The same forecast, twice, resolving both ways

The week at a glance

Every number here is transcribed from the capture of that window. Nothing is inferred and nothing is computed after the fact.

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. Mult is an output of that regime, never a setting, and every window here reads auto. The three-way shares occasionally sum to 101 because each is rounded to a whole percent; four of these ten do. Red folder USD releases are marked against the day; both printed at 08:30 ET, half an hour before the window opens.

Morning by morning

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

Monday 10 August

The durations split, and the one that stayed quiet was right

60m Atlas mode ✗ · 30m abstained and its rail landed

Two windows, one morning, opposite calls. The 30 minute box came in at 11.8 points against a five-day average of 24, compressed enough that the ladder left the multiplier at full, and it leaned up at 45% against 31%. The 60 minute box, formed over the same hour, leaned down at 53%.

The 60 minute call cleared both gates comfortably, 53 of the 66% that resolve being 80% of the decided share, and Atlas mode fired short. Price went the other way, through the up rail. The 30 minute window abstained, because 45 of 76 is only 59% of its decided share and the gate wants 62, and the rail it quietly favoured is the one that landed.

It is a useful first entry in the record. The abstain is not the indicator having no view; it is the indicator declining to promote a view it does not rate highly enough, and on this morning the view it declined to promote was the correct one.

• 30m box11.8 pt · compressed
• 60m box23.4 pt · accumulation
• 30m three-way45 / 31 / 24
• 60m three-way13 / 53 / 34
• 30m modeabstain
• 60m modeAtlas, short
• Outcomeup rail, both windows
• Durationsdisagreed

Tuesday 11 August

The same 30 minute forecast as Monday, resolving the other way

Both windows abstained · favoured rail 1 of 2

The 30 minute window returned 45 / 31 / 24, out of a sample of 261 sessions. Those are Monday's numbers exactly, the same cell of the same table, because the box landed in the same place on the same two axes the lookup uses. Monday it resolved up. Tuesday it resolved down.

That is the whole argument for publishing these pages. A 45% call is not a prediction that something will happen; it is a statement that out of 261 comparable mornings, this is how they divided. Two of them landed in the same week and split.

The 60 minute window is the more instructive abstain of the two so far. Its direction was strong, 28 of the 40% that resolve being 70% of the decided share, comfortably over the 62 the gate wants. It abstained anyway, because only 40% of comparable sessions reached a rail at all and the second gate wants 50. The rail was reached in the end. The gate was not wrong to be cautious about whether it would be.

• 30m box11.5 pt · compressed
• 60m box20.4 pt · accumulation
• 30m three-way45 / 31 / 24
• 60m three-way12 / 28 / 61
• 30m modeabstain, direction gate
• 60m modeabstain, resolution gate
• Outcomedown rail, both windows
• Durationsdisagreed

Wednesday 12 August

CPI expanded both boxes, and the ladder cut the multiplier to a quarter

30m Atlas mode ✓ · 60m abstained, no target hit🔴 CPI · 08:30 ET

Core CPI and CPI printed at 08:30 ET, half an hour before the window opens. So the box does not measure the spike; it measures the hour of repricing after it, which is exactly what it should do. Both boxes came in wide: 28.3 points on the 30 minute against a 17.9 point five-day average, and 34.7 on the 60 minute, the widest single box of the week.

This is the auto ladder's whole purpose and it is worth watching it work. A box tagged expanded gets the multiplier cut, so the 30 minute window's target came down to 7.1 points rather than the 28 a full multiple would have demanded. At that distance only 2% of comparable sessions fail to resolve, the direction call was a clean 76% short, and it paid.

The 60 minute window took the other branch. Its multiplier only stepped to a half, its target sat 17.3 points out, its direction was an unconvincing 41 against 29, and it abstained. Price spent the rest of the day inside the box and reached neither rail. Both windows read the same expansion; only one of them found a distance it could resolve at.

