XAUUSD Historical Data: Gold Price OHLC Data Download

XAUUSD historical data records the price of one troy ounce of gold in US dollars as a series of bars, each with an open, high, low and close (OHLC), on four time frames (hourly, daily, weekly and monthly), each downloadable as CSV. Both “Change” columns compare a bar’s close with that same bar’s open, not with the previous bar’s close. That choice changes daily return figures, volatility estimates and any strategy test built on them.

What each row measures

XAU is the ISO 4217 code for one troy ounce of gold, which weighs 31.1035 grams. XAU is the base and USD is the quote, so a reading of 2,000.00 means one ounce costs 2,000 US dollars.

Each row in the tables is one bar:

The high and low tell you how far price traveled. The open and close tell you where it started and finished. A bar with a wide high-to-low range and a close near its open describes a period when gold moved a lot and went nowhere. A narrow bar that closes near its high describes a quiet, one-directional period.

How the change columns are calculated

The figures in the tables reconcile to these formulas:

• Change (pips) = (Close − Open) ÷ 0.10
• Change (%) = (Close − Open) ÷ Open × 100

So a pip in this dataset is ten cents per ounce, and a one-dollar move equals 10 pips. Some brokers define a gold pip as 0.01 instead, so the same move can appear as 100 pips in one place and 1,000 in another. When sizing a position, match your broker’s pip definition; a pip calculator for XAUUSD handles the conversion.

Open-to-close versus close-to-close

Most published “daily change” figures, including the headline change on a live quote, compare the current price with the previous close. The tables here compare close with open of the same bar. The two numbers differ whenever a bar opens at a different level from the prior close, which happens regularly after weekends and after the daily trading break.

Hypothetical example (not market data):

• Previous daily close: 2,010.00
• Today: Open 2,000.00, High 2,030.00, Low 1,985.00, Close 2,020.00

The table would show gold up 1.00% for the day, while the change from yesterday’s close is half that. Summing the open-to-close percentages across many bars will not reproduce the overall change in price, because every gap between one bar’s close and the next bar’s open drops out of the total. For returns over a period, use closes: (Close at end ÷ Close at start) − 1.

Time frames and how bars are built

Spot gold trades over the counter almost around the clock from Sunday evening to Friday evening (US time), with a short daily break. The way a data provider cuts that continuous stream into periods decides what each bar contains.

Hourly bars each cover sixty minutes. The hourly table includes a row from Sunday evening, when trading reopens after the weekend, and none for the closed hours of Saturday. The newest row may still be forming: a bar where open, high, low and close are identical usually means the hour has just started and only one price has been recorded. Treat the latest row of any time frame as provisional until its period ends.

Daily bars appear only for weekdays. Sunday evening trading is folded into Monday’s bar: the low of the Monday daily bar matches the low of the Sunday evening hourly bar that preceded it. If you build your own daily bars from hourly data using midnight-to-midnight calendar days, you will get a different Monday and a stray Sunday bar, so your OHLC will not match these.

Weekly bars are labeled with the first trading day of the week, normally a Monday. When Monday is a US holiday, the label can fall on Tuesday. Monthly bars are labeled with the first trading day of the month, which is why a month can carry the date of the 2nd or 3rd.

In the downloads, 120 hourly rows cover roughly one trading week, 260 daily rows about a year of weekdays, 156 weekly rows about three years, and 60 monthly rows five years.

Gaps between bars

The open of a bar does not have to equal the close of the bar before it. A gap appears when price moves while the market is closed (over the weekend or during the daily break) or when the provider’s first recorded quote of a new period differs from its last quote of the previous one. Weekend gaps in gold can be large after geopolitical news. A stop-loss order placed inside a gap fills at the first available price when trading resumes, not at the stop level, and historical bars will not show you that fill.

Working with the CSV file

The downloaded file contains the same seven columns as the on-screen table. To use it in Excel, Google Sheets or a scripting language:

1. Sort by date ascending. The tables list the newest bar first. Most calculations and charting tools expect the oldest bar first.
2. Check number formatting. Prices use a comma as the thousands separator. Spreadsheets set to a European locale may read the comma as a decimal mark, turning 2,000.00 into 2.00. Import with the separator set explicitly, or strip the commas.
3. Recompute what you need. Keep the change columns for reference, but calculate your own close-to-close returns if you want daily performance.

