If (early_high − open) ≥ (open − early_low): High First, else Low First
Broken definition
Post-arm H1 bar exceeds early extreme
Any H1 bar after 12:00 with High > early_high (if High First) or Low < early_low (if Low First)
H1 data range
2010-02-01 → 2026-03-13
99,999 bars
Weeks analysed
199
2022-05-08 → 2026-03-01 enriched window
Overall Summary
199
Weeks analysed
36.7%
Held (unbroken)
63.3%
Broken
73
Held count
126
Broken count
112.5h
Avg hrs to close
111.0h
Median hrs to close
14.9h
Avg hrs to break
8.0h
Median hrs to break
Reading these tables:Held = the early-window extreme level was NOT exceeded by any post-arm H1 bar — it remained the FINAL weekly extreme in that direction. Broken = at least one post-arm H1 bar exceeded the early extreme (went further in the same direction). A high Held% implies the Monday early session is setting a structural anchor for the week's primary direction.
Left: overall broken vs held donut. Right: by early-window direction.
Stability by Early-Window First Extreme Type
First Extreme
N Weeks
N Broken
N Held
Broken %
Held %
High
106
70
36
66.0
34.0
Low
93
56
37
60.2
39.8
High First = early window moved up more from Monday open. Low First = early window moved down more.
Stability by Candle Type
Candle Type
N Weeks
N Broken
N Held
Broken %
Held %
BullExpansion
32
10
22
31.2
68.8
BearExpansion
25
13
12
52.0
48.0
BullRejection
47
36
11
76.6
23.4
BearRejection
51
42
9
82.4
17.6
NeutralRange
44
25
19
56.8
43.2
Left: counts — held (green) vs broken (red). Right: broken% per type.
First Extreme Timing — Hour within Monday Early Window
Hour (UTC)
N_Weeks
N_Broken
Broken%
Held%
0.0
35.0
16.0
45.7
54.3
1.0
14.0
5.0
35.7
64.3
2.0
4.0
0.0
0.0
100.0
3.0
17.0
9.0
52.9
47.1
4.0
6.0
5.0
83.3
16.7
5.0
9.0
3.0
33.3
66.7
6.0
5.0
1.0
20.0
80.0
7.0
4.0
3.0
75.0
25.0
8.0
10.0
8.0
80.0
20.0
9.0
13.0
10.0
76.9
23.1
10.0
27.0
17.0
63.0
37.0
11.0
55.0
49.0
89.1
10.9
Left: how many weeks set their first extreme at each hour. Right: broken% by hour — lower = more stable at that hour.
Break Timing — When Do Extremes Break?
Day
N_Breaks
% of All Breaks
Avg Break Hour
Avg Hrs to Break
Monday
92
73.0
14.1
6.1
Tuesday
22
17.5
10.1
27.3
Wednesday
10
7.9
14.1
56.8
Thursday
2
1.6
14.0
75.5
Left: which day of the week was the early extreme broken. Right: distribution of hours from first extreme to the break bar.
Ultimate Extreme Day (from Enriched HighTime / LowTime)
This section uses HighTime/LowTime from the enriched weekly CSV — the timestamps of the ABSOLUTE weekly High and Low. For each week, the 'ultimate extreme day' is the day the FINAL weekly extreme in the same direction as the early-window first extreme was set. If a Monday early extreme matches the ultimate extreme → the extreme never moved further. If the ultimate extreme is on Friday → the extreme was expanded late in the week.
Day extreme finalised
N_Weeks
N_Broken
Broken%
Held%
Monday
108
35
32.4
67.6
Tuesday
50
50
100.0
0.0
Wednesday
28
28
100.0
0.0
Thursday
12
12
100.0
0.0
Friday
1
1
100.0
0.0
Left: how many weeks had their ultimate extreme set on each day. Right: broken% of early-window extremes by day the final extreme was set.
Sequence Alignment — Early Window vs Full-Week Enriched Sequence
The enriched Sequence column records whether the final weekly HIGH or final weekly LOW was set first. This section shows how the early-window direction (first 12 hours) aligns with the full-week final sequence. High agreement = the