How to measure gold volatility? ATR, hourly range and the 60-second payoff, measured

"Is gold volatile right now?" maps onto at least three completely different numbers: how many dollars it moved in a minute, what the ATR reading is, and what the annualised volatility is in percent. They are not interchangeable, and mixing them produces inverted conclusions — using a daily range to judge whether 60 seconds is tradeable, or using ATR to work out the spread ratio, are both standard mistakes.

This article does one thing: it separates the three measures, gives the formula and the measured reading for each, and lands on one executable sentence — in this minute, is there enough room to cover my cost. Which hour to trade is a separate article, see "When is the best time to trade gold? The three sessions and the best window, measured". The division of labour is: that article sets the time, this one sets the ruler. Fix the hour with that piece first, then use this one to work out how much of each dollar of room at that hour is actually yours.

1. Three numbers first: the conclusions are all here

• Median 1-minute range 1.225 USD (p75 1.855, p90 2.670), which is 1.85× the current spread of 0.662. In the typical minute, the high-low gap is just under twice the spread.
• But only 42.8% of minutes see a net move larger than the spread. The median net move (absolute close − open) is only 0.520 USD, smaller than the spread. A wide range does not mean you can capture it — this is the most overlooked truth about 60-second products.
• ATR14: 3.663 USD on the 5-minute chart, 1.328 USD on the 1-minute chart; realised volatility over the same window annualises to 21.27%. The daily range over the five days ran from 35.32 to 97.31 USD, the largest 2.75× the smallest.

Live gold prices and the current spread are on the today's gold price page, refreshed every minute; the numbers here are a cross-section from the timestamp above, which is exactly why they are traceable and exactly why they must go stale.

2. Three measures: do not mix them

Volatility is not a number, it is a family. Short-term trading needs only three of them, and each answers one question.

One year of 1-minute periods is 60×24×365 = 525,600; switch to 5-minute periods and the count becomes 105,120, so the same price series annualises lower — comparability of annualised figures depends on the period, so fix the period before comparing across measures.

2.1 Range: the most intuitive, and the most deceptive

Range only looks at the high-low gap, never the path. If price runs from 3,990 to 3,993 and back to 3,990 within a minute, the range is 3 USD and the net move is zero. For a directional trader those 3 USD are visible but unreachable; for a market maker they are the source of profit. So range answers "is there room", not "whose room is it". Every coverage calculation below uses range, because what the spread consumes is room, not direction.

2.2 ATR: the only measure that carries gaps

ATR differs from range in two ways: it pulls in the previous close (|H−C₋₁|, |L−C₋₁|), and it smooths over 14 periods. The result is that gaps, overnight jumps and momentary spikes all get counted, while a simple high-low difference does not. That is why on the same day ATR(5m) at 3.663 is clearly above the 5-minute median range of 3.045 — the 0.618 USD difference is essentially the contribution of gaps and spikes.

It is also why ATR can set a stop distance but cannot compute a spread ratio: a stop guards against the worst candle, while cost is calculated on the average minute.

2.3 Realised volatility: for comparing across time

The virtue of an annualised reading of 21.27% is that it can be compared across instruments and across months (gold in the low 20s is a normal range; above 40% is event-driven). The drawback is that it has almost no operational value for short-term trading — annualised 21% converts to a one-minute standard deviation of 0.00032, about 1.28 USD/oz of price movement, a number with no direct bearing on the P&L of your next trade.

3. Measurement one: hourly 1-minute range (the full 24-hour table)

The 7,423 minutes are binned by Beijing-time hour, 300 samples per bin (5 days × 60 minutes), with the median range for that hour and the coverage multiple "range ÷ spread in that period".

The 0.000 in the 05:00 row is not a typo: the quote store holds no valid quotes at all between 05:00 and 06:00 across those five trading days, matching the platform's daily maintenance window (the endpoint's weekend_note says 20:00–24:00 UTC, i.e. 04:00–08:00 Beijing; the actual gap concentrates at 05:00–06:00). That hour is excluded from all statistics.

Two peaks are worth remembering: 09:00–10:00 at 2.203 USD (3.33×) and 21:00–22:00 at 2.165 USD (3.27×). The first is the hour after Tokyo's real session opens; the second is where US data and the US equity open overlap. Meanwhile the "8pm to midnight" window quoted endlessly in Chinese-language material measures 1.980 at 20:00 but has already fallen to 1.415 by 23:00 — the back half of that window decays sharply, and the genuinely thick part is 20:00–23:00.

