THE SHORT ANSWER
A reproducible gold price starts with the observation, not the formatted number
GoldObserve validates a real provider observation, preserves its source and times, normalizes it to a fine-troy-ounce basis, applies currency, mass and purity transformations in a fixed order, and rounds only at the presentation boundary. Historical calculations use returned observations rather than invented points. Failed sources remain unavailable, and revised macro series may change research results because the current platform uses the latest available vintage.
Exactly one troy ounce for precious-metal conversion.
Gross mass becomes fine-metal mass before valuation.
Nearest date only where the research method documents it.
Unavailable stays unavailable or becomes a labelled manual input.
LIVE CALCULATION AUDIT
Trace one gold value from source observation to final result
The worksheet requests the same public gold endpoint used by GoldObserve tools, exposes source and timing metadata, and then shows every mass and purity transformation. Change the assumptions, copy a share link, or download a machine-readable audit record. If the source fails, the tool does not insert a remembered or estimated price.
Price reference
Shows the source and observation time of the metal price used here. Product premiums, dealer quotes and taxes are separate.
- Source
- No provider
- Observed
- Observation time unavailable
- Age
- Checking age…
- Delivery
- No data layer available
Checking the current market reference.
Enter a positive price and gross weight, with fineness from 0.1 to 1000.
This is a transformation audit, not a dealer quote. It excludes premium, bid-ask spread, tax, making charge, assay, shipping and storage. A local-currency result may include a provider-supplied or separate reference FX conversion; it is not an executable foreign-exchange rate.
THE REQUIRED RECORD
A price is comparable only after seven questions have answers
01What instrument? Indicative XAU spot, an LBMA auction benchmark, a futures contract, a coin and a dealer bid are different objects.
02Which field? Last, reference, bid, ask, high and low cannot be substituted without changing the claim.
03Which currency and unit? USD per fine troy ounce is not a local-currency gram quote until both transformations are complete.
04Which provider? The adapter that succeeded is retained; the first provider in the configured order is not assumed to have answered.
05When observed? This is the source's measurement or publication time.
06When received? This separates market age from network and cache timing.
07What state? Fresh, provider-stale, independently aged, invalid and unavailable require different displays.
Provider identities, current endpoint status, cache windows, terms and historical coverage are maintained separately in the data-source register. This page documents what happens after a source is selected.
VALIDATION PIPELINE
Six gates stand between an upstream response and a displayed value
Request a supported instrumentMetal, currency and unit must match the public route contract.
Require a successful responseTransport and upstream HTTP failures advance to the next eligible provider.
Reject malformed valuesThe price must be numeric, finite and positive; missing optional fields remain null.
Preserve provenanceProvider, observed time, receipt time and source stale state travel with the quote.
Transform in a fixed orderFine-ounce basis, currency, weight unit, fineness and scenario costs stay conceptually separate.
Fail closedAfter all eligible providers fail, the route returns unavailable with no fabricated replacement price.
UNIT, CURRENCY AND PURITY
The normalization order is part of the result
Currency timestamps can differ from the metal timestamp. A converted CNY price is not the Shanghai Gold Benchmark Price, and a converted INR price is not a tax-inclusive jewellery quote. Denomination changes do not transform an indicative reference into a local benchmark or executable offer.
TIME-SERIES CONSTRUCTION
Display sampling is not data creation
A daily history response is validated point by point: the date must parse and the closing value must be positive and finite. For a dense screen chart, GoldObserve may retain real observations at a regular stride plus the final observation. It does not interpolate a smooth path, copy the last close into missing days or invent a weekend price.
Series identity is preserved before any calculation. The default daily fallback is the independent NBU route, which stores gold, silver, platinum and palladium accounting values in UAH per troy ounce plus official UAH per USD for the same effective date, then derives USD per ounce by division when read. Page and CSV requests query only the reviewed local database; the offline updater checks the official API and requires an exact candidate hash before writing. The World Bank route stores monthly-average USD gold and silver. GoldObserve does not splice these observation definitions into one artificial path.
The order of operations matters: align valid levels first, then calculate changes for the matched sample. Calculating two independent return series and pairing them by array position could compare different calendar periods.
RETURNS AND PATH RISK
Each statistic answers a narrower question than its label suggests
These figures use the returned market-price path. A personal result can differ because purchase premium, sale discount, fees, tax, storage, cash flows, product purity and transaction dates belong in the position-level calculation.
TECHNICAL INDICATORS
Indicators describe the returned history; they do not predict the next price
Equal weight; blank until the full window exists.
α = 2 ÷ (n + 1), seeded with the first n-period SMA.
Initial averages use 14 changes, followed by Wilder smoothing.
Calculated from the complete trailing return window.
Warm-up periods remain empty instead of being backfilled with partial-window estimates. A moving-average crossover, an RSI reading above 70 or below 30, rising volatility and a prior drawdown are historical descriptions. None guarantees direction, timing or magnitude of a future gold move. Indicators are not investment advice or price forecasts. The technical-analysis workspace exposes these series and their downloadable observations.
MACRO DATA AND REVISIONS
The latest published value is not always the value first reported
GoldObserve macro routes currently use Federal Reserve Economic Data for the 10-year TIPS real yield, broad U.S. dollar index, CPI, M2 money stock, Federal Reserve balance sheet and WTI oil. CPI year-over-year is transformed as (current CPI ÷ CPI twelve months earlier − 1) × 100. Bitcoin comparison data comes from CoinGecko and is not treated as a FRED series.
