Two first sweeps from five, the weakest gold week on the record, and both sides landed on a 4% rating

The week at a glance

Every rating in this table is transcribed from the close capture of that morning; the Modal column is the largest of the four shares and the High and Low columns are the two marginals summed from them, as the note explains.

Four-way reads high only / low only / both / inside and always sums to 100. The ~ on Monday's cell is the indicator's era flag: the cell's shares spread more than ten points across market eras, so they are quoted as approximate. High and Low carry the marginal probability that side is taken at all, which is that side's exclusive share plus the both-sides share, not the number printed on the level itself. These captures are 5-minute charts, so an outcome time is the bar in which the level was crossed and is good to the bar. Two red folder USD releases fell inside the session: Core PCE with the preliminary GDP print at 08:30 on Wednesday, and Fed Chair Warsh at Jackson Hole with the preliminary benchmark payrolls revision at 10:00 on Friday. Both are marked against the day.

Morning by morning

Open capture on the left, close capture on the right. Click any chart to enlarge.

Monday 24 August

Wednesday 12 August's era-flagged cell again, and this time it went the way the table leaned

First sweep ✓ · modal branch ✓ · high only, as rated

~55 / ~22 / ~10 / ~13, first sweep 72 / 28, out of 640 sessions: the cell Wednesday 12 August drew, the one carrying the tilde. The tilde means the cell's shares spread more than ten points across market eras, so the indicator quotes them as approximate rather than to the point. On the 12th the session took both sides. This time it took the high at ~09:00 and nothing else.

Eurasia had printed 44.5 points, above average, and swept its predecessor's high, topping out on a spike at 07:45 that left the open a few points above EQ. US AM climbed from the first bar, crossed the high on the 09:00 bar, and then sat five to ten points above it until the close.

Both marginals resolved as stated: the high at 65% went, the low at 32% did not. The same lookup on two mornings has now produced both sides once and high only once, which is one draw more than the record had on this cell a fortnight ago and no more informative than that.

• Eurasia range44.5 pt, above average
• Prior sweepSwept high
• Open vs Eurasia EQAbove EQ
• Four-way~55 / ~22 / ~10 / ~13
• First sweep▲ 72% · ▼ 28%
• P(high taken)65% → taken ~09:00
• P(low taken)32% → not taken
• Samplen 640, High

Tuesday 25 August

Tuesday 18 August's cell, the same 65% call on the high, and the low went instead

First sweep ✗ · modal branch ✗ · low only, rated 24%

45 / 24 / 27 / 4, first sweep 65 / 35, out of 729 sessions: the table Tuesday 18 August drew, when a single 09:30 bar took the low and the high went ten minutes later. This time the session sold off from the open, took the low on the 09:15 bar, bottomed near 4660 at 10:00 and climbed back inside the range to 4696 by the close.

Two draws from one cell and two different answers: both sides on the 18th, low only on the 25th. What the two mornings share is that the 65% call on the high going first was wrong on both of them. Eurasia was 33.3 points and inside its predecessor, so the levels were not far away; the high sat seventeen points above the open and was never reached.

Per side, one from two: the low at 51% went and the high at 72% did not. The week's first double miss, on a cell that has now missed its first-sweep call on both appearances. Two is not a pattern, and it is written down so that the third draw, if it comes, has something to be compared with.

• Eurasia range33.3 pt, below average
• Prior sweepInside
• Open vs Eurasia EQAbove EQ
• Four-way45 / 24 / 27 / 4
• First sweep▲ 65% · ▼ 35%
• P(high taken)72% → not taken
• P(low taken)51% → taken ~09:15
• Samplen 729, High

Wednesday 26 August

The low on the PCE bar, thirty minutes in, and nothing else for the rest of the morning

First sweep ✓ · modal branch ✓ · low only, as rated🔴 Core PCE + GDP · 08:30 ET

Core PCE and the preliminary GDP print landed at 08:30 ET, thirty minutes into the session. The table had the low at 45% and 62% on it going first, after a Eurasia session that had spent the whole night falling and swept its predecessor's low. The 08:30 bar took the Eurasia low.

After that the session did very little for an hour, held a few points either side of the level, bounced to 4688 near 10:00 and then slid to 4640 into the close. The high at 4706.5 was never within fifteen points of price after the first bar.

Both marginals resolved as stated: the low at 59% went and the high at 43% did not. Low only was the modal branch and the first-sweep call agreed with it, so the morning scores on all three axes. The release sits on the day because it fell inside the session, and the level it took was the one fourteen points from the open on the side the night had already been moving.