4. Measurement two: how to read and use ATR

Readings this time: ATR14(5m) = 3.663, ATR14(1m) = 1.328 (last 5-minute candle closed at 2026-10-24 01:40 Beijing). Converted into directly usable distances:

• Stop: 1.5 × ATR(5m) ≈ 5.49 USD. Place it at 3 USD and normal noise takes you out — the 5-minute median range alone is 3.045, with a p90 of 6.320.
• Target: 2–3 × ATR(5m) ≈ 7.3–11.0 USD. A target below 1.5× means betting a 50% win rate on insufficient payoff.
• Match the horizon: ATR(1m) 1.328 belongs to 60 seconds, ATR(5m) 3.663 belongs to 5 minutes. Use a 5-minute stop on a 60-second trade and the clock expires before the stop is reached; use a 60-second stop on a 5-minute trade and you will be stopped out for certain.

The most common error when reading ATR is treating it as directional. ATR says only how far a candle can travel, not where. When ATR jumps from 3.6 to above 5, the correct response is to lengthen the horizon or stand aside, not to add size — the source of that volatility is usually an event like nonfarm payrolls, CPI or FOMC, and slippage inside an event window runs 8–23× the normal spread (method in section 4 of "Gold trading costs, fully broken down").

5. Measurement three: coverage and break-even win rate by session

Grouping the hours into sessions, and adding the break-even win rate — under the assumption that stop = target = the session's median range and each entry and exit pays half the spread, the win rate needed just to break even: p* = 0.5 + spread ÷ (2 × target size).

Two counter-intuitive things in this table deserve pulling out.

First, the Asian session's 1-minute range is not small (1.475, second only to the Europe/US overlap). It is concentrated in the 09:00–10:00 hour (2.203), where Tokyo's real session and Asian block orders collide. So the conclusion "you can't do 60 seconds in Asia" depends on the clock: across 08:00–15:00, 53.1% of minutes see a net move above the spread, essentially level with the overlap (54.2%) and clearly better than late US (29.4%). What genuinely does not work is the quiet 04:00–08:00 stretch — median range there is 0.501, smaller than the spread, so any direction is just working for the spread.

Second, the overlap's value only shows fully at 5 minutes and above. Its 5-minute median range of 4.650 and p90 of 8.875 are a tier above Asia's (3.620 / 6.430), while at the 1-minute level the gap between them is small. In other words: for 60 seconds, the Asian morning and the Europe/US overlap can break even; for 5 minutes and beyond, the overlap is clearly thicker. That matches the ranking the session article derives on a 15-minute basis — different measures, same direction, so the two corroborate each other.

6. The payoff maths of 60 seconds: 59.5% and 116%

Two sets of arithmetic. Do not confuse them.

6.1 Fixed-payoff products: the break-even line is set by the payoff alone

The site's /api/config returns rates.60 = 0.68, meaning a winning 60-second trade pays 68% of stake and a loser forfeits it. Break-even win rate p* = 1 ÷ (1 + 0.68) = 59.52%. That number is independent of volatility: however violently price moves, the payoff does not change and neither does the break-even line. Volatility only determines whether you can push your win rate above 59.5% — it is a function of entry quality, not of the payoff.

6.2 Spot / CFD: the break-even line floats with the spread and the target

p* = 0.5 + spread ÷ (2 × target size). With the spread fixed at 0.662 and the target set to that session's median range, the last column of the previous table follows: Europe/US overlap 68.0%, Asia 72.4%, Europe 75.9%, late US 91.9%, quiet 116.1%. 116.1% is an impossible number — it means the quiet stretch offers no mathematical basis for short-term trading at all, and that is not a failure of your technique.

6.3 The most uncomfortable number: 42.8%

Across all 7,423 minutes in the window, only 42.8% saw a net move (|close − open|) larger than the spread, with a median net move of 0.520 USD, p75 1.092 and p90 1.850. Translated into trading language: if you enter and exit at a random minute, six times out of ten price does not even clear the fee. A range of 1.225 looks 85% bigger than the spread, but that is the high-low gap, not the gap you capture — you open in the middle and close in the middle, giving up half the spread at each end.

Which is why the decisive factor in a 60-second product is not an indicator but choosing hours with sufficient coverage, and only trading the right tail of the net-move distribution. How thick that tail is can be computed: the p90 net move here is 1.850, or 2.79× the spread — meaning only one minute in ten offers room wide enough for a decent target.

7. When volatility lies to you

7.1 Change the window and the conclusion changes

The same method, three different windows:

The two 48-hour windows differ by 33.6% in median range, and the share of minutes clearing the spread rises from 38.4% to 51.1%. So whenever someone hands you a number like "gold moves X dollars a minute", the first questions are always: which window, how many samples, which measure. Without those three, the number is unusable.