FRED shows the latest available vintage. Source agencies may revise observations, series definitions and seasonal adjustments. FRED documents those revisions, while ALFRED preserves what was known at earlier real-time dates. GoldObserve uses the current FRED response and does not currently preserve an ALFRED vintage with every saved calculation. A later rerun can therefore differ even when the requested date range is unchanged.
A serious event-study or backtest must define whether it uses latest-vintage data or only values known at the historical decision date. GoldObserve labels current correlation work as ex_post_latest. A future point_in_time_vintage dataset may be generated only by a controlled offline ALFRED job using a secret API key; neither PHP nor browser code may carry that key or silently backfill today's revised values into an earlier information set.
ROUNDING, EXPORTS AND REPRODUCTION
Round presentation last and preserve the assumptions beside the answer
Headline currency values are formatted for reading, but the calculation chain keeps additional precision. Repeating a formula from a screenshot can produce a small difference if the screenshot contains rounded intermediate values. The live audit above exposes extended decimals and downloads the exact numeric inputs, constant, formulas, outputs, provider state and generation time as JSON.
01Save the observation. Record provider, observed time, receipt time, currency, unit and state.
02Save every user input. Weight, fineness, cost and scenario assumptions must not be hidden in interface state.
03Name the constants. In this chain, the material constant is 31.1034768 grams per troy ounce.
04Separate raw from formatted. JSON outputs keep numbers; screen and CSV formatting can round for usability.
05Expect live results to move. A later quote, FX observation or revised macro vintage is a different input record.
RETAIL AND SETTLEMENT BOUNDARY
Theoretical metal value is not a complete transaction price
A physical product can add fabrication, brand, packaging, scarcity, dealer margin, payment, shipping, insurance and tax. A resale can subtract spread, assay, damage, minimum fees and settlement delay. Spot settlement conventions also do not determine when a particular retailer transfers title or cash.
LBMA publishes distinct auction benchmark and value-date material, and use of benchmark data can require permission or a license. GoldObserve does not relabel its indicative live feed as LBMA Gold Price. For accounting, contracts or settlement, use the benchmark and license that the document actually specifies; for a purchase or sale, use the counterparty's same-time executable quote.
PRIMARY REFERENCES
Constants, revision behavior and benchmark boundaries
- NIST precious-metals conversion information for the exact 31.1034768-gram troy ounce used in this methodology.
- FRED database update and validation documentation for source checks, latest-vintage display and ALFRED archiving.
- FRED real-time-period documentation for revisions to observations, series and sources.
- ALFRED API documentation for retrieving data as it was known at earlier real-time dates.
- National Bank of Ukraine Developer API and open-data terms for the raw date-matched accounting-price and FX inputs in the independent daily reference library.
- World Bank Commodity Markets and its CC BY 4.0 dataset record for the self-owned monthly-average library.
- LBMA auction-price documentation for the separately timed benchmark and licensing boundary.
- LBMA value-date guidance for the general two-London-business-day spot convention and institution-specific exceptions.
RELATED WORKSPACES
Follow the methodology into a working product
FAQ
Gold price methodology questions
How does GoldObserve calculate gold price per gram?
It divides the selected-currency price per fine troy ounce by exactly 31.1034768 grams. A product's metal value then uses its fine-gold mass, not automatically its gross mass.
Why does fineness matter in a gold value calculation?
Fineness is parts of gold per thousand by mass. A 10-gram item marked 750 contains 7.5 fine grams before any non-gold deductions; a 999.9 item contains 9.999 fine grams.
Does GoldObserve round every intermediate calculation?
No. Calculations retain JavaScript numeric precision through the transformation chain and round only for display or export formatting. The audit table exposes more decimals than the headline result.
Does chart sampling create or interpolate gold prices?
No. Display sampling is not data creation. When a long series is thinned for a chart, GoldObserve retains real returned observations at a regular stride and the final observation. It does not draw calculated closes into the dataset.
How are two time series matched?
Daily series normally use exact calendar-date matches. Lower-frequency macro research may use the nearest valid gold date inside a documented maximum distance. GoldObserve preserves both source dates, the signed day offset and aggregate match evidence; unmatched observations are omitted rather than silently forward-filled.
What does the NBU history represent?
The NBU official-reference route contains separately named gold, silver, platinum and palladium accounting series derived from date-matched metal-in-UAH and UAH-per-USD inputs stored locally. It is not an executable live quote, and the definitions are never silently spliced with monthly or third-party display data.
Can a FRED result change after it was first published?
Yes. FRED shows the latest available vintage and some source series are revised. GoldObserve currently uses that latest vintage and does not currently preserve an ALFRED vintage for every calculation.
Is annualized volatility a forecast?
No. It is the sample standard deviation of returned daily percentage changes multiplied by the square root of 252 and shown as a percentage. It describes the selected historical sample.
Is the calculated metal value the amount a dealer will pay?
No. Theoretical metal value excludes the dealer's bid, spread, assay, product condition, taxes, fees and settlement terms. Use an actual same-time counterparty quote for a transaction.