XAUUSD has no real volume — and your indicators know it

What tick volume actually counts

On an exchange-traded instrument, volume is a fact about participation: how much was actually bought and sold. On XAUUSD, tick volume is a fact about how busy your broker's price feed was. Those two things correlate under normal conditions and diverge precisely when it matters.

Feed activity depends on your broker's liquidity providers and how they quote, on server load and how updates are aggregated and throttled, on session overlaps and news events, and on the broker's own feed configuration. So any indicator built on volume is, on XAUUSD, a function of quote-update frequency. It is not measuring how much gold changed hands, because nobody knows how much gold changed hands.

Five things this breaks

• Volume indicators measure the wrong thing. OBV, VWAP, volume-weighted averages, volume confirmation, volume filters and volume profile are all functions of quote-update frequency on XAUUSD. The tell: tick volume for the same instrument and the same minute differs between brokers. Real volume would be identical everywhere, because it would be the same market.
• Market impact cannot be modelled. A backtest fills you at the quoted price. On an exchange you can at least inspect depth to judge whether that holds for your size; on XAUUSD there is no depth to inspect. For retail lots this is usually a reasonable simplification — but it is an assumption, not a verified fact, and it degrades silently in fast markets.
• The spread is broker-determined, not exchange-determined. On an exchange the spread emerges from competing orders; on XAUUSD your broker sets it. It varies by broker, account type, time of day and news. A backtest's spread assumption is a guess unless you measured it on your own broker's feed over the period you are testing.
• Overnight financing is an estimate. Swap on an OTC CFD is set by the broker, not by an exchange settlement process. Long-horizon backtests that approximate or omit it are systematically optimistic for strategies that hold across the roll.
• Gaps are feed artefacts, not market events. Weekend gaps, session gaps and sudden tick gaps reflect when your broker's feed was live and how it handled the close and reopen. A gap-fill strategy tuned to these artefacts is tuned to your data vendor, not to the gold market.

XAUUSD and COMEX gold futures are not the same instrument

GC trades on COMEX, a real exchange with real volume, a real order book and real settlement. XAUUSD is a spot-referenced CFD priced off the same underlying but traded over the counter. They track each other closely, and they differ in trading hours, settlement, contract size, margin, financing and — at any given moment — price.

The consequence for evaluation is what matters most: a strategy generated and tuned on COMEX gold futures daily data is not thereby validated on XAUUSD hourly data. The two datasets have different bar boundaries, different sessions, different gaps and different costs. Scoring one against the other produces out-of-sample statistics that look meaningful and are not. We treat that as an error condition rather than a convenience.

There is a third case worth naming. Some datasets are constructed from a gold-backed token traded on a crypto exchange, used as a price proxy for continuous 24/7 coverage. That is a legitimate proxy for some purposes and it is not the same thing as broker XAUUSD quotes. Mixing the two without labelling them is how a strategy ends up looking robust when it has only ever seen one of them.

How to check this yourself in ten minutes

None of the above has to be taken on faith. Every claim is verifiable, and the first check takes about a minute.

• Open the same instrument on two different brokers' MT5 terminals and compare the volume reading for the same minute. They will differ. Real volume cannot differ between two views of the same market.
• Check whether your broker's XAUUSD symbol shows an exchange name in the specification window. If the venue is blank or broker-internal, there is no exchange.
• Look for depth-of-market on XAUUSD. If the DOM is empty, or shows only your broker's own quotes, there is no order book.
• Compare XAUUSD and GC over the same period in your platform and look at bar boundaries and session gaps. They will not align.
• Read your broker's contract specification for spread and swap terms, and note whether they are fixed, floating, or marked as indicative.

What to do about it

• Use tick volume only as a relative measure of quote activity within a single broker's feed — never as an absolute measure of liquidity, and never as a cross-broker comparison.
• If your strategy genuinely depends on volume, test it on an instrument where volume exists: exchange-traded gold futures, or major crypto pairs on large venues. Both give true volume, and both come with a different market microstructure.
• Measure your own broker's spread and slippage rather than assuming them. Two traders running the same strategy with different cost assumptions will get different results, and neither is 'the' backtest.
• Record which dataset a strategy was evaluated on, and never mix instruments in one evaluation.
• Re-run with conservative cost and financing assumptions, especially for strategies that trade frequently or hold overnight.

How EasyQuant labels datasets

Because dataset provenance is not a detail, we treat it as structure rather than documentation. Datasets are labelled by source — broker MT5 data, exchange futures data, and proxy datasets carry distinct keys and are documented separately rather than pooled. Known limitations are recorded per dataset, so the absence of real volume on XAUUSD is a stated property of that dataset rather than something a user discovers after building on it.

Cross-dataset evaluation is treated as an error: a strategy is evaluated against its home dataset, and multi-market testing is a deliberate, labelled operation rather than a silent fallback. Instruments that cannot be independently re-verified are marked as such rather than given a misleading pass. And cost and financing assumptions are explicit inputs you can see and change, not hidden constants.

This is the same principle that runs through the rest of the platform: the failure modes should be visible on the page, not buried in a footnote.