How to Backtest Non USD Spot FX With USD Account Balance?

[kbs-code]

I am attempting to back-test on FX spot currency pairs, and for a non-USD pair, such as EUR/CHF, it gives me this error:

2024-06-13T05:03:00.000000000Z [ERROR] BACKTESTER-001.Cache: Cannot calculate exchange rate: ValueError('Quote maps must not be empty')
2024-06-13T05:03:00.000000000Z [ERROR] BACKTESTER-001.Portfolio: Cannot calculate account state: insufficient data for CHF/USD

My account balance is USD.
What is the recommended way to deal with this? Can I side-load USD/CHF daily prices and specify that they should be used for currency conversions while the EUR/CHF prices are used for backtesting?

bts_fx.py

[faysou]

I've done a PR for supporting cross currency pairs, it will likely get merged at some point after the next release.

[kbs-code]

Ok, thank you.

[faysou]

Response to GitHub Discussion #3358

Solution: Backtesting Non-USD Spot FX with USD Account Balance

Good news! Recent changes have added support for cross-currency pairs in the portfolio. The system now has two mechanisms for handling currency conversions:

1. Graph-based DFS algorithm - Can find indirect conversion paths through intermediate currencies
2. Direct cross-currency conversion - Uses cross_notional_value method for currency pairs with explicit conversion prices

The exchange rate system can automatically find conversion paths like:

• Direct: CHF → USD (if CHF/USD is available)
• Indirect: CHF → EUR → USD (if EUR/CHF and EUR/USD are available)
• Cross-currency: Direct conversion using quote/base prices when provided

There are two recommended approaches to solve this:

Solution 1: Add Conversion Pair Quote Data (Recommended)

Add a currency pair instrument that provides a conversion path from CHF to USD. You have two options:

Option A: Direct path - Add CHF/USD:

from nautilus_trader.model.currencies import CHF, USD
from nautilus_trader.model.identifiers import InstrumentId, Symbol, Venue
from nautilus_trader.model.instruments import CurrencyPair
from nautilus_trader.model.objects import Price, Quantity

# Create CHF/USD conversion pair
venue = Venue("SIM")  # or your venue
chfusd = CurrencyPair(
    instrument_id=InstrumentId(Symbol("CHF/USD"), venue),
    raw_symbol=Symbol("CHF/USD"),
    base_currency=CHF,
    quote_currency=USD,
    price_precision=5,
    size_precision=2,
    price_increment=Price.from_str("0.00001"),
    size_increment=Quantity.from_str("0.01"),
    lot_size=None,
    max_quantity=None,
)

# Add the instrument to the engine
engine.add_instrument(chfusd)

# Add quote data for CHF/USD (can be daily bars or ticks)
chfusd_bars = load_chfusd_bars()  # Your data loading logic
engine.add_data(chfusd_bars)

Option B: Indirect path - Add EUR/USD (since you already have EUR/CHF):

from nautilus_trader.model.currencies import EUR, USD
from nautilus_trader.model.identifiers import InstrumentId, Symbol, Venue
from nautilus_trader.model.instruments import CurrencyPair
from nautilus_trader.model.objects import Price, Quantity

# Create EUR/USD conversion pair
venue = Venue("SIM")  # or your venue
eurusd = CurrencyPair(
    instrument_id=InstrumentId(Symbol("EUR/USD"), venue),
    raw_symbol=Symbol("EUR/USD"),
    base_currency=EUR,
    quote_currency=USD,
    price_precision=5,
    size_precision=2,
    price_increment=Price.from_str("0.00001"),
    size_increment=Quantity.from_str("0.01"),
    lot_size=None,
    max_quantity=None,
)

# Add the instrument to the engine
engine.add_instrument(eurusd)

# Add quote data for EUR/USD
eurusd_bars = load_eurusd_bars()  # Your data loading logic
engine.add_data(eurusd_bars)

The system will automatically find the conversion path: CHF → EUR → USD (via EUR/CHF and EUR/USD quotes).

Solution 2: Use Mark Exchange Rates

If you have static exchange rates or want to set them manually, you can use mark exchange rates:

from nautilus_trader.model.currencies import CHF, USD
from nautilus_trader.portfolio.config import PortfolioConfig

# Configure portfolio to use mark exchange rates
portfolio_config = PortfolioConfig(
    use_mark_xrates=True,
    convert_to_account_base_currency=True,
)

# After creating the engine, set the mark exchange rate
cache = engine.cache

# Set CHF/USD exchange rate (e.g., 1 CHF = 1.10 USD)
cache.set_mark_xrate(
    from_currency=CHF,
    to_currency=USD,
    xrate=1.10,  # Your CHF/USD rate
)

# The inverse rate (USD/CHF) is automatically set

Note: With mark exchange rates, you don't need to provide quote data for the conversion pair, but you'll need to update the rates manually if they change over time.

How It Works

Recent changes (commit b8b7f633d) added cross-currency pair support in the portfolio. The system now has two mechanisms:

1. Graph-based DFS algorithm (in xrate.rs) - Uses a depth-first search to find conversion paths between currencies. When you add a CurrencyPair instrument, it's automatically added to the exchange rate symbol table, and the system can:

   • Use direct quotes (e.g., CHF/USD)
   • Build indirect paths (e.g., CHF → EUR → USD if EUR/USD and EUR/CHF are available)

2. Direct cross-currency conversion (new in portfolio) - The portfolio now uses cross_notional_value method for currency pairs, which allows direct conversion using quote_price and base_price parameters. This means:

   • For EUR/CHF positions, you can convert to USD by providing CHF/USD and EUR/USD conversion prices
   • The portfolio's _calculate_currency_pair_exposure method automatically uses this when calculating exposure for currency pairs with a target currency
   • This avoids requiring the conversion pairs to be in the exchange rate graph

The key addition is the cross_notional_value method in the Position class, which enables more efficient cross-currency conversions for currency pair positions.

Recommendation

For backtesting, Solution 1 is generally preferred because:

• It uses realistic market data for conversions
• It automatically handles rate changes over time
• It's more accurate for historical analysis
• The system will automatically find the best conversion path

If you only have static rates or want to simplify, Solution 2 works well and requires less data preparation.

Both approaches will resolve the "Quote maps must not be empty" and "insufficient data for CHF/USD" errors you're encountering.

[faysou]

I was curious about your question so I asked cursor about it using my PR branch as context.