Gold trading costs, fully broken down: spread, overnight financing and slippage

The mid price on screen is 4,125.925. You hit buy and you fill at 4,126.256. Nobody explains that 0.662 gap, and it never shows up in a "fees" line on your statement — it was taken out of the account the instant you clicked. That is the spread: the one charge in gold trading that is levied on every trade and hidden every time. For how much platforms differ and how to measure it yourself, see gold spread comparison.

Most people lose money in their first year of gold trading not on direction but on costs they never counted. Someone who is right 60% of the time can still lose, because every trade pays a toll up front and every overnight hold pays for the room as well. When that bill is unclear, the account slowly leaks.

No mysticism here, just bookkeeping. Every number is calculated from the live market at 2026-10-21 13:32 Beijing time (UTC+8), so you can follow along with a calculator and then check your own account — most people find they hand far more to the market each month than they pay the platform in commission.

Remember four sentences; the nine sections below are their expansion:

1. Cost has three layers: the trade layer (spread, commission, slippage), the holding layer (overnight financing) and the account layer (deposits, withdrawals, FX). Most people only watch the first.
2. The spread scales linearly with lots; overnight interest accumulates with time and notional value. One lot = 100 oz, so every 1 USD move is 100 USD of P&L — that is the anchor for every conversion here.
3. Leverage does not change the spread in dollars, only its size relative to your capital. At 400x the spread is 6.42% of margin, and a round trip is 12.8%.
4. Low frequency, use physical; high frequency, use contracts. Round-trip cost on physical is about 3% against 0.032% on a contract — roughly 94 round trips to break even. Frequency picks the tool, not the other way round.

1. Put one trade's bill on the table

Say you go long 1 lot of XAUUSD at the current price — 100 oz, 412,592.50 USD notional (about CNY 887,600 at USD/CNY 6.6913 and CNY 887.61/g). No stop, no exit; just watch the cost grow.

Data note: prices from the XAUAD live feed (source Swissquote), snapshot 2026-10-21 13:32 Beijing time (UTC+8). Overnight interest is estimated at an annualised 3.5% on the long side; actual rates are published daily by each platform and move with dollar rates.

Look at the last row: hold for 10 nights and the cost is 461.80 USD, which means gold has to rise 4.62 USD just for you to get back to even. The average daily range over the past 20 sessions is 55.75 USD, so that 4.62 is only 8.3% of one day's range — which sounds small, except it is 8.3% handed over for free even when your direction is completely right. Do that ten times in a month and the number becomes 83%.

2. Spread: the only charge levied on every trade

The spread is the gap between bid and ask. Right now it is 0.662 USD, which is 1.6 basis points (0.016%) of the mid — about 6.62 bp. It looks negligible, but it is charged both ways: you buy at the ask on entry and sell at the bid on exit, crossing the gap twice.

In lots it becomes concrete:

CNY figures at 6.6913, for comparison only; your actual result depends on your deposit currency and the platform's settlement rate.

Now convert it into a per-gram price: 0.662 ÷ 31.1034768 × 6.6913 ≈ CNY 0.142/g. Compare that directly with the premium at a jeweller, which for investment bars is often CNY 3–30 per gram — twenty to thirty times more. So "contracts are cheaper than physical" is true on a single trade, but only if you do not hold overnight and do not trade frequently.

When the spread widens

The spread is not a constant; it follows liquidity. When liquidity is good, market makers quote tight; when it thins or a risk event lands, quotes widen fast to protect them. In Beijing time you can roughly remember it like this:

The ranges above are rules of thumb for a sense of scale; actual spreads come from the platform's live quote and vary by account type and volume tier. The current 0.662 USD is normal for the Asian session.

The practical value of that table: if you trade intraday, moving your session to 20:00–24:00 can take the spread from 0.66 down to about 0.35, nearly halving the cost of every trade. That requires no analytical skill at all, just a change of habit. For the session breakdown, check the live session table, which shows in Beijing time which market is currently trading.

3. Leverage turns the spread into a percentage of your capital

Many people think leverage only multiplies P&L. What it changes first is cost as a share of capital. The spread is a fixed 66.20 USD per lot, but the margin you post to hold that lot changes with leverage, and the ratio changes completely:

Margin = notional ÷ leverage, computed at the 4,125.925 mid. Actual margin requirements are set by the platform and account type and may be adjusted with volatility.

This explains something you see all the time: why short-term traders on 400x leverage do not last, even with a hit rate above 50%. Not because they get direction wrong, but because every open and close pays 12.8% of capital as an entrance fee, and eight of those finishes the account. The real cost of leverage is not the probability of liquidation, it is the linear rise in the cost ratio.