• Eurasia range37.7 pt, below average
• Prior sweepSwept low
• Open vs Eurasia EQBelow EQ
• Four-way29 / 45 / 14 / 12
• First sweep▲ 38% · ▼ 62%
• P(high taken)43% → not taken
• P(low taken)59% → taken ~08:30
• Samplen 649, High

Thursday 27 August

Thursday 13 August's cell, a 70% call on the high, and the low went at 09:45, exactly as it did then

First sweep ✗ · modal branch ✗ · low only, rated 19%

48 / 19 / 4 / 29, first sweep 70 / 30, out of 573 sessions: Thursday 13 August's table. That morning the session took the low at 10:30 and nothing else, against a 70% call on the high. This morning it took the low at ~09:45 and nothing else, against the same 70% call on the high. Same cell, same outcome, same two misses.

Eurasia had printed 52.6 points, above average, and the wide range is why inside carried 29%, level with the 13th and the highest the gold record had shown until the next morning. The session did not stay inside either. It drifted down from the open, crossed the low on the 09:45 bar, bottomed near 4624 and recovered to 4647 by the close.

Per side, one from two, and the wrong one: the low at 23% went and the high at 52% did not. Two draws of this cell have now produced two low-only mornings on a 19% branch. That is the kind of thing a running record is for, and also the kind of thing that two draws cannot tell you anything about.

• Eurasia range52.6 pt, above average
• Prior sweepSwept low
• Open vs Eurasia EQAbove EQ
• Four-way48 / 19 / 4 / 29
• First sweep▲ 70% · ▼ 30%
• P(high taken)52% → not taken
• P(low taken)23% → taken ~09:45
• Samplen 573, High

Friday 28 August

Inside rated 33%, the highest yet, and the session took both levels ten minutes apart around 10:00

First sweep ✗ · modal branch ✗ · both sides, rated 4%🔴 Warsh (Jackson Hole) + payrolls revision · 10:00 ET

Eurasia had swept both of its predecessor's levels and printed 44.3 points, and the table put 33% on the session staying inside, the highest inside share the gold record has shown, above Friday 21 August's 26%. The low carried 40% and 64% on going first. Both sides was rated 4%, the smallest share on the table.

For ninety minutes the session held between 4645 and 4665. Then it climbed to the high on the 09:55 bar, Fed Chair Warsh spoke at Jackson Hole at 10:00 alongside the preliminary benchmark payrolls revision, and the 10:00 bar fell from 4665 to 4630. The low went on the 10:05 bar, price reached about 4580 by 10:15 and recovered to 4619 into the close.

Every axis missed. The first sweep was called low at 64% and the high went first, by ten minutes. The modal branch was low only and both sides landed. Per side, the high at 27% went and the low at 44% went, two from two on the marginals, on a morning where the four-way rated the pair together at 4%. A 4% branch landing is what a 4% branch does about one morning in twenty-five; this page records that it was this one.

• Eurasia range44.3 pt, above average
• Prior sweepSwept both
• Open vs Eurasia EQBelow EQ
• Four-way23 / 40 / 4 / 33
• First sweep▲ 36% · ▼ 64%
• P(high taken)27% → taken ~09:55
• P(low taken)44% → taken ~10:05
• Samplen 587, High

What the week actually showed

• Two first sweeps from five, the weakest gold week on the record, and all three misses were calls of 64 to 70%. Tuesday's 65% on the high, Thursday's 70% on the high and Friday's 64% on the low all went the other way. Over three gold weeks the first-sweep axis stands at ten from fifteen, which is a lower number than the two previous weeks left it at and is published for exactly that reason.
• Three repeat cells, and one of them resolved exactly as it did the first time. Thursday drew Thursday 13 August's table and went low only again, against the same 70% call on the high. Tuesday drew Tuesday 18 August's table and missed the same 65% call on the high again, low only this time where the 18th was both sides. Monday drew Wednesday 12 August's era-flagged cell and went high only where the 12th went both. The forecast is a lookup, and a lookup that repeats an outcome twice has told you it repeated, nothing more.
• Both sides landed once, on the smallest share of the week. Friday rated the pair together at 4% and took the high at ~09:55 and the low at ~10:05, either side of the 10:00 Warsh speech. Across fifteen gold sessions the both-sides branch has now landed five times against a mean rating of about 13%. That remains the number this record exists to watch, and fifteen sessions is still far too few to say whether it is a run or a miscalibration.
• Levels went more often than stated for the third week running, by less. Six of ten against a mean stated 47%, after eight of ten against 54% and six of ten against 48%: twenty of thirty against a mean stated of about 50% over three weeks. Two thirds realised against a half stated is a gap of about one and a half standard errors on thirty events, which is recorded because it belongs in the record and not because it means anything yet.
• Nothing stayed inside, on the two highest inside ratings the gold record has shown. Thursday carried 29% and Friday 33%, both on above-average Eurasia ranges, and both sessions took a level. Across fifteen gold sessions the inside branch has averaged roughly 13% and landed zero times. The wide-prior-session mechanism that pushes the inside share up has not yet produced a contained morning on gold, which at fifteen draws is not yet surprising.