Part 3: The week and the news · Chapter 8 of 14 · 15 min
US data releases
At 08:30 New York time the US data come out and the five-minute candle is worth a median 1.5 times those of the previous hour and a half, but on 7% of days (15 out of 210) it is worth at least 5 times. The first Friday of the month does so in 4 cases out of 9, against 2 out of 32 on other Fridays (p = 0.009). After the first candle the effect lasts half an hour, and in the following hour the price tends to give back part of the move (but there are only 14 cases).
Why gold watches US data
Gold pays no interest. When real yields (the yields of US government bonds net of inflation) rise, holding gold costs more in lost yield, and the price tends to fall; when they fall, it tends to rise. The same goes for the dollar: a strong dollar makes gold more expensive for anyone buying in other currencies. US data move both, because they change expectations about what the Fed will do (its rate decisions: chapter 9).
The mechanism in one line: data stronger than expected suggest higher rates, and usually push the dollar and yields up and gold down; weaker data do the opposite. It is a rule, not a law: there are days when gold rises on strong data because the market feared worse, and days when weak data push it down because they bring fear about the economy and generalised selling.
The point beginners skip is this: the price moves on the surprise, not on the number. Before a release analysts publish a forecast (the "consensus"). If inflation comes out at 3.0% and the consensus was 3.0%, the price often moves little: it was already priced in. If it comes out at 3.3%, that three-tenths difference moves the market. Reading a number without knowing the consensus and the previous value tells you nothing about what the price will do.
The data that move gold most often come out almost all at 08:30 New York time (14:30 in Rome, 13:30 in the weeks when daylight saving is not aligned, chapter 1). At 10:00 second-tier data come out.
| Release | What it measures | When it usually comes out |
|---|---|---|
| Employment report (often called NFP) | New jobs, unemployment, wages | First Friday of the month, 08:30 |
| Consumer inflation (CPI) | How prices paid by consumers change | Once a month, mid-month, 08:30 |
| Producer inflation (PPI) | Prices at the factory gate | Once a month, 08:30 |
| Retail sales | How much consumers spend | Once a month, 08:30 |
| Jobless claims | New layoffs, week by week | Every Thursday, 08:30 |
| GDP and consumer spending (PCE) | Growth and the Fed's preferred inflation measure | Once a month, 08:30 |
| ISM, job openings (JOLTS), consumer confidence | Activity indexes, labour market, mood | Often at 10:00 |
Dates must be checked every week on an economic calendar: they change with holidays and with the statistical offices' choices. A note on our data: we could not obtain the official release calendar in a verifiable way (the statistical office's site does not allow automated access), so in this chapter we do not attribute a candle to a specific release. We measure the time and the day of the week, which can be observed, and leave the rest to the calendar.
What we measured
For each weekday we compared the five-minute candle at 08:30 with the average of the five-minute candles of the previous two and a half hours (07:00-08:30 New York time), to account for the volatility level of that period. We call a strong day a day on which the 08:30 candle is worth at least 5 times that reference. One day is excluded: Friday 3 April 2026 (Good Friday), when the reference was 0.33 $ per candle and the market was practically at a standstill.
Over 210 days the 08:30 candle is worth a median 1.5 times the reference (mean 2.3). On 15% of days (32) it is worth at least 3 times, on 7% (15) at least 5, on 2% (5) at least 10. By day of the week:
| Day | Median ratio | Strong days (at least 5×) |
|---|---|---|
| Monday | 1.1× | 0 (0%) |
| Tuesday | 1.5× | 1 (2%) |
| Wednesday | 1.5× | 4 (10%) |
| Thursday | 1.8× | 4 (10%) |
| Friday | 1.7× | 6 (15%) |
Strong days are few: 15 in ten months, on average one every three weeks. For the rest of the mornings the 08:30 candle is only a little larger than the others. The danger of a release is this: most of the time nothing happens, and once in a while a lot does.
Read the chart in words
Daily gold chart from 3 August to 13 September 2026 with, below, the bars of the 08:30 New York candle ratio.
The first Friday of the month
Friday is the day with the most strong days (6 out of 42). In chapter 6 we saw that the 08:30 candle is the widest on Fridays. Here you see where that comes from: the employment report usually comes out on the first Friday of the month.