4. Overnight interest: the longer you hold, the more it costs, and Wednesday counts triple

Overnight interest (the swap) is the cost or credit for holding a position through the night. It does not come from platform markup but from an interest-rate differential: gold is priced in dollars, so being long is like borrowing dollars to buy gold, which means paying dollar interest; gold's own lease rate (what you earn lending it out) offsets part of that. The difference is the swap rate, and the platform adds its own markup on top.

The formula is simple, but plenty of people use the wrong base — it is charged on notional value, not on margin:

Thirty nights of interest exceeds the entire margin at 400x leverage — that is the arithmetic reason you cannot run a long-term position on a high-leverage contract.

Three details people miss:

• Wednesday usually counts triple. FX and metals settle T+2, so a position held through Wednesday crosses the weekend with a settlement date of the following Monday, and three days of interest are charged at once. The Wednesday-to-Thursday night costs three times normal — a short-term strategy that happens to span it will show an ugly number.
• Shorting does not always earn interest. In theory shorts receive it, but platforms add their markup to the swap rate, so often shorts pay too, especially when gold lease rates are low. Check the platform's swap table before you trade rather than assuming.
• Swap-free accounts are not free. So-called swap-free accounts (common in some religious accounts) usually move the cost into a wider spread or a fixed "management fee". The cost of money does not disappear, it just gets collected somewhere else.

So the first principle of short-term trading is not "get the direction right", it is avoid holding overnight. Close intraday and the whole overnight interest line is zero; on the same trade, one night versus ten nights is an eightfold difference in cost. No forecasting skill required.

5. Slippage: invisible, and the most expensive

Slippage is the gap between the price you saw and the price you got. It is not on any rate card and only appears when the market is moving violently, which is why it gets underestimated. A simple way to gauge its scale is against the average daily range. Over the 20 sessions measured here the average daily range is 55.75 USD, 39.69 USD over the last 5, with a 20-day high of 4,235.83 and low of 4,073.93; on 14 October the single-day range reached 97.00 USD.

Slippage ranges are empirical and depend on liquidity, order type and how the platform executes; during a gap a stop order can fill far worse than its trigger price.

Note the scale of the last two rows: one data-release slippage event can eat an entire year's spread budget in one go. So "do not trade the data" is not conservatism, it is a calculated decision. If you must trade it, at least use limit orders and accept "might not fill" rather than a market order and "will fill at a bad price".

6. Account-level costs: the layer overseas Chinese investors miss most

The three layers above all live inside the trading account, where you can see them. There is another layer outside it, and the amounts are often far larger than the spread: how the money gets in, how it gets out, and how many times it is converted on the way. If you live outside China, this layer is unavoidable.

Put those numbers into a concrete case: you wire 5,000 USD equivalent from a Singapore bank. The wire costs 25 USD and the FX conversion 0.5% (25 USD), so 50 USD in total — exactly 1% of capital. What is that 1% worth? At 0.1 lots with a 6.62 USD spread per trade, 50 USD equals the friction of seven and a half trades; at 1 lot it equals 0.76 of a trade.

Which is why "deposit small to test the water" is inefficient on cost: the fixed account-layer fees do not shrink with lot size. Wire 500 USD to test and the 25 USD wire is already 5% — you are down 5% before your first trade. The sensible approach is to fund in one properly sized amount (within what you can afford) rather than making repeated small transfers.

7. Compared with physical gold: 94 round trips to break even

Put the two instrument types in one table and the conclusion is obvious. Physical gold's cost is mostly the purchase premium and the buy-back discount (bars typically cost 1%–3% over, and buy-back is 1%–3% under), while a contract's cost is the spread and overnight interest. The former is one-off and predictable; the latter accumulates per trade and per day.

Work the break-even point with this article's numbers: 3% round-trip cost on physical against 0.032% on a contract (0.662 × 2 ÷ 4,125.925), so 3% ÷ 0.032% ≈ 94. In other words you have to complete 94 full round trips on a contract before its cost catches up with buying physical bars once. Conversely, if your plan is "buy twice a year and hold for three years", the contract has no cost advantage at all and adds liquidation risk; if your plan is "trade three times a week", physical is impossible — the spread alone would eat all the profit.

One more difference that gets ignored: physical gold has no overnight interest, a contract does. Hold a contract for a year and at 3.5% the financing is 3.5% of notional, already more than the round-trip cost of physical. So "contracts are cheaper" has strict conditions — short holding period, low frequency, no overnight. Outside those conditions the arithmetic reverses. For a finer comparison of instrument types see gold ETFs vs physical gold.

