Reading Market Data
CSV, yfinance & NSEpy

Every algo trading system starts with data. In India, you'll work with three main free sources before connecting to a paid broker API: CSV files (from NSE website), yfinance (Yahoo Finance Python library), and NSEpy (NSE-specific historical data).

Loading Data from CSV

NSE's daily Bhavcopy CSV format is the most reliable offline data source. Here's how to load, clean, and use it with Pandas.

# Sample NSE Bhavcopy CSV structure:
# SYMBOL,SERIES,OPEN,HIGH,LOW,CLOSE,TOTTRDQTY,TIMESTAMP
# NIFTY 50,EQ,21450,21620,21390,21580,0,15-JAN-2024

import pandas as pd

# --- Load from CSV ---
df = pd.read_csv(
    "nifty_daily.csv",
    parse_dates=["TIMESTAMP"],    # auto-parse date column
    dayfirst=True                   # DD-MMM-YYYY format
)

# Rename columns to lowercase standard
df.rename(columns={
    "OPEN":       "open",
    "HIGH":       "high",
    "LOW":        "low",
    "CLOSE":      "close",
    "TOTTRDQTY":  "volume",
    "TIMESTAMP":  "date"
}, inplace=True)

# Set date as index and sort
df = df.set_index("date").sort_index()

# Keep only OHLCV columns
df = df[["open", "high", "low", "close", "volume"]]

print(df.head())
print(f"\nRows: {len(df)}, Range: {df.index[0].date()} → {df.index[-1].date()}")

Fetching Data with yfinance

yfinance is the quickest way to get historical data in Python. For NSE stocks, append .NS to the ticker. For indices, use ^NSEI (NIFTY 50) or ^NSEBANK (BANKNIFTY).

import yfinance as yf
import pandas as pd

# ── NIFTY 50 Index (1 year of daily data) ──
nifty = yf.download(
    tickers  = "^NSEI",
    start    = "2024-01-01",
    end      = "2024-12-31",
    interval = "1d",        # daily candles
    progress = False
)

nifty.columns = [c.lower() for c in nifty.columns]
print(nifty.tail(5))

Common yfinance Tickers for Indian Markets

# ── Intraday 5-minute data (last 60 days max) ──
nifty_5m = yf.download(
    tickers  = "^NSEI",
    period   = "5d",          # last 5 trading days
    interval = "5m",          # 5-minute candles
    progress = False
)
print(f"5-min candles loaded: {len(nifty_5m)}")
print(nifty_5m.tail(3))

# ── Multiple tickers at once ──
watchlist = yf.download(
    tickers  = "RELIANCE.NS TCS.NS HDFCBANK.NS INFY.NS",
    start    = "2024-01-01",
    end      = "2024-06-30",
    interval = "1d",
    progress = False
)
# Access individual stock close prices
reliance_close = watchlist["Close"]["RELIANCE.NS"]
print(reliance_close.tail(3))

Using NSEpy for NSE Data

NSEpy fetches data directly from NSE India's website. It's useful for getting stock-specific historical data and F&O contract prices.

from nsepy import get_history
from datetime import date
import pandas as pd

# ── Equity historical data ──
reliance = get_history(
    symbol    = "RELIANCE",
    start     = date(2024, 1, 1),
    end       = date(2024, 6, 30),
    series    = "EQ"
)

# Rename to standard OHLCV columns
reliance = reliance[["Open", "High", "Low", "Close", "Volume"]]
reliance.columns = ["open", "high", "low", "close", "volume"]
print(reliance.tail(5))

# ── NIFTY Futures historical data ──
nifty_fut = get_history(
    symbol      = "NIFTY",
    start       = date(2024, 1, 1),
    end         = date(2024, 1, 31),
    index       = True,
    futures     = True,
    expiry_date = date(2024, 1, 25)   # last Thursday of month
)
print(nifty_fut[["Open","High","Low","Close","Volume"]].head())

Data Cleaning — The Essential Step

Raw market data always has issues: missing rows, zero prices, duplicate timestamps, wrong dtypes. Always clean before analysis.

import pandas as pd

def clean_ohlcv(df: pd.DataFrame) -> pd.DataFrame:
    """Standard OHLCV cleaning pipeline for any data source."""

    # 1. Remove duplicate index timestamps
    df = df[~df.index.duplicated(keep="last")]

    # 2. Sort chronologically
    df = df.sort_index()

    # 3. Remove rows with zero or negative prices
    price_cols = ["open", "high", "low", "close"]
    df = df[(df[price_cols] > 0).all(axis=1)]

    # 4. OHLC integrity check (high must be highest)
    valid = (
        (df["high"] >= df["open"]) &
        (df["high"] >= df["close"]) &
        (df["low"]  <= df["open"]) &
        (df["low"]  <= df["close"])
    )
    bad_rows = (~valid).sum()
    if bad_rows > 0:
        print(f"⚠️  Removed {bad_rows} candles failing OHLC check")
    df = df[valid]

    # 5. Ensure numeric types
    for col in ["open","high","low","close","volume"]:
        df[col] = pd.to_numeric(df[col], errors="coerce")

    # 6. Drop any remaining NaN rows
    before = len(df)
    df = df.dropna(subset=price_cols)
    dropped = before - len(df)
    if dropped:
        print(f"⚠️  Dropped {dropped} rows with NaN prices")

    print(f"✅ Clean rows: {len(df)}")
    return df

# Use it on any data source
df_clean = clean_ohlcv(nifty)
print(df_clean.dtypes)

Saving Cleaned Data to CSV

# Save to CSV for offline use (no internet needed next run)
df_clean.to_csv("nifty_clean.csv")

# Reload later
df_reload = pd.read_csv("nifty_clean.csv", index_col=0, parse_dates=True)
print(f"Reloaded {len(df_reload)} rows")

Full Data Pipeline — Fetch, Clean & Prepare

import yfinance as yf
import pandas as pd
from pathlib import Path

def fetch_nse_data(symbol: str, period: str = "1y", interval: str = "1d") -> pd.DataFrame:
    """Fetch, clean and return OHLCV data for an NSE symbol."""

    # Map symbol to yfinance ticker
    ticker_map = {
        "NIFTY":       "^NSEI",
        "BANKNIFTY":   "^NSEBANK",
        "RELIANCE":    "RELIANCE.NS",
        "TCS":         "TCS.NS",
        "HDFCBANK":    "HDFCBANK.NS",
        "INFY":        "INFY.NS",
    }
    ticker = ticker_map.get(symbol, symbol + ".NS")

    # Check cache first
    cache_path = Path(f"data/{symbol}_{interval}.csv")
    if cache_path.exists():
        print(f"📂 Loading {symbol} from cache...")
        return pd.read_csv(cache_path, index_col=0, parse_dates=True)

    # Fetch from yfinance
    print(f"📡 Fetching {symbol} ({ticker}) from yfinance...")
    raw = yf.download(ticker, period=period, interval=interval, progress=False)
    raw.columns = [c.lower() for c in raw.columns]

    # Clean
    raw = raw[["open","high","low","close","volume"]].dropna()

    # Save to cache
    cache_path.parent.mkdir(exist_ok=True)
    raw.to_csv(cache_path)
    print(f"✅ Saved to cache: {cache_path}")
    return raw

# Fetch NIFTY and BANKNIFTY daily data
nifty_df    = fetch_nse_data("NIFTY")
banknifty_df = fetch_nse_data("BANKNIFTY")

print(f"\nNIFTY    : {len(nifty_df)} days")
print(f"BANKNIFTY: {len(banknifty_df)} days")

Quiz

Exercises

Lesson Summary