Trading gold on NFP day:the 2026 release calendar and what the volatility actually measures

NFP is the easiest gold session of the year to get wrong — not because direction is hard to guess, but because it compresses the volatility into a few dozen seconds. On the release tick the spread can jump from USD 0.70 to USD 43.20, be back at 0.70 thirty seconds later, and by then price has travelled 40 to 100 dollars. The feel you built up in normal conditions stops working, because nothing about your judgement changed — the cost structure did.

This article does three things only: it lists the exact time of every remaining 2026 release (the “first Friday at 20:30” rule of thumb is wrong seven times in 2026); it uses COMEX gold futures 5-minute candles to measure how far price actually travels on the release instant; and it uses tick-level bid/ask from XAUAD’s own quote database on 2026-10-02 to show what happens to the spread during those ten minutes. It closes with rules you can execute between 60 seconds and 5 minutes.

1. NFP is three numbers, and the market really trades the third

The Employment Situation report releases three sets of figures at once: headline payrolls, the unemployment rate, and average hourly earnings. Most commentary only covers the first, yet gold is most sensitive to the third.

• Headline payrolls decide whether the economy is hot. A beat sends the dollar and Treasury yields up and gold down — that is the textbook path.
• The unemployment rate comes from the same household survey and often diverges from payrolls (on 2026-08-07 payrolls were strong and unemployment rose too). When they diverge, the market freezes for about 30 seconds before pricing anything.
• Average hourly earnings point straight at inflation: wages up → services inflation sticky → inflation expectations up → real yields (nominal minus expected inflation) repriced. Gold’s opportunity cost is the real yield, which is why this set of numbers tends to have a longer-lasting effect than the headline.

There is one more set that almost nobody reads and every institution does: the revisions to the previous two months. The initial print is a sample estimate and gets revised twice; revisions routinely run 50,000 to 100,000. When payrolls come in at 120k but the prior two months are revised down by 80k, the market concludes that hiring is weaker than it looked and gold can rally on a “strong” print. That is why “good data, short gold” so often gets taken out by the reversal.

Follow the chain to the end and the point becomes clear: NFP moves gold not through one dimension called “the economy” but because it simultaneously rewrites growth expectations and inflation expectations. The first moves the dollar, the second moves real yields, and the two legs often kick in opposite directions. CPI, released on a different day, runs through a more direct channel; the division of labour between them is set out in CPI and gold: what happens to the price on inflation day.

2. All twelve 2026 releases, converted to Beijing time

NFP is fixed at 8:30 US Eastern (per the official BLS calendar). Converting to Beijing time takes two steps: add 13 hours during US daylight time (EDT) or 14 hours during standard time (EST). US daylight time in 2026 runs from 8 March to 1 November, so Beijing time is:

Source: bls.gov, Schedule of Selected Releases 2026, pulled 2026-10-24. Beijing time = 8:30 US Eastern + 13 or 14 hours.

Check that table against the most repeated rule of thumb — “first Friday of the month at 20:30” — and seven of the twelve 2026 dates do not fit: 9 January is the second Friday and 21:30; 11 February is a Wednesday; 6 March is 21:30; 8 May is the second Friday; 2 July is a Thursday; and both 6 November and 4 December are 21:30. There are two kinds of error: a date off by a day to a week, and a time off by a full hour — and the latter is enough to make you miss the entire move on a 60-second or 5-minute product.

So the first discipline on a data day is not technical analysis, it is writing the calendar down. After daylight time ends on 1 November, the New York open also shifts from 20:00 to 21:00 Beijing, pushing the overlap window later; the session-by-session detail is in the best hours to trade gold, measured session by session.

3. Measured: how far gold travels in the five minutes after the release

Take the GC=F 5-minute candles for the two most recent NFP releases in 2026 (4 September and 2 October). Both land on the 12:30 UTC bar (20:30 Beijing); candle labels are open time, and the close is the price five minutes after the release.

Multiple = the bar’s 5-minute range ÷ the median 5-minute range across the whole sample, USD 3.90 (13,639 bars, 2026-08-14 to 10-23). Source: Yahoo Finance GC=F, pulled 2026-10-24.

4 September is the textbook case: price first pushed USD 5.8 higher from 4,517.70 to 4,523.50, then collapsed to 4,419.00 inside five minutes and closed at 4,441.90. The bar’s range of 104.50 dollars is 26.8× the median 5-minute range — that single candle absorbed the movement of twenty-seven normal ones. 2 October was gentler at 42.50 dollars, still 10.9× the median.