8. Seven ways to cut the cost

1. Move your main session to 20:00–24:00 Beijing time. The spread drops from 0.66 to about 0.35, nearly halving per-trade cost, with no analytical skill required.
2. Close intraday, do not hold overnight. That removes the whole overnight line (39.56 USD per lot per night) and avoids the Wednesday triple and weekend gaps.
3. Use limit orders, not market orders. Change "accept any price" into "only accept this price", and slippage goes from uncontrollable to controllable.
4. Do not trade in the five minutes around a data release. Slippage at NFP, CPI and FOMC is 8–23× the normal spread and can eat a whole year's spread budget in one trade.
5. Bring leverage below 100x. The spread falls from 6.42% of margin to 1.60%, effectively handing every trade an extra 4.8 percentage points of room for error.
6. Choose USD as the account currency and fund in one go. That avoids the 1%–3% double conversion and spreads the fixed wire cost thinner.
7. Once a month, compute cost ÷ gross profit. Add up all spread, overnight fees and funding costs for the month and divide by gross profit. Above 30% means your strategy's frequency or holding period does not match the instrument — change the instrument, do not add size.

The seventh is the one worth doing. Almost nobody audits this line, so they never learn whether their strategy is "bad direction" or "eaten by cost". Audit it once and you find that many strategies which look like a 50% hit rate are negative expectancy after costs, and the problem is not the entry signal but the holding time and trade frequency.

9. Three common misconceptions

Misconception one: a zero-spread account is cheaper

"Raw spread" or "zero spread" accounts usually charge commission instead, say 3–7 USD per lot. Converted properly that may be neither more nor less expensive; what matters is how many lots you trade. Commission scales with lots and so does the spread, so the two are interchangeable. What you should compare is cost per unit (dollars per lot), not whether the label says "commission".

Misconception two: swap-free means saving money

An account with no overnight interest usually puts the cost into a wider spread, higher commission or a fixed account fee. The cost of money does not vanish, it moves. The test is simple: compute the total cost of both account types over the same holding period, rather than checking whether one line item is zero.

Misconception three: the cost is too small to matter

At 0.016% per trade it certainly is small. But cost accumulates with frequency and time: five round trips intraday is 0.16% a day; twenty trading days is 3.2% a month; add overnight financing at 3.5% annualised if you hold. Over a year the cost can easily exceed 30% of capital — at which point talking about returns is pointless.

10. Frequently asked questions

Q1. What is a normal gold spread?
A: In this article's live quote of 2026-10-21 13:32 Beijing time (UTC+8), the bid was 4,125.594 and the ask 4,126.256, a spread of 0.662 USD/oz, about 6.6 basis points or 0.016%. In lots: 66.20 USD on 1 lot, 6.62 USD on 0.1 lot, 0.66 USD on 0.01 lot. It is usually tighter in the London–New York overlap and clearly wider in the Asian afternoon and at the instant of a data release.

Q2. How is overnight interest calculated, and why is it charged?
A: The formula is notional value × annualised swap rate ÷ 365 (some platforms use 360). One lot is about 412,592.50 USD notional, so at a 3.5% annualised long rate it is about 39.56 USD per night and about 118.69 USD for three nights. It comes from the gap between dollar rates and the gold lease rate: being long is like borrowing to buy gold, so you pay; shorts can sometimes receive. Most platforms charge triple on Wednesday because settlement at T+2 crosses the weekend.

Q3. Does higher leverage make trading more expensive?
A: The spread in dollars does not change, but its share of margin rises linearly with leverage. One lot at 20x needs 20,629.62 USD of margin, making the spread 0.32%; at 100x margin is 4,125.93 USD and the spread is 1.60%; at 200x it is 3.21%; at 400x margin is only 1,031.48 USD and the spread is 6.42%, with a round trip at about 12.8% of capital. High-leverage short-term trading therefore has to manage the spread as its number one cost.

Q4. Can slippage be avoided?
A: Not entirely, but it can be controlled. Limit orders instead of market orders, staying out of the five minutes around data releases, and avoiding gap windows around the open and close all reduce it significantly. For scale: the average daily range over the last 20 sessions was 55.75 USD with a maximum single day of 97.00 USD; at an NFP or CPI release, 5–15 USD of slippage is common, 8–23× the normal spread.

Q5. What other costs do overseas Chinese traders overlook?
A: Account-layer costs are often dearer than the spread. A bank wire is typically 20–40 USD plus 10–25 USD in correspondent fees, FX conversion is 0.3%–1%, and withdrawal 0–25 USD; a non-USD account pays 0.5%–1.5% per conversion, so 1%–3% round trip. On a 5,000 USD deposit, a 25 USD wire plus 0.5% conversion is about 50 USD — exactly 1% of capital, equal to the spread on seven and a half 0.1-lot trades.

Quotes from Swissquote; prices move with the market and the figures here are a snapshot at 2026-10-21 13:32 Beijing time (UTC+8). Spreads, swap rates and margin requirements are as published live by each platform.