Data Sources

Learn how to configure and use data sources for market data in your trading strategies.

Overview

The Investing Algorithm Framework uses DataSource objects to define what market data your strategies need. Data sources are configured with symbols, timeframes, and warmup windows, and the framework automatically handles data retrieval from the appropriate data providers.

DataSource Configuration

Basic DataSource

A DataSource defines the market data requirements for your strategy:

from investing_algorithm_framework import DataSource

data_source = DataSource(
    identifier="btc_hourly",
    symbol="BTC/EUR",
    time_frame="1h",
    warmup_window=100,
    market="BITVAVO"
)

DataSource Parameters

Parameter	Type	Description
identifier	str	Unique identifier to reference this data source in your strategy.
symbol	str	Trading pair (e.g., "BTC/EUR", "ETH/USDT").
time_frame	str	Candlestick timeframe (e.g., "1h", "4h", "1d").
warmup_window	int	Number of historical data points needed for indicator warmup.
market	str	Exchange/market identifier (e.g., "BITVAVO", "BINANCE").
data_type	DataType	Type of data (default: DataType.OHLCV).
pandas	bool	Return data as pandas DataFrame (default: False).

⚠️ Deprecation Notice: The window_size parameter is deprecated and will be removed in release 0.8.0. Please use warmup_window instead.

Using Data Sources in Strategies

Single Data Source

from investing_algorithm_framework import TradingStrategy, TimeUnit, DataSource, PositionSize

class MyStrategy(TradingStrategy):
    time_unit = TimeUnit.HOUR
    interval = 1
    symbols = ["BTC"]
    trading_symbol = "EUR"

    data_sources = [
        DataSource(
            identifier="btc_1h",
            symbol="BTC/EUR",
            time_frame="1h",
            warmup_window=200,
            market="BITVAVO"
        )
    ]

    position_sizes = [
        PositionSize(symbol="BTC", percentage=0.5),
    ]

    def generate_buy_signals(self, data):
        df = data["btc_1h"]  # Access by identifier
        close = df["Close"]
        ma50 = close.rolling(50).mean()
        ma200 = close.rolling(200).mean()

        # Golden cross signal
        buy_signal = (ma50 > ma200) & (ma50.shift(1) <= ma200.shift(1))
        return {"BTC": buy_signal}

    def generate_sell_signals(self, data):
        df = data["btc_1h"]
        close = df["Close"]
        ma50 = close.rolling(50).mean()
        ma200 = close.rolling(200).mean()

        # Death cross signal
        sell_signal = (ma50 < ma200) & (ma50.shift(1) >= ma200.shift(1))
        return {"BTC": sell_signal}

Multiple Assets

For strategies using multiple assets, multiple timeframes, or mixing different markets, see Multiple Market Data Sources.

Supported Markets

The framework ships with built-in data providers for multiple markets. Set the market parameter on your DataSource to tell the framework where to fetch data from.

Choosing the Right Market

Use Case	Recommended Market	Why
Crypto trading (live)	CCXT exchanges (BINANCE, BITVAVO, etc.)	Direct exchange connection, supports order execution
US stocks / ETFs (research & backtesting)	YAHOO	Free, no API key, broad coverage
US stocks (production-grade)	POLYGON	High reliability, real-time data, official exchange feeds
Stocks + forex + crypto (lightweight)	ALPHA_VANTAGE	Simple API, good for prototyping
Local / offline data	CSV or Pandas provider	Full control, no network dependency

---

Yahoo Finance — market="YAHOO"

Best for: Stocks, ETFs, indices, forex, and crypto research and backtesting.

Free, no API key required. Powered by the yfinance library using data from Yahoo Finance.

Installation: pip install investing-algorithm-framework[yahoo]

DataSource(
    identifier="aapl_daily",
    market="YAHOO",
    symbol="AAPL",
    data_type="OHLCV",
    time_frame="1d",
    warmup_window=200,
)

Supported timeframes: 1m, 2m, 5m, 15m, 30m, 1h, 1d, 1W, 1M

Symbol format: Standard ticker symbols — AAPL, MSFT, GOOGL, BTC-USD, EURUSD=X

⚠️ Yahoo Finance is great for research and backtesting but is not recommended for production live trading due to rate limits and data delays.

---

Polygon.io — market="POLYGON"

Best for: Production-grade US stock, options, forex, and crypto data.

Requires an API key from Polygon.io. Powered by the polygon-api-client library.

Installation: pip install investing-algorithm-framework[polygon]

from investing_algorithm_framework import MarketCredential

app.add_market_credential(
    MarketCredential(market="POLYGON", api_key="your_polygon_api_key")
)

DataSource(
    identifier="aapl_daily",
    market="POLYGON",
    symbol="AAPL",
    data_type="OHLCV",
    time_frame="1d",
    warmup_window=200,
)

Supported timeframes: 1m, 5m, 15m, 30m, 1h, 1d, 1W, 1M

Symbol format: Standard ticker symbols — AAPL, MSFT, X:BTCUSD (crypto), C:EURUSD (forex)

You can also configure the API key via environment variable: export POLYGON_API_KEY=your_key

---

Alpha Vantage — market="ALPHA_VANTAGE"

Best for: Prototyping and lightweight stock/forex/crypto strategies.