• 30m box28.3 pt · expanded
• 60m box34.7 pt · expansion
• 30m three-way22 / 76 / 2
• 60m three-way41 / 29 / 31
• 30m target7.1 pt at ×0.25
• 60m target17.3 pt at ×0.5
• Outcome30m down rail · 60m no target
• Durationsdisagreed

Thursday 13 August

The only morning both windows agreed, both fired, and both paid

Both durations Atlas mode ✓🔴 PPI · 08:30 ET

Core PPI and PPI at 08:30 ET, and unlike Wednesday the boxes came in ordinary: 18.4 and 31.9 points, both tagged normal against their five-day averages. Not every release expands the hour after it.

Both windows fired short and both were paid, the only morning of the week where the two durations agreed, both cleared the gate, and both landed. The 30 minute call was the stronger at 70% of a 92% decided share.

The 60 minute window is the one to look at twice. Its forecast was 13 / 53 / 34 out of 203 sessions, which is Monday's cell, exactly. Monday it fired short and price went up. Thursday it fired short and price went down. Same table, same lookup, same call, opposite results, four days apart. Neither outcome tells you anything about the table; the pair of them together tells you what 53% means.

• 30m box18.4 pt · normal
• 60m box31.9 pt · normal
• 30m three-way22 / 70 / 9
• 60m three-way13 / 53 / 34
• 30m modeAtlas, short
• 60m modeAtlas, short
• Outcomedown rail, both windows
• Durationsagreed

Friday 14 August

A repeat pair that agreed, and an outcome mark that waits for Sunday

30m Atlas mode ✓ · 60m abstained, no target hit

The 30 minute window was the week's only long call. A normal 16.6 point box, up favoured 63% against 28, Atlas mode fired, and the 8.3 point target went inside the first half hour after the box closed.

The 60 minute window returned 41 / 29 / 31 from 244 sessions, which is Wednesday's cell. This time the repeat agreed: both abstained, and both ended with neither rail reached. Three repeat cells this week, two that split and one that did not.

One thing on this chart is not a fault. The 60 minute outcome column is blank rather than showing a no-target-hit check. A rail being hit stamps immediately, but the absence of a hit can only be confirmed once the race window has shut, and that check is drawn at the next session open. For a Friday that means Sunday evening. Friday's close had already settled it.

• 30m box16.6 pt · normal
• 60m box27.3 pt · normal
• 30m three-way63 / 28 / 9
• 60m three-way41 / 29 / 31
• 30m modeAtlas, long
• 60m modeabstain
• Outcome30m up rail · 60m no target
• Durationsagreed

What the week actually showed

• Three windows returned a forecast the indicator had already given, and two of the three pairs resolved in opposite directions. Monday and Tuesday's 30 minute windows were the same cell from the same 261 sessions and split up against down. Monday and Thursday's 60 minute windows were the same cell from the same 203 and split the same way. Wednesday and Friday's 60 minute windows were the same cell and both ended with no target hit. Nothing demonstrates more plainly that these are lookups into a fixed table rather than anything fitted to the day in front of them.
• The two gates do genuinely different jobs, and Tuesday showed why both are there. The 60 minute window that morning had a strong direction, 70% of its decided share, and abstained anyway because only 40% of comparable sessions reached a rail at all. The rail was reached in the end. Being cautious about whether a race finishes is a separate question from being confident about which way it finishes, and a single threshold cannot ask both.
• CPI expanded the boxes and the ladder cut the multiplier, which is the mechanism doing what it was built for. Wednesday's 30 minute box came in 58% over its five-day average, the multiplier stepped down to a quarter, and the resulting 7.1 point target was close enough that only 2% of comparable sessions fail to resolve it. The 60 minute window read the same expansion, only stepped to a half, sat its target 17.3 points out and never got there. Same day, same release, two different distances, two different answers.
• The durations disagreed on three of five mornings. They are not two views of one question. A 30 minute box and a 60 minute box are different boxes with different rails at different distances, and treating their agreement as confirmation would be double-counting the same hour.
• Five calls from ten windows, four of which landed, on a week where the unfiltered favoured rail went six from eight. Both numbers are far too small to mean anything and are published because the record has to start somewhere. The place to judge calibration is the indicator's own realised versus expected rows, which sat between 68 and 73% against a 66 to 71% expectation across this week.