Which side of gold's first hour breaks first?
Gold, one-minute bars, every day from 2008-01-29 to 2026-08-14 with all sixty minutes of the 8:20 to 9:20 hour. The hour's high and low are the range; the end that printed first is the one that formed first; the 9:20 close sits in the upper or lower half of the range. Then the bars from 9:20 to 3:59: the first end price trades through is the one that broke first. A touch is not a break.
What the file says
Most of it is where price sits at 9:20, not which end formed first. When the low formed first, the high broke first on 1,492 of 2,335 days, 63.9%. When the 9:20 close sat in the upper half of the range, whichever end formed first, the high broke first on 1,664 of 2,401, 69.3%. Given a low-first hour, price was in the upper half at 9:20 on 79.4% of days, so the first number is mostly the second one. Inside the upper half, low-first adds 7.0 points over high-first (70.9% against 63.9%); formation order is not nothing, it is the smaller of the two. The mirror is the same: high first, low breaks first 63.5%.
Every day, by what formed first and where price sat
Counts and shares. The rows in white are the ones the finding is read from.
| the hour | days | high broke first | low broke first | neither by 3:59 |
|---|---|---|---|---|
| low formed first | 2,335 | 1,492 · 63.9% | 759 · 32.5% | 84 · 3.6% |
| high formed first | 2,402 | 780 · 32.5% | 1,526 · 63.5% | 96 · 4.0% |
| 9:20 close in the upper half, any order | 2,401 | 1,664 · 69.3% | 646 · 26.9% | 91 · 3.8% |
| 9:20 close in the lower half, any order | 2,316 | 600 · 25.9% | 1,628 · 70.3% | 88 · 3.8% |
| 9:20 close at the middle exactly | 20 | 8 · 40.0% | 11 · 55.0% | 1 · 5.0% |
| low first and upper half | 1,853 | 1,314 · 70.9% | 464 · 25.0% | 75 · 4.0% |
| low first and lower half | 472 | 175 · 37.1% | 288 · 61.0% | 9 · 1.9% |
| high first and upper half | 548 | 350 · 63.9% | 182 · 33.2% | 16 · 2.9% |
| high first and lower half | 1,844 | 425 · 23.0% | 1,340 · 72.7% | 79 · 4.3% |
The same, on the last three years only
2023-09-06 to 2026-08-14, 757 days: the window a shared version of this stat was cut on.
| the hour | days | high broke first | low broke first | neither by 3:59 |
|---|---|---|---|---|
| low formed first | 370 | 251 · 67.8% | 110 · 29.7% | 9 · 2.4% |
| high formed first | 387 | 138 · 35.7% | 238 · 61.5% | 11 · 2.8% |
| upper half at 9:20, any order | 394 | 295 · 74.9% | 89 · 22.6% | 10 · 2.5% |
| lower half at 9:20, any order | 361 | 93 · 25.8% | 258 · 71.5% | 10 · 2.8% |
By year
Low formed first, the high broke first.
| year | days read | excluded | low formed first | high broke first | share |
|---|---|---|---|---|---|
| 2008 | 234 | 7 | 111 | 70 | 63.1% |
| 2009 | 249 | 8 | 121 | 66 | 54.5% |
| 2010 | 253 | 5 | 121 | 76 | 62.8% |
| 2011 | 254 | 4 | 129 | 79 | 61.2% |
| 2012 | 253 | 5 | 133 | 90 | 67.7% |
| 2013 | 254 | 4 | 119 | 64 | 53.8% |
| 2014 | 252 | 6 | 117 | 65 | 55.6% |
| 2015 | 256 | 2 | 108 | 73 | 67.6% |
| 2016 | 258 | 0 | 132 | 82 | 62.1% |
| 2017 | 256 | 1 | 121 | 80 | 66.1% |
| 2018 | 256 | 2 | 134 | 91 | 67.9% |
| 2019 | 257 | 1 | 128 | 77 | 60.2% |
| 2020 | 259 | 0 | 147 | 106 | 72.1% |
| 2021 | 258 | 0 | 119 | 81 | 68.1% |
| 2022 | 257 | 1 | 136 | 88 | 64.7% |
| 2023 | 256 | 1 | 124 | 75 | 60.5% |
| 2024 | 258 | 1 | 127 | 87 | 68.5% |
| 2025 | 257 | 1 | 136 | 99 | 72.8% |
| 2026 | 160 | 0 | 72 | 43 | 59.7% |
By year, low-first to high-first ran from 53.8% (2013) to 72.8% (2025); every year of nineteen is above 53%.
What it does not say, and how it was done
What this does not measure. How far the break went, where a stop would have sat, or what a trade on it would have cost: "the high broke first" is a fact about a level, not about money. Set aside and counted: 38 days missing a minute of the hour, 10 with both ends on one bar, 1 with both ends broken on the same minute. The three-year window a widely shared version of this stat was cut on, 2023-09-06 to 2026-08-14, is the second table; on it the low-first number is 67.8%, and the reading is the same.
How it was measured. The measurement, its definitions and six predictions were written down before it ran (SETUP 02, September 9, 2026; five of the six held, and the miss is the finding above: position at 9:20 carries more than formation order). Script gc_ib_first_break-2026-09-09.py v0.1 on the one-minute gold file: 6,489,511 lines, 4,786 dates with a bar in the window, 4,737 read. The script's own print is the record for this page, gc_ib_first_break.txt; every number here is read from it.
Raw market data: FirstRate Data (firstratedata.com). A measurement on a file, not advice.