Requires a free or premium API key from Alpha Vantage. Powered by the alpha_vantage Python library.

Installation: pip install investing-algorithm-framework[alpha_vantage]

from investing_algorithm_framework import MarketCredential

app.add_market_credential(
    MarketCredential(market="ALPHA_VANTAGE", api_key="your_av_api_key")
)

DataSource(
    identifier="aapl_daily",
    market="ALPHA_VANTAGE",
    symbol="AAPL",
    data_type="OHLCV",
    time_frame="1d",
    warmup_window=200,
)

Supported timeframes: 1m, 5m, 15m, 30m, 1h, 1d, 1W, 1M

Symbol format: Standard ticker symbols — AAPL, MSFT, IBM

⚠️ The free tier is limited to 25 API calls per day. For heavier usage, consider a premium plan.

---

CCXT Exchanges — Crypto

Best for: Live crypto trading and backtesting on specific exchanges.

The CCXT provider supports 100+ cryptocurrency exchanges. No additional API key is needed for public OHLCV data, but order execution requires exchange credentials.

DataSource(
    identifier="btc_hourly",
    market="BITVAVO",       # any CCXT-supported exchange
    symbol="BTC/EUR",
    data_type="OHLCV",
    time_frame="1h",
    warmup_window=200,
)

Popular exchanges:

Exchange	Market Identifier	Website
Binance	BINANCE	binance.com
Bitvavo	BITVAVO	bitvavo.com
Coinbase	COINBASE	coinbase.com
Kraken	KRAKEN	kraken.com
Bybit	BYBIT	bybit.com
OKX	OKX	okx.com
KuCoin	KUCOIN	kucoin.com
Bitfinex	BITFINEX	bitfinex.com

Symbol format: Trading pairs with slash — BTC/EUR, ETH/USDT, ADA/BTC

See the full list of supported exchanges in the CCXT documentation.

---

CSV and Pandas — Local Data

Best for: Offline backtesting, custom data pipelines, or data from unsupported sources.

CSVOHLCVDataProvider

from investing_algorithm_framework.infrastructure import CSVOHLCVDataProvider

csv_provider = CSVOHLCVDataProvider(
    csv_file_path="./data/btc_eur_1h.csv",
    symbol="BTC/EUR",
    time_frame="1h",
    market="bitvavo"
)

app.add_data_provider(csv_provider, priority=1)

CSV files must contain columns: Datetime, Open, High, Low, Close, Volume.

PandasOHLCVDataProvider

from investing_algorithm_framework.infrastructure import PandasOHLCVDataProvider
import pandas as pd

df = pd.read_csv("btc_data.csv", index_col=0, parse_dates=True)

pandas_provider = PandasOHLCVDataProvider(
    dataframe=df,
    symbol="BTC/EUR",
    time_frame="1h",
    market="bitvavo"
)

app.add_data_provider(pandas_provider, priority=1)

---

Data Provider Priority

When multiple data providers can serve the same DataSource, the framework uses priority to select the best one. Lower number = higher priority:

from investing_algorithm_framework import create_app
from investing_algorithm_framework.infrastructure import (
    CCXTOHLCVDataProvider,
    PandasOHLCVDataProvider
)

app = create_app()

# Priority 1 = highest priority (used first)
app.add_data_provider(pandas_provider, priority=1)

# Priority 3 = lower priority (fallback)
app.add_data_provider(CCXTOHLCVDataProvider(), priority=3)

To learn how to build your own data provider for a custom API, see Custom Data Providers.

Warmup Window

The warmup_window parameter is crucial for strategies using technical indicators.

Why Warmup Matters

Many indicators require historical data before producing valid values:

# If you use a 200-period moving average, you need at least 200 data points
# before the indicator produces valid values

data_sources = [
    DataSource(
        identifier="btc_data",
        symbol="BTC/EUR",
        time_frame="1h",
        warmup_window=250,  # 200 + buffer
        market="BITVAVO"
    )
]

Calculating Warmup Window

Indicator	Typical Period	Recommended Warmup
SMA/EMA (20)	20	30
SMA/EMA (50)	50	60
SMA/EMA (200)	200	220
RSI (14)	14	30
MACD (12,26,9)	26 + 9 = 35	50
Bollinger Bands (20)	20	30

# Example: Strategy using multiple indicators
# Longest indicator is 200-period MA, so warmup should be at least 200 + buffer
data_sources = [
    DataSource(
        identifier="btc_data",
        symbol="BTC/EUR",
        time_frame="4h",
        warmup_window=250,  # Covers 200-period MA + buffer
        market="BITVAVO"
    )
]

Data Types

OHLCV Data (Default)

from investing_algorithm_framework import DataSource, DataType

ohlcv_source = DataSource(
    identifier="btc_ohlcv",
    symbol="BTC/EUR",
    time_frame="1h",
    warmup_window=100,
    market="BITVAVO",
    data_type=DataType.OHLCV  # Default
)

Ticker Data

ticker_source = DataSource(
    identifier="btc_ticker",
    symbol="BTC/EUR",
    market="BITVAVO",
    data_type=DataType.TICKER
)

Next Steps

• Learn about Multiple Market Data Sources for multi-asset, multi-timeframe, and cross-market strategies
• Explore External Data to load CSV, JSON, or Parquet from URLs
• Check out Trading Strategies for strategy implementation