Nine of the ten windows declined to call, the most on the gold record, and the one that called was right

The week at a glance

Every number here is transcribed from the capture of that window. Times are read off the 5-minute bars, so an outcome time is the bar in which the rail was crossed.

Regime is the table's tag, which measures this box against the whole history; the tag printed on the box itself measures it against the last five sessions, so the two legitimately differ. They are the same three buckets named twice, compressed against accumulation, expanded against expansion, normal against normal. Seven of this week's ten boxes land in different buckets on the two rulers, against none at all on the week of 7 September, and that is a fact about the preceding five sessions being unusually narrow rather than about the boxes. Friday's hour is the sharpest case: 29.1 points, compressed against the whole history and expanding against the last five sessions. Mult is an output of the table's regime, never a setting, and every window here reads auto. Monday's two windows were tagged normal and drew a half multiple; the other eight were tagged compressed and drew a full one. Nothing was touched between Monday and Tuesday. The three-way shares do not all sum to 100, because each is rounded to a whole percent: Tuesday's hour and Friday's hour sum to 99, Wednesday's hour and Friday's half hour to 101. Charts are 5-minute, so an outcome time is the bar in which the rail was crossed and is good to the bar; the session close is the 16:55 ET bar on all five days. The tick and cross in the outcome column mark whether the favoured rail was the one that went, for calls and abstains alike; only the one Atlas row counts towards the Atlas score. Friday's half hour reached its lower rail first and its upper rail three hours and forty minutes later; only the first is scored.

Morning by morning

The 30 minute window on the left, the 60 minute on the right. Click any chart to enlarge.

Monday 14 September

The only Atlas call of the gold week, and the 11:30 bar paid it

30m Atlas mode ✓ · 60m abstained and its favoured rail landed

The two boxes share a low. Gold put in 4293.2 on the 09:05 bar and never went back to it inside the window, so the half hour and the hour have the same floor at 4293.2 and differ only at the top: 4317.8 by 09:25, 4321.7 by 09:55. That makes the hour the wider of the two at 28.5 points against 24.6, and it is why the hour's rails sit further out.

The half hour is the only window of the week Atlas mode called. It read 63 / 28 / 9 from 306 sessions, 69 percent of the decided share against a 62 percent gate and 91 percent of comparable sessions resolving against a 50 percent gate, so it fired long with a 12.3 point target. Price fell to 4303.5 on the 10:20 bar, about ten points above the box low and nowhere near the lower rail at 4280.90, with no bar after the window trading below the box low at all, then climbed and the 11:30 bar put its high of 4335.2 through 4330.10. The same cell, 63 / 28 / 9 out of the same 306 sessions, came up once before, on Friday 14 August. It fired long then and the upper rail went then too. Two draws of a cell is two draws and separates nothing.

The hour missed the gate by a point. 41 / 27 / 32 gives 61 percent of the decided share where the gate is 62, so it abstained even though 68 percent of comparable sessions resolved and the up side was the larger share. Five minutes after the half hour's mark the 11:35 bar took the hour's rail at 4336.00 on a high of 4338.8. One move, two windows, and the mark on the hour is bare because no call was made. That cell, 41 / 27 / 32 from 290 sessions, also drew on Tuesday 18 August, where it abstained the same way and the lower rail went instead.

• 30m box24.6 pt · normal
• 60m box28.5 pt · normal
• 30m three-way63 / 28 / 9
• 60m three-way41 / 27 / 32
• 30m modeAtlas, long
• 60m modeabstain, direction gate at 61
• Outcome30m up rail 11:30 · 60m up rail 11:35
• Durationsagreed

Tuesday 15 September

The largest unresolved share the gold tables have printed, and the hour duly resolved nothing

30m abstained and the other rail landed · 60m abstained, no target hit

The hour's 62 percent unresolved share is the largest the gold tables have printed on this record, ahead of the four 61 percent readings behind it, and the window went on to resolve nothing. That is one draw of a 62 percent branch and it proves nothing about the branch. It is recorded because it happened, and because a reader looking at a 25 percent up figure should understand that the largest of the three numbers on that label was the one that says neither rail.

Both of Tuesday's windows abstained, and they failed on different gates. The half hour was 35 / 43 / 22 from 270 sessions: 78 percent of comparable sessions resolved, comfortably clear, but the down side was 55 percent of the decided share against a 62 percent gate, so the direction gate stopped it. The hour was 25 / 12 / 62 from 326 sessions: 68 percent of the decided share, clear of the direction gate, but only 38 percent decided against a 50 percent gate. One window was too close to call, the other had too little to call it on.

The half hour then did both things. It ran down to 4313.1 on the 10:50 bar, which is 2.3 points above the lower rail at 4310.80 and the closest either Tuesday window came to the side it leaned towards, and then spent three hours and twenty-five minutes climbing until the 14:15 bar took the upper rail at 4346.50. The scorer records the first rail reached and nothing else, so this counts against the lean. There is no credit for 2.3 points.

