--- title: "Units and conventions" description: "The units and conventions every Fincanva API and MCP result follows: CAGR against arithmetic return, the signs of VaR and drawdown, weights and fractions." canonical_url: "https://fincanva.com/developers/concepts/units-and-conventions" last_updated: "2026-10-03" md_url: "https://fincanva.com/developers/concepts/units-and-conventions.md" --- # Units and conventions Every figure a Fincanva operation returns has a unit, and where the same word means different things in finance — an annual return can compound or average, a loss can be written as a positive or a negative number — it follows one of the conventions below. They are stated once, here, and each operation's reference links the convention every field follows. ## What does CAGR mean in a Fincanva result? A field labelled `cagr_geometric` is the compound annual growth rate: the constant yearly rate that takes the starting value to the [ending value](/glossary/final-value) over the period. It compounds, so it is never an average of yearly returns. Applies to: [backtest](/glossary/backtest) metrics, [per-strategy analytics](/glossary/strategy-analytics), [start-date sensitivity](/glossary/start-date-sensitivity). Not to be confused with: The arithmetic annualised return, which averages instead of compounding and is higher whenever returns vary. Label: `cagr_geometric`. The definition: [CAGR](/glossary/cagr). ## What is the arithmetic annualised return, and where does it appear? The arithmetic annualised return is the mean daily return multiplied by 252. Backtest trust, contribution analysis and the [stress test](/glossary/stress-test) report it, labelled `return_arithmetic_annualised`. Because it averages instead of compounding, it is higher than the CAGR of the same returns whenever they vary, so the two are never compared with each other. Applies to: backtest trust, contribution analysis, stress test. Not to be confused with: CAGR, which compounds. Label: `return_arithmetic_annualised`. ## How are volatility and returns annualised? Daily figures are annualised with 252 trading days a year — [volatility](/glossary/volatility) by multiplying the daily standard deviation by the square root of 252 — and monthly figures with 12 periods a year. A field that is annualised says so in its description; a field that does not is a value for its own period. Applies to: every operation that returns a volatility or a return. Label: `annualised_252`. The definition: [Annualization](/glossary/annualization). ## What sign do VaR, CVaR and drawdown carry? [Value at risk](/glossary/value-at-risk) and [conditional value at risk](/glossary/conditional-value-at-risk) are positive numbers that express a loss: a VaR of 0.021 is a loss of 2.1%. Drawdown is negative: a maximum drawdown of -0.34 is a fall of 34% from the previous peak. Compare their sizes, never their signs. Applies to: backtest metrics, contribution analysis, [instrument](/glossary/instrument) analytics. Label: `loss_positive_drawdown_negative`. The definition: [Drawdown](/glossary/max-drawdown). ## In what unit is a contribution to the Sharpe ratio? A contribution to the Sharpe ratio is in Sharpe points: the contributions of the parts add up to the whole's Sharpe ratio. A contribution of 0.12 is 0.12 of Sharpe ratio, not 12%. Applies to: contribution analysis. Label: `sharpe_points`. The definition: [Sharpe ratio](/glossary/sharpe-ratio). ## How are weights expressed and normalised? Weights are fractions of the capital they apply to. Where shorting is allowed they are signed — negative is short — and normalised by the sum of their absolute values, so the absolute weights add up to 1 before [leverage](/glossary/leverage); leverage then scales them as a multiple of capital. Applies to: allocation results, current holdings, allocation history. Label: `weights_gross_normalised`. ## Are percentages returned as fractions? Yes. Every return, weight, rate, drawdown and threshold the app shows as a percentage is returned as a fraction: 0.07 is 7%, -0.3 is -30% and 1.5 is 150%. [Screener](/glossary/screener) filter values follow the same rule — a percentage filter's threshold of 0.02 is the 2.0 % the screener shows — and each target's `unit` says `fraction`. Applies to: every operation. Label: `fraction`. ## How is a series labelled? Every series carries its frequency — daily, weekly, monthly or yearly — and says whether its values are annualised. Price levels are never returned as a series; returns, drawdowns, rolling statistics and indexed growth are. Applies to: every operation that returns a series. Label: `series_labelled`. ## How are dates written? Dates are ISO 8601 calendar dates, `YYYY-MM-DD`, naming the trading day. Timestamps are ISO 8601 in UTC. Applies to: every operation. Label: `iso_date`. ## What is the data version on a result? `dataVersion` names the nightly data build a result was computed from, so you can tell which data a figure reflects. A result that computes nothing from market data, such as a catalogue listing, carries null. Applies to: every operation. Label: `data_version`. ## In which currency are amounts of money? Amounts of money are in the base currency of your simulation settings, unless the field's unit is `usd`, which marks figures the screener reports in US dollars, such as market capitalisation. Applies to: every operation that returns an amount of money. Label: `base_currency`. The definition: [Base currency](/glossary/base-currency). ## How does a result tell you what it assumed? When a request leaves out a value the operation needs, the operation uses the default its reference documents and lists it in `assumed` — the field as a JSON pointer, the value used and the reason. Nothing is assumed silently. Applies to: every operation. Label: `assumed_stated`. ## What does the unit native mean? `native` means the value is in the unit of what it describes: the unit named by the `unit` field beside it, or the series a rule reads — index points for the VIX, a plain ratio for the VIX ratio. Applies to: catalogue operations, [risk condition](/glossary/risk-condition) presets. Label: `native_unit`. Fincanva is for education and illustration only. It is not personalised financial advice, and past or simulated results do not predict future ones. [Read the Terms Addendum](https://fincanva.com/terms/addendum#section-3)