Universes

A Universe identifies the assets and market context evaluated by a Study. Keeping this definition with the study makes it clear which market produced a result and allows the same strategy to be tested against different asset sets without mixing their evidence.

Each study targets exactly one universe. A universe can describe one symbol, a fixed basket, or the market context used by a strategy that selects symbols dynamically.

Define a universe

from investing_algorithm_framework import Universe

crypto_majors = Universe(
    key="crypto_majors_eur",
    symbols=["BTC/EUR", "ETH/EUR"],
    trading_symbol="EUR",
    market="BITVAVO",
    metadata={"selection": "largest EUR pairs"},
)

Field	Purpose
key	Stable identifier used in persisted results and indexes.
symbols	Assets included in the evaluation.
trading_symbol	Quote or settlement currency, such as EUR.
market	Exchange, broker, or venue identifier.
metadata	Optional provenance or selection details.

When key is omitted, the framework derives one from the universe definition. Use an explicit key when the selection has a durable business meaning or when you want a stable label in reports.

Use a universe in a study

from investing_algorithm_framework import Study, StudySampleType

study = Study(
    name="momentum_majors",
    universe=crypto_majors,
    backtest_windows=windows,
    sample_type=StudySampleType.IN_SAMPLE,
)

The universe is persisted with the study in the Open Backtest Format, so reports and external readers can identify the assets and market behind every result.

Universe out-of-sample testing

Use separate studies to test whether a strategy generalizes to unseen assets:

development_study = Study(
    name="momentum_development",
    universe=Universe(
        key="development_assets",
        symbols=["BTC/EUR", "ETH/EUR"],
        trading_symbol="EUR",
        market="BITVAVO",
    ),
    backtest_windows=windows,
    sample_type=StudySampleType.IN_SAMPLE,
)

held_out_study = Study(
    name="momentum_held_out",
    universe=Universe(
        key="held_out_assets",
        symbols=["SOL/EUR", "ADA/EUR"],
        trading_symbol="EUR",
        market="BITVAVO",
    ),
    backtest_windows=windows,
    sample_type=StudySampleType.OUT_SAMPLE_UNIVERSE,
)

Keeping these as separate studies prevents in-sample and held-out evidence from being pooled accidentally.

Related guides

• Studies combines a universe, windows, engines, and assumptions.
• Backtest Windows defines the time periods to evaluate.
• Cross-Sectional Pipelines dynamically ranks and filters symbols within a strategy.