Useful derived columns:

• Simple return: Closeₜ ÷ Closeₜ₋₁ − 1
• Log return: ln(Closeₜ ÷ Closeₜ₋₁). Log returns add up across periods, which makes them convenient for volatility work.
• True range: the largest of (High − Low), |High − previous Close| and |Low − previous Close|. Unlike the plain high-low range, it captures gaps.
• Average true range (ATR): an average of true range over a set number of bars, commonly 14. ATR in dollars, divided by your pip size, gives a typical move in pips for sizing stops.
• Realized volatility: the standard deviation of daily log returns, multiplied by the square root of the number of trading days in a year (about 252 is the usual convention) to annualize it.

Backtesting with historical OHLC

OHLC bars are a summary. They tell you the high and low of a period, not the order in which they occurred. If a backtest places both a stop and a target inside the same bar, the bar alone cannot say which was hit first, and testers that assume the favorable one overstate results. Testing on a shorter time frame than the strategy trades reduces this problem.

A few further adjustments make a test closer to live trading:

• Costs. Historical prices typically reflect a single quote, not the separate bid and ask your broker charges. Subtract the spread and any commission from every trade.
• Contract size. Many brokers define one standard gold lot as 100 ounces, which would make a one-dollar move worth 100 dollars per lot. Contract specifications vary, so use your broker’s figure. A position size calculator converts a stop distance and risk amount into lots.
• Overnight financing. Positions held past the daily rollover accrue swap charges or credits; add these to each trade’s result.

The file has no volume column, and because spot gold trades over the counter there is no single consolidated volume figure; the “volume” in broker platforms is tick volume, a count of price updates on that broker’s feed.

Minute-level data

Five-minute and one-minute bars for XAUUSD come mainly from trading platforms and broker servers. MetaTrader 5, for example, lets you request bar history for a symbol from your broker and export it to CSV, and that history will follow your broker’s prices, session times and server time zone. Futures exchanges also sell intraday data for gold futures, which trade at a different price from spot because the contract includes carrying costs to its delivery date. Whichever source you use, keep it consistent between backtesting and live trading.

Why XAUUSD prices differ between sources

Two datasets of “gold price history” rarely match to the cent. The usual reasons are:

• Instrument. Spot gold, the nearest gold futures contract and the twice-daily LBMA benchmark price are different prices.
• Quote type. A file may record bid, ask, mid or last-traded prices.
• Session cut. One provider closes the day at New York 5 p.m., another at midnight UTC, another at the broker’s server time. Each produces different daily highs, lows and closes from the same underlying trades.
• Time zone of timestamps. An hourly bar labeled 14:00 in one file can be 09:00 in another.

The data comes from Yahoo Finance, with update times in UTC. Small differences from broker data in highs, lows and daily closes are expected for these reasons and do not by themselves indicate a bad feed.

What moves gold against the dollar

Gold pays no interest or dividend, so its price responds to the return available on alternatives and to the value of the currency it is quoted in. The main drivers are:

• Real interest rates. When US yields after inflation rise, holding a non-yielding asset costs more in forgone income, and gold has tended to come under pressure. Falling real yields work the other way.
• The US dollar. Because gold is priced in dollars, a stronger dollar makes it more expensive for buyers using other currencies. The relationship is not fixed, and it can be checked against other markets on the XAUUSD correlation page.
• Federal Reserve policy and US data. Rate decisions, Fed speeches, inflation (CPI) and employment releases change rate expectations and often produce the widest hourly bars in the data.
• Risk and geopolitics. Wars, financial stress and political shocks have historically drawn buyers looking for an asset outside the banking system.
• Physical demand. Central bank reserve purchases, jewelry demand and investment in bars, coins and exchange-traded funds add or remove buying over longer horizons.

On a chart these drivers show up differently by time frame. A data release appears as one wide hourly bar; a shift in rate expectations appears as a run of monthly bars in one direction.

A brief history of the gold price

Free-market price history starts in August 1971, when President Nixon suspended the dollar’s convertibility into gold at a fixed 35 dollars per ounce. Since then gold has climbed to about 850 dollars in January 1980, bottomed near 250 dollars around 1999 to 2001, reached around 1,900 dollars in 2011, fallen below 1,100 by late 2015 and first crossed 2,000 dollars in 2020. Long-run statistics on gold therefore depend heavily on the window chosen: a study that starts in 1980 and one that starts in 2001 will reach opposite conclusions about the same instrument.