More interesting is the hour after the release. After the drop to 4,419.00 on 4 September, price did not keep going: it chopped between 4,430 and 4,456 for half an hour, spiked to 4,412.00 on the 13:00 bar, then climbed all the way back to 4,461.00 by 13:30 and closed 13:45 at 4,476.90 — 64.90 dollars above the low. Short the bottom and by 13:45 you had given back every dollar of profit. 2 October was the mirror image: after the push to 4,259.00 price ground lower, back to 4,220.40 by 13:20 and 4,216.10 at 13:40 — precisely the pre-release level, a complete round trip.

Both cases say the same thing: the direction of the first five minutes after NFP and the direction of the next hour are frequently opposite. Trading 60 seconds to 5 minutes means trading the first leg; trading 30 minutes or more means trading the second. Carrying a first-leg conclusion into a second-leg position is the most common way to lose money on a data day.

4. The volatility is compressed into one hour — the day itself is not necessarily bigger

Many people assume that “it is NFP today, so the whole session needs watching”. Splitting the release-hour range from the full-day range across the last six releases says otherwise:

Method: the release hour is the 12:00 UTC bar of the GC=F 60-minute series (covering 12:00–13:00 UTC, including the 12:30 release); the prior-20-day median is the median range of the same 12:00 UTC hour over the twenty preceding sessions; the full-day multiple is the day’s high-low range ÷ the median full-day range of the prior 20 days. Source: Yahoo Finance, pulled 2026-10-24.

In five of the six cases the release hour ran 2.2 to 4.5× a normal hour in that slot (median 3.11×), while the full-day range was only 0.90 to 1.68× — and on 8 May it was 0.90×, meaning the day NFP printed was quieter end to end than the twenty before it.

That is the true shape of a data day: not “wild all day”, but “a whole day’s range stuffed into one hour”. The implication for order placement is direct — you do not need to scale up for the full session, you need full attention from 20:00 to 21:00 and normal pace the rest of the time. The rest of the day is often easier, because the volatility has been drained out of it.

5. What actually happened to the spread during those ten minutes

The table below is a tick-by-tick extract from XAUAD’s quote database for 2026-10-02 (Swissquote feed, one tick every 30 seconds, 21 ticks in ten minutes), showing eight of them:

Baseline: the same feed between 11:00 and 12:20 UTC, 160 ticks, median spread 1.00 USD, p90 3.20, maximum 5.00. Source: our quotes table, sampled 2026-10-24 22:05 Beijing.

That 43.20 tick is the single most important number in this article. Its ask of 4,240.08 had already jumped to the new post-release level while the bid of 4,196.88 was still sitting at the pre-release level — one side updated, the other had not, and the 43.20 gap opened between them. This is not a platform cheating you; it is a quote desynchronisation. In the second the data lands, market makers pull their orders and then reprice, and in between there is a vacuum nobody dares to fill. Anyone who sends a market order into that vacuum fills at 4,240.08 and hands the 43.20 straight to the counterparty.

The very next tick, at 12:30:30, shows a bid of 4,224.58 and an ask of 4,225.28 — a spread of 0.70, narrower than the 1.00 baseline. So the widening lasted less than thirty seconds before returning to normal; but between 12:30 and 12:38 the wide quotes were still 5.50 to 13.60, which is 1.7× to 4.3× the baseline p90 of 3.20. The top of book came back. The depth did not.

That distinction matters a great deal for how you place orders: a narrow spread means the top level will fill; missing depth means a large order will move price hard. So a data day is not “no trading” — it is no large market orders at that moment.

6. How to place orders between 60 seconds and 5 minutes: four rules

Put those three data sets together and the rules are derived, not guessed.

Rule one: do not trade the 12:30:00 tick. The total range of the 2 October 5-minute candle was 42.50 dollars while the spread at the release instant was 43.20 — spread ÷ range = 101.6%. Even with perfect direction, price has to cover the entire candle before you break even. The expected value of that trade is negative and it has nothing to do with your analysis.

Rule two: wait 30 to 60 seconds for the book to recentre. At 12:30:30 the spread is back to 0.70, which is 17.9% of the median 5-minute range (3.90) and only 1.6% of that day’s 42.50 candle. Same trade, thirty seconds later, cost falls from 101.6% to 1.6% — the highest-return action described anywhere in this article.

Rule three: never use an expiry shorter than 300 seconds. XAUAD’s yields by expiry (/api/config, 2026-10-24 22:10 Beijing) convert into the break-even win rates below:

Break-even win rate = 1 ÷ (1 + yield). For 60 s: 1 ÷ 1.68 = 59.5%, meaning you need six correct out of ten just to break even. Yields come from the platform configuration and can change; the figure shown on the page at the time governs.