In the data the 9 first Fridays (day of the month from 1 to 7) have a median ratio of 4.4× (mean 7.0×) and 4 of them (44%) are strong days. The other 32 Fridays have a median ratio of 1.6× (mean 2.3×) and only 2 (6%) are strong days. The permutation test on the mean ratio gives p = 0.009: chance is hard to believe in.
The rule has exceptions, though, and the data show them: the first Friday of January (the 2nd) has a ratio of 2.0× and the strong candle of the month is on Friday 9 January (6×); the first Friday of July (the 3rd, a US holiday) has 1.7×, while on Thursday 2 July the 08:30 candle is worth 14× (65.8 $ of range). The strong candle of the month does not always fall on the first Friday, which is why the calendar must be checked and not remembered. We did not verify these two cases against the official calendar: what we know is where the strong candles fall.
The strongest days in the sample, by ratio and size of the five-minute candle: 4 September 28× (109.9 $), 14 July 17× (74.8 $), 2 July 14× (65.8 $), 11 September 14× (68.0 $), 12 August 14× (56.3 $). The 4 September one covered 110 $ in five minutes, more than the range of an average day (98 $).
The shape of the reaction
The five-minute profile, on the 15 strong days, is clear-cut (in multiples of the overall median candle, 5.37 $):
| Time (New York) | 08:25 | 08:30 | 08:35 | 08:40 | 08:45 | 08:50 | 09:00 | 09:15 | 09:30 |
|---|---|---|---|---|---|---|---|---|---|
| Strong days | 2.0× | 9.2× | 2.8× | 2.2× | 2.1× | 2.0× | 2.6× | 1.6× | 2.1× |
| All days | 1.3× | 2.3× | 1.5× | 1.4× | 1.4× | 1.3× | 2.0× | 1.4× | 2.2× |
The peak is a single candle. In blocks, on strong days the mean of the five-minute candles is 4.73× between 08:30 and 08:45, 2.01× between 08:45 and 09:00 and 1.71× between 09:00 and 09:30 (on other days 1.50×, 1.30× and 1.49×). So the effect lasts half an hour: after 09:00 the market is back to normal (the rise at 09:00 and 09:30 is there on normal days too: at 09:30 the New York stock market opens). The 08:30 candle occupies a median 88% of the range of the hour 08:30-09:30: most of the movement happens at once, and whoever arrives later finds a different situation.
The strong candle is wide (median range 42.2 $) but has a solid body (the body is 67% of the range, as a median): the price starts in one direction and mostly stays there at least until the five minutes close.
After the first move
The question everyone asks: after the first candle, does the price continue or come back? In the 14 strong days with a first move of at least 1 $, the net move of 08:35-09:25 goes in the same direction as the first candle in 4 cases (29%, ± 24 points) and in the opposite direction in 10. In 5 cases (36%, ± 25) the price returns to touch the opening level of the 08:30 candle, that is it cancels the whole move. Measured as a share of the first move, the next hour's move is a median −0.23 (the mean is −0.24 ± 1.00): as a median the price gives back about a quarter of the first candle.
A control: all other five-minute candles of the same size (body of at least 23.3 $) outside release times (137 cases): the move afterwards goes in the same direction in 65 cases (47%) and returns to touch the open in 62 (45%). Almost a coin. After the data, in 14 cases, there is a hint of a return, but it is not proven: the mean is inside the error.
Read the chart in words
H1 gold chart of 4 September 2026 with the moment of the 08:30 New York data and two lines: the level before the release and the one at 09:30.
| When | What | Price |
|---|---|---|
| 4 Sept 2026 12:00 UTC | 08:30 data (New York) | 4,482.23 |
| 4 Sept 2026 11:00 UTC | before the data 4478.2 | 4,478.17 |
| 4 Sept 2026 12:00 UTC | at 09:30 4416.5 | 4,416.49 |
Times in UTC.
Read the chart in words
H1 gold chart of 2 October 2026 with the moment of the 08:30 New York data and two lines: the level before the release and the one at 09:30.
| When | What | Price |
|---|---|---|
| 2 Oct 2026 12:00 UTC | 08:30 data (New York) | 4,228.70 |
| 2 Oct 2026 11:00 UTC | before the data 4189.8 | 4,189.75 |
| 2 Oct 2026 12:00 UTC | at 09:30 4197.4 | 4,197.42 |
Times in UTC.
