Universe construction
What is universe construction?
Universe construction is the process of defining which instruments are eligible to be included in an analysis, model, index, or backtest and how that set is maintained over time. It turns broad market coverage into a concrete, rule-based list of symbols. Good universe construction makes assumptions explicit (venue, instrument type, liquidity constraints, currency, etc.) and makes results reproducible.
Why it matters
Nearly every metric you compute depends on the universe: returns, volatility, correlations, breadth, and even “market average” spread. A poorly defined universe can introduce hidden biases (survivorship, lookahead) or inconsistent coverage (symbols appear/disappear without explanation). In crypto, frequent listings, delistings, and contract changes make universe rules especially important.
What criteria are typically used for universe construction?
Typical criteria include venue, instrument type, quote currency, minimum history length, and liquidity/activity thresholds. Many workflows also require a minimum quality of data coverage (for example, continuous quotes or trades). The best criteria depend on the question you are answering and the frequency of your analysis.
How does universe construction affect backtest results?
It affects both what opportunities exist and what risks are measured. Excluding hard-to-trade or delisted instruments can artificially improve performance and reduce drawdowns. Conversely, including instruments without reliable coverage can add noise and make execution simulations unrealistic.
How do you make a universe definition reproducible?
Write the inclusion/exclusion rules as code, version them, and make them time-dependent when necessary. Store either membership intervals or periodic snapshots so you can reconstruct the universe as of any timestamp. Document any data-quality filters and how missing data is handled.
Real-world example
You want to measure average bid-ask spreads across “all USD spot pairs” on multiple exchanges. If you include illiquid pairs with only sporadic quotes, your average spread can explode and comparisons across exchanges become meaningless. A robust universe definition might require continuous quote coverage and a minimum number of trades per day, producing a more stable and interpretable dataset.