7.2 The tail matters more than the mean

Full-day range (high − low, split on Beijing calendar days) across the five complete trading days: 10-19 50.58, 10-20 43.84, 10-21 35.32, 10-22 97.31, 10-23 77.43 USD. The calmest day and the most violent differ by 2.75×. Size positions off the mean of 60.9 and you will look far too conservative on a day like 10-21 and completely under-sized on 10-22 — short-term position size should be set from p90, not the mean.

7.3 Gaps and outages

Quotes are empty from 05:00 to 06:00 (Beijing) on all five consecutive days, matching the platform maintenance window; weekends are a full closure — the endpoint states "stop Friday 20:00 UTC, resume Monday 00:00 UTC", i.e. stop Saturday 04:00 Beijing, resume Monday 08:00. The first candle after any of those three points has its ATR pushed up by the gap, so it looks "very volatile" when in fact that is a hole, not liquidity. Trading 60 seconds into a gap means paying maximum cost for minimum certainty.

8. Applying the three numbers from 60 seconds out to 5 minutes

Three rules, each checkable against the figures in this article.

1. Below 2× coverage, do not trade 60 seconds. From the hourly table, the hours with a coverage multiple ≥ 2 are 08, 09, 10, 13, 14, 15, 19, 20, 21, 22 and 23 — eleven of them; in the other thirteen, the 1-minute range does not even reach twice the spread, so entering is simply working for the spread.
2. At 5 minutes and beyond, set stops from ATR, not from a fixed number of points. With ATR(5m) at 3.663, the stop is 5.49 and the target 7.3 or more. A fixed 3 USD stop, against a 5-minute distribution with a median range of 3.045 and a p90 of 6.320, hands more than half your trades to noise.
3. If ATR suddenly widens by 40% or more, shorten the horizon before discussing direction. 3.663 × 1.4 ≈ 5.13, so once ATR(5m) holds above 5 you are in an event window; slippage risk on 60 seconds multiplies, so extend the horizon to 15 minutes or simply wait for the event to pass (for handling before and after events see the FOMC piece and the CPI piece).

One more point on ordering: fix the session first, then the ruler, and direction last. The session decides which pool you are swimming in (see the session article), the ruler decides whether this shot is worth taking (this article), and direction is only the final layer. A lot of people reverse that order, spend six months practising direction, then discover that six of ten trades landed on hours with coverage below 1× — that is not a technique problem, it is a venue problem. For a more detailed short-term framework, continue with gold short-term tactics.

9. Frequently asked questions

Q1. What is a normal volatility level for gold?
A: It depends on the measure. On the tick data from 2026-10-18 22:00 to 10-24 01:42 (Beijing) used here, annualised realised volatility is 21.27%; in point terms, the 1-minute median range is 1.225 USD, the 5-minute 3.045 USD, and the daily range 35.32–97.31 USD. Different windows vary by more than a third, so always quote the interval with the number.

Q2. What is the difference between ATR and range, and why do the two numbers differ?
A: Range takes only the period high minus the period low; ATR additionally brings in the previous close (|H−C₋₁|, |L−C₋₁|) and applies 14-period Wilder smoothing, so gaps and spikes get counted. Here ATR(5m) at 3.663 exceeds the 5-minute median range of 3.045, and the 0.618 USD difference comes mainly from gaps and long wicks.

Q3. Does higher volatility make it easier to make money?
A: No. Volatility decides only whether there is enough room to cover cost, never direction. In the measurement here only 42.8% of minutes see a net move above the spread, meaning random entries fail to clear cost six times out of ten; mistaking range for capturable profit is the most common short-term illusion.

Q4. What is the break-even win rate on a 60-second product?
A: It is set by the payoff. The site's /api/config returns rates.60 = 0.68, so the break-even win rate is 1 ÷ 1.68 = 59.52%, independent of volatility. Converted to a spot basis (stop = target = that session's median range), the Europe/US overlap needs 68.0%, Asia 72.4%, late US 91.9% and the quiet stretch 116.1%.

Q5. Why is the 05:00 range zero, and when can I trade at the weekend?
A: 05:00–06:00 (Beijing) is the platform's daily maintenance window, and the quote store holds no valid quotes in that hour across the five trading days, so it is excluded. The weekend closure is configured as Friday 20:00 UTC to Monday 00:00 UTC, i.e. stop Saturday 04:00 Beijing, resume Monday 08:00; the first candle after the restart contains a gap and is not a good place to take a 60-second trade.

The data here is a cross-section of the quote store and live endpoint readings at 2026-10-24 01:42 Beijing (UTC+8); prices come from Swissquote and quotes and spreads move with the market. Page figures do not update on refresh — always rely on the live quote in the trading interface.