--- title: "What data your backtests run on" description: "What a backtest reads: one daily-refreshed catalogue of tradable instruments and reference-only series, keeping delisted names and dated index membership." canonical_url: "https://fincanva.com/docs/data-methodology/what-data-your-backtests-run-on" last_updated: "2026-08-30" md_url: "https://fincanva.com/docs/data-methodology/what-data-your-backtests-run-on.md" --- # What data your backtests run on Every backtest in Fincanva reads one shared market-data catalogue: daily price history for the [instruments](/docs/getting-started/instrument) a strategy can hold, plus reference-only series it can measure but never buy. The catalogue is maintained so that a run reads history **as it actually stood** — companies that later disappeared are still in it, and index membership is recorded per date rather than as it looks today. That property, not the size of the catalogue, is what makes a backtest result worth reading. ## What is in the catalogue a backtest reads Three things, and the difference between them decides what a strategy can do with each. {/* VISUAL: svg-diagram — the catalogue split three ways (tradable instruments · reference-only instruments · special data series) and which of them a strategy can hold — tracked in VISUAL_BACKLOG */} - **Tradable instruments** — the ones a strategy can actually hold, spanning several asset classes. Each carries a [permanent identifier](/docs/data-methodology/permanent-instrument-identifier) and a [coverage window](/docs/data-methodology/coverage-window): the first and last dates for which it has prices. - **Reference-only instruments** — indices and macro series that exist for measurement, not ownership. They carry the **Not tradable** badge in instrument search, and the pickers that decide what a strategy *holds* are restricted to tradable ones, so a reference series cannot become a position by accident. See [Tradable and Not tradable](/docs/data-methodology/tradable-and-not-tradable). - **Special data series** — market-wide references kept beside instrument prices, such as interest rates, inflation and valuation indicators. See [special data series](/docs/data-methodology/special-data-series). The catalogue is refreshed daily, and a backtest ends at the latest available market close rather than at today's calendar date — so the data, not the clock, closes a run. See [data freshness and frontier](/docs/backtesting/data-freshness-and-frontier). ## Why does the catalogue keep companies that no longer trade? Because dropping them would quietly rig every backtest in your favour. An instrument that stopped trading — acquired, taken private, or failed — stays in the catalogue with the history it had, and stays usable for the period it did trade. It carries the **Delisted** badge in instrument search. A catalogue that kept only today's survivors would let a strategy be tested exclusively on names that made it, which is [survivorship bias](/docs/getting-started/the-nine-biases-fincanva-helps-you-avoid#how-does-fincanva-handle-survivorship-bias) — the single most flattering error a backtest can contain. Keeping delisted names is what removes it. [Delisted](/docs/data-methodology/delisted) covers the term itself. ## Why an instrument keeps its history when its ticker changes Because the catalogue identifies an instrument by a [permanent identifier](/docs/data-methodology/permanent-instrument-identifier) rather than by its ticker. Tickers get reassigned and renamed; an identity that changed with them would split one company's price history into two unrelated fragments, or silently join two different companies under one symbol. Lists and index membership are recorded against that identifier too, so a member that renamed mid-membership stays one continuous entry. You never type this identifier, but you can see its absence: a strategy that references an instrument the library cannot resolve fails with "Referenced instruments do not exist in the platform data library". Add the instrument again from the search field in **Asset selection** so the strategy points at an entry that exists. ## What does the catalogue know about index membership? It knows who was in an index **on each past date**, not just who is in it now. Membership is stored per index as a joined/left history, so a run can read the index as it stood at the time rather than as it stands today — the same reason delisted names are kept, applied to lists instead of instruments. In the app this reaches you as one of the universe filters rather than as a data screen: alongside **Asset type**, **Country**, **Exchange**, **Sector** and **Currency**, the **Index** filter is hinted "Index membership". [Index lists and point-in-time constituents](/docs/data-methodology/index-lists-and-point-in-time-constituents) covers what that record contains. ## Limits and edge cases - **Nothing in the app shows you an instrument's coverage window.** There is no first-price-date column, and Fincanva does not move your simulation start year forward to match a short history — so a run can span two decades while one of its instruments only existed for the last three years. This matters enough to read before trusting a long-history number: [coverage window](/docs/data-methodology/coverage-window). - **Instruments from exchanges with different holidays share one date axis.** How they are placed onto it is covered in [market-day and trading-calendar alignment](/docs/data-methodology/market-day-and-trading-calendar-alignment). - **The catalogue records a required data level per instrument, but no plan-derived earliest-year floor is applied to a run today** — every account reaches the same history. See [data-tier gating](/docs/data-methodology/data-tier-gating). - **[SymbolsLists](/docs/data-methodology/symbolslist) ship with the market data and are not yours to create or edit**, and unlike index lists they record membership as it stands now, with no joined/left history. ## Related To choose which of these instruments a strategy holds, see [choosing the instruments your strategy holds](/docs/strategies/choosing-the-instruments-your-strategy-holds); for what a run does with them, see [how backtesting works](/docs/backtesting/how-backtesting-works). A catalogue free of survivorship bias makes a backtest honest, not predictive — Fincanva does not tell you what to hold, see [Is this financial advice?](/docs/investing-theory/is-this-financial-advice#is-this-financial-advice).