A 60-second expiry needs a 59.5% win rate just to break even, yet in the hour after the 4 September release the 5-minute candle ranges swung between 8.5 and 29.8 dollars while price oscillated between 4,412.00 and 4,481.80 — the outcome of a 60-second trade depends far more on which stretch of those sixty seconds you landed in than on the direction of NFP. Take the same judgement to 900 or 3600 seconds and the thresholds fall to 54.1% and 52.6%, at which point directional skill starts to matter.

Rule four: size so that one loss cannot affect the next trade. Cap single-trade risk at 1% of the account, with a maximum of one or two trades on a data day. Two consecutive false breakouts in opposite directions is the norm at these moments (4 September rallied first and then dumped), and adding to a loser turns a 1% risk into 4%.

7. The three most common ways to lose

One: grabbing the first tick. Watching the clock and pressing at 12:30:00 fills you at 4,240.08 when fair value was around 4,224. You are down 16 dollars from the moment of the fill, before direction is even in play.

Two: chasing the first candle. The first 5-minute candle on 4 September closed at 4,441.90 and looked like a confirmed downtrend, so people shorted it — and price then spent seventy minutes climbing back to 4,476.90. The first candle reflects who reacted first, not who is right; institutions typically finish building positions on the second or third candle.

Three: trading a data day on a single-link logic. “Strong jobs → strong dollar → gold down” is only a moderate relationship on normal days: measured over the last 252 sessions, the correlation between daily gold returns and daily dollar index returns is −0.374, explaining just 14.0% of the variance, and the two moved in the same direction on 35.7% of days. On a data day that link is weaker still, because inflation expectations, risk appetite and central bank buying are layered on top. The full correlation study is in the dollar index and real yields: how they really move gold.

8. Two releases left in 2026, and the other moments with the same mechanics

Per the BLS calendar, the remaining 2026 releases are 6 November (October data) and 4 December (November data), both on standard time and both 21:30 Beijing. After 1 November the New York session shifts an hour later as a whole, moving the overlap window to 21:00–24:00, so anyone trading data days needs to recalibrate their timetable.

Three other event types share the mechanics, all landing at 20:30 or 21:30 Beijing: CPI (same 8:30 ET slot, running through the inflation-expectations leg), the FOMC decision (02:00 or 03:00 Beijing, running through the nominal-rate leg), and PCE. Their volatility structure closely mirrors NFP — the spread explodes in the first minute, the first candle overreacts, and price mean-reverts over the following 30 to 70 minutes. The four rules above transfer almost unchanged; what is specific to the FOMC (the dot plot and the press-conference second shock) is covered separately in the Fed, FOMC meetings and gold.

If all you want to know is whether to act right now, the live bid/ask and spread on the today’s gold price page will tell you your cost more usefully than any forecast. For practice between data days, see short-term tactics and measured session volatility; both use the same range methodology as this article, so the numbers can be cross-checked.

9. FAQ

Q. What time is NFP released, and what is that in Beijing time?
A. 8:30 US Eastern, which is 20:30 Beijing during daylight time and 21:30 during standard time. Both remaining 2026 releases (6 November, 4 December) are 21:30 because the US switches on 1 November. Do not derive dates from “the first Friday” — seven of the twelve 2026 dates do not fit.

Q. Does gold always rise or fall after NFP?
A. There is no fixed direction. Five minutes after the 2026-09-04 release gold was down 75.80 dollars (−1.68%); five minutes after 2026-10-02 it was up 36.60 (+0.87%). Direction is set by actual minus expected, and it frequently reverses within 70 minutes: the 4 September low was 4,412.00 and by 13:30 UTC price was back at 4,461.00.

Q. Can you trade NFP on a 60-second expiry?
A. Not between 12:30:00 and 12:31:00. On 2026-10-02 the spread at the release instant was 43.20 dollars while the whole 5-minute candle ranged only 42.50, so the spread was 101.6% of the range — perfect direction merely breaks even. Thirty seconds later the spread is back at 0.70, or 1.6% of the range, and only then is the cost structure workable.

Q. Does the spread stay wide all day on an NFP day?
A. No. Measured on 2026-10-02: 43.20 at 12:30:00, back to 0.70 by 12:30:30 — narrower than the 1.00 baseline median. But wide quotes between 12:30 and 12:38 were still 5.50 to 13.60, or 1.7× to 4.3× the baseline p90 of 3.20. Depth recovers more slowly than the spread.

Q. Which moves gold more, NFP or CPI?
A. They work through different mechanisms, so a straight comparison is misleading. NFP rewrites both growth expectations (moving the dollar) and inflation expectations (moving real yields), while CPI mainly hits the real-yield channel. Decide which of the two you are actually trading before choosing which to watch; the CPI read-through is broken down step by step in the CPI and gold article.