The two cases are what you would expect after a candle of 40-110 $: on 4 September the first move (−78.9 $) largely holds (in the following hour the price recovers 17.2 $); on 2 October the +35.5 $ move is largely given back (−27.9 $ in the following hour, with the price returning to touch the starting level). In 14 days both stories appear, and such a small sample does not say which one wins: in chapter 11 we will use it as the hypothesis of a strategy (going against the first move) and test it with a control.
The 10:00 data
The 10:00 data (ISM, job openings, consumer confidence) move much less: the 10:00 candle is worth a median 1.9× the 07:00-08:30 reference, and exceeds 5× on only 3 days out of 210 (1%); 38 days (18%) exceed 3×. On the first days of the month (calendar days 1-3, 21 cases) the median ratio is 2.1×: once a month, a few more days, but nothing like the employment report.
Read the chart in words
H1 gold chart of 7 August 2026 with four candles marked with the letters A, B, C and D.
| When | What | Price |
|---|---|---|
| 7 Aug 2026 11:00 UTC | Point A | 4,332.55 |
| 7 Aug 2026 12:00 UTC | Point B | 4,373.38 |
| 7 Aug 2026 14:00 UTC | Point C | 4,355.54 |
| 7 Aug 2026 17:00 UTC | Point D | 4,346.30 |
Times in UTC.
What to do with it
- Check the calendar every week. Strong days are 15 out of 210 and they are predictable: they are the days of the important releases. Write down the time and day of the release and the consensus.
- Before the release, do not enter and do not leave tight stops. The 08:30 candle is 9.2× the median on strong days, 42 $ of range as a median and up to 110 $. A 10 $ stop sits inside a single candle and is filled wherever it happens (as over the weekend). Better to stay out or reduce size.
- Do not chase the first candle. The peak is a single candle and the rest of the half hour is worth 2-3 times normal: entering on the most extreme candle means buying the worst price.
- Wait for the first five minutes to close, better the first fifteen. Then see whether the price holds or gives back: in our 14 strong days a partial return is common (median −0.23 of the first candle).
- After 09:00 the market is normal. If you missed the move there is no need to force it: in half an hour the range is back to its usual size.
- The sign of the number is not enough. The price reacts to the surprise against the consensus, and the reaction can reverse in a few candles.
Procedure in five steps
- On Monday look at the week's calendar and mark the 08:30 releases (and the 10:00 ones) with the consensus.
- A few minutes before the release, close or reduce positions with a close stop and do not open new orders.
- Let the first five-minute candle pass without trading.
- After the first fifteen minutes, judge whether the price holds the level of the move or gives it back, with sensible stops (wider than usual).
- At 09:00-09:30 the market is back to normal: resume your usual plan.
Exercises
Chapter exercises
1. On the chart the letters A-D are on four H1 candles of Friday 7 August 2026 (first Friday of the month, UTC hours). Which letter is the candle that contains the 08:30 New York data?
Read the chart in words
| When | What | Price |
|---|---|---|
| 7 Aug 2026 11:00 UTC | Point A | 4,332.55 |
| 7 Aug 2026 12:00 UTC | Point B | 4,373.38 |
| 7 Aug 2026 14:00 UTC | Point C | 4,355.54 |
| 7 Aug 2026 17:00 UTC | Point D | 4,346.30 |
Times in UTC.
2. In the data the first Fridays of the month are strong days (08:30 candle at least 5 times the reference) in 4 cases out of 9, against 2 out of 32 on other Fridays (p = 0.009). What do you take from it for your work?
3. In the 14 strong days the price, in the hour after the first candle, goes in the opposite direction in 10 cases; at equal size outside the data the same happens in 53% of cases. How do you treat it?
4. The five-minute candle of strong data has a median range of 42 $ and reaches 110 $. You hold a position with a 10 $ stop at the time of the release. What usually happens?
Have a question or want to share your exercise?
Post in the community, or join the free signals room on Telegram.
Educational content, not financial advice. Trading involves risk.