• 30m box11.9 pt · compressed
• 60m box15.8 pt · compressed
• 30m three-way35 / 43 / 22
• 60m three-way25 / 12 / 62
• 30m modeabstain, direction gate
• 60m modeabstain, decided gate
• Outcome30m up rail 14:15 · 60m no target
• Durationsdisagreed

Wednesday 16 September

Both rails that went were lower ones, four hours apart, and the second sits inside the FOMC block

30m abstained and the other rail landed · 60m abstained and its favoured rail landed🔴 FOMC rate decision, projections and statement 14:00 ET · press conference 14:30

The week's only red folder landed here: the Federal Funds Rate, the FOMC Economic Projections and the FOMC Statement at 14:00 ET, with the press conference at 14:30. Every window on this page opens at 09:00, so no box on the record contains the release. One scored outcome sits inside the block, the hour's lower rail on the 14:10 bar, and it is the only one on the page that does. This page is not claiming the release caused that move, and the other four days have nothing to do with it.

The half hour resolved long before any of that. 42 / 36 / 22 from 287 sessions is 54 percent of the decided share against a 62 percent gate, close enough to a coin that the table declined to call it, and the 10:05 bar put its low of 4365.0 through the lower rail at 4367.80. Thirty-five minutes after the box closed, against the larger of the two shares, and counted as a miss for the lean.

The hour read 12 / 31 / 58 from 320 sessions. The down side was 73 percent of the decided share, well clear of the direction gate, but only 42 percent of comparable sessions resolved at all against a 50 percent gate, so it abstained on the decided gate with the down side as the larger share. Price sat inside the rails for four hours and ten minutes. The 14:00 bar then covered 4414.1 down to 4355.4, a little under 59 points on one 5-minute bar, stopping 4.8 points above the rail, and the 14:10 bar took it at 4348.8. The session low of 4273.1 came later still, on the 15:10 bar. Both durations ended on their lower rails while favouring opposite sides, which is the arithmetic of a coin-flip half hour and a one-sided hour, not agreement between them.

• 30m box17.0 pt · compressed
• 60m box25.6 pt · compressed
• 30m three-way42 / 36 / 22
• 60m three-way12 / 31 / 58
• 30m modeabstain, direction gate
• 60m modeabstain, decided gate
• Outcome30m down rail 10:05 · 60m down rail 14:10
• Durationsdisagreed

Thursday 17 September

The one window of the week that failed both gates, and the slowest resolution on the page

30m abstained and its favoured rail landed · 60m abstained, no target hit

The hour is the only window of the ten that failed both gates. It read 25 / 17 / 58 from 244 sessions: the up side is 59 percent of the decided share where the gate is 62, and only 42 percent of comparable sessions resolved where the gate is 50. Too close to call and too little resolved to call it on, at once. Price then spent the afternoon walking down and stopped 3.6 points above the lower rail on the 16:30 bar. That same cell, 25 / 17 / 58 out of 244, drew on Monday 7 September and reached neither rail then either, although that Monday was a shortened Labor Day session, so the two are not measured over the same number of hours.

The half hour drew 35 / 43 / 22 from 270 sessions, which is the cell Tuesday's half hour drew, out of the same sample, in the same week. Both abstained on the direction gate at 55 percent of the decided share. Tuesday the upper rail went; Thursday the lower one did. Two draws of one cell, opposite outcomes, four days apart. This is the clearest thing on the page about what a stated share is and is not.

Thursday's lower rail at 4385.50 went on the 15:25 bar, on a low of 4382.3, five hours and fifty-five minutes after the box closed. That is the slowest resolution of the week, against twenty minutes for the fastest on Friday. Nothing on this page rewards speed, and nothing penalises it either, but it is worth a reader knowing that a window scored as reached can sit unresolved through an entire session first.

• 30m box15.1 pt · compressed
• 60m box20.0 pt · compressed
• 30m three-way35 / 43 / 22
• 60m three-way25 / 17 / 58
• 30m modeabstain, direction gate
• 60m modeabstain, both gates
• Outcome30m down rail 15:25 · 60m no target
• Durationsdisagreed

Friday 18 September

An 8 percent up share, the smallest on the record, and the hour still could not call it

30m abstained and its favoured rail landed · 60m abstained, no target hit

The hour's 8 percent up share is the smallest single direction share on the gold record, and the 81 percent of the decided share it implies is the strongest direction reading of the week, comfortably over the 62 percent gate. It abstained anyway, because only 44 percent of comparable sessions resolved against a 50 percent gate. Strong lean, thin base. The two gates are there so that one of them cannot carry a call on its own, and this is what that looks like on a chart.

Friday's hour is also the one box on the page where the two rulers land two buckets apart, rather than the one bucket that separates them on the other six that differ. At 29.1 points it is the widest window of the week and it is expanding against a 22.7 point five session average, while the table, which measures it against the whole history, calls it compressed. Both are true. Gold's previous five sessions were narrow enough that a below-average box by the long ruler is an above-average one by the short.

The half hour needed twenty minutes. 34 / 44 / 23 from 278 sessions, 57 percent of the decided share, abstained on the direction gate, and the 09:50 bar put its low of 4385.8 through the lower rail at 4387.60. Price then turned and reached the upper rail at 4428.10 as well, on the 13:30 bar, three hours and forty minutes later. The scorer records only the rail that goes first, which is the lower one, so this counts for the lean. A window that eventually touches both rails is not a stronger result than one that touches one, and the page notes it so nobody reads the outcome column as the whole day.

• 30m box13.5 pt · compressed
• 60m box29.1 pt · compressed
• 30m three-way34 / 44 / 23
• 60m three-way8 / 35 / 56
• 30m modeabstain, direction gate
• 60m modeabstain, decided gate
• Outcome30m down rail 09:50 · 60m no target
• Durationsagreed

What the week actually showed

• Nine of the ten windows abstained, which is the most of any gold week published here. The previous high was five, on three separate weeks, and the low was two on the week of 31 August. The nine split cleanly by cause: five were stopped by the direction gate alone (Monday's hour at 61 percent of the decided share, then all four of Tuesday, Wednesday, Thursday and Friday's half hours between 54 and 57), three by the decided gate alone (Tuesday, Wednesday and Friday's hours, at 38, 42 and 44 percent decided), and one, Thursday's hour, by both at once. The mean stated favourite across the ten was 39, the lowest of the six weeks and well under the 49 the record averages. A week of weak, undecided tables produced a week of silence, which is what the gates are for.
• The one call landed. Monday's half hour read 63 / 28 / 9 out of 306 sessions, cleared both gates at 69 and 91, fired long with a 12.3 point target and the 11:30 bar took the upper rail at 4330.10. That is one race. It is the same cell that fired long on Friday 14 August and took the upper rail then too, which is two races. Nothing in this paragraph distinguishes a 63 percent branch from a 50 percent one and nothing was ever going to.
• Seven of ten reached a rail, and all three that reached nothing were hours. Tuesday's, Thursday's and Friday's 60 minute windows were unresolved at the 16:55 bar, and every half hour on the week resolved, the slowest of them at 15:25 on Thursday and the fastest at 09:50 on Friday. The three unresolved hours carried unresolved branches of 62, 58 and 56 percent, which are three of the four largest on the page: Wednesday's hour also carried 58 and reached its rail. Seven of ten is neither the best week here nor the worst: three weeks came in at nine and one at five.
• No box on the week was expanded on the table's ruler, a first in six weeks. Eight of the ten were compressed and the other two, Monday's pair, were normal. Every previous week has carried at least one expanded box and the week of 31 August carried six. Because the multiplier follows the regime, that is also why all ten targets ran at a full or half multiple and none at a quarter. Read that as a description of gold's range over these five sessions, not as a property of the indicator.
• One red folder block, one result inside it. The FOMC rate decision, the economic projections and the statement at 14:00 ET on Wednesday, and the press conference at 14:30, are the only high impact USD entries of the week, and every window on this page opens at 09:00 and closes by 10:00. Wednesday's hour reached its lower rail on the 14:10 bar, inside that block, and that is the only scored result on the page that sits in a news window. Wednesday's half hour had already resolved at 10:05, four hours earlier. No claim is being made here about cause.
• One cell repeated inside the week itself and three repeated across weeks. The intra-week draw is Tuesday's and Thursday's half hours, both 35 / 43 / 22 out of 270 sessions, both abstaining on the direction gate at 55 percent, and the upper rail went on Tuesday while the lower went on Thursday. The three across weeks: Monday's half hour, 63 / 28 / 9 out of 306, drew on 14 August and fired long into the upper rail both times; Monday's hour, 41 / 27 / 32 out of 290, drew on 18 August and took the lower rail then against the upper rail now; Thursday's hour, 25 / 17 / 58 out of 244, drew on 7 September and reached nothing both times. Four repeats, two of which agree with themselves and two of which do not. That is exactly what a shallow sample of a wide distribution looks like.
• Sixty windows over six weeks: forty-seven reached a rail, the favoured rail won thirty of those, Atlas mode fired thirty times and was right on sixteen of the twenty-four that resolved. These are published because the record has to accumulate somewhere, and sixty races remain far too few to mean anything. The place to judge calibration is still the indicator's own realised versus expected rows, which this week ran 69% to 74% against a 66% to 71% expectation on the general hit rate, every z between +0.3 and +1.0, and 65% to 76% against a 72% to 75% expectation on the Atlas rate, every z between -0.9 and +0.5.
• Gold and the index book behaved nothing alike this week. On the same five mornings the NQ tables fired Atlas mode on eight of their ten windows; the gold tables fired on one. Two instruments, two window times, five sessions and ten windows each, and the only thing that follows from it is that the gates were reading very different tables. It is recorded because both pages are published from the same week.