case 6: Calculate Correct Adjustment Factors for Custom Analysis
Problem: You're building a custom price normalization and need to understand how to apply cascading adjustments (splits on top of bonuses, etc.).
Query:
SELECT
symbol,
ex_date,
action_type,
value_ratio,
adjustment_factor,
cumulative_adjustment_factor
FROM fact_corporate_action_event
WHERE symbol = 'TATASTEEL'
ORDER BY ex_date ASC;
Example Result:
symbol | ex_date | action_type | value_ratio | adjustment_factor | cumulative_adjustment_factor
----------|------------|-------------|-------------|-------------------|-------------------------------
TATASTEEL | 2018-02-01 | BONUS | 1:5 | 0.1667 | 0.1667
TATASTEEL | 2020-05-20 | SPLIT | 1:2 | 0.5 | 0.0834
TATASTEEL | 2023-11-10 | DIVIDEND | 15.0 | 1.0 (no split) | 0.0834
How to Use It:
- Cumulative adjustment tells you how many new shares equal 1 original share
- TATASTEEL 2018 holder: 1 original share → 5 bonus shares (1:5 bonus) → 2.5 post-split shares → 2.5 current shares
- Cumulative factor = 0.0834 means 1 original share = 0.0834 current shares (or 1 current share = 12x original share)
- For a price from 2017, multiply by 0.0834 to make it comparable to 2024 prices
Apply to Custom Normalization:
# Pseudo-code
def normalize_historical_price(symbol, price, price_date):
"""Convert a historical price to current-day comparable"""
cumulative_factor = query_cumulative_adjustment(
symbol=symbol,
as_of_date=price_date
)
return price * cumulative_factor
# Example: TATASTEEL on 2017 at ₹100
# normalized = 100 * 0.0834 = ₹8.34 (comparable to 2024 prices)
General Tips for Using the Space
1. Always Filter by Date Range
Corporate actions span decades. Without a date filter, queries are slow and results are overwhelming. Use WHERE ex_date BETWEEN '2020-01-01' AND CURRENT_DATE to stay focused.
2. Use Symbol or ISIN for Precise Queries
- Query by symbol for current events
- Query by ISIN for historical events (symbols change, ISINs don't)
3. Understand Ex-Date vs. Payment Date
- Ex-Date: First day you DON'T get the benefit (price drops ~dividend amount on this date)
- Record Date: Technical cutoff (you must own by this date to get the benefit)
- Payment Date: When you receive cash/shares
- For backtesting, use ex-date; for corporate records, use record-date
4. Validate Against Price Movements
Always cross-check corporate actions against actual bhavcopy price data. If a ₹20 dividend causes a 0.5% price drop instead of 1.5%, something else moved the market that day.
5. Export and Transform Locally
The Space is for exploration. For production, export results to CSV/Parquet and load into your backtest engine locally. Keep your own versioned copy of the mappings you use.
What's NOT in This Dataset
- Real-time feeds — monthly snapshots only
- BSE data — NSE equities only (Phase 1)
- Intraday data — daily OHLCV only
- Forecasts — historical truth layer only
- Fundamental data — no earnings, PE, etc.
Next Steps
- Explore the Space: Run queries for your favorite stocks
- Test on Your Data: Match our corporate action dates against your price data
- Integrate: Use the mappings in your backtest pipeline
- Feedback: Report discrepancies or missing events in the discussion thread
Questions?
Ask in the HuggingFace Discussion or check the methodology docs for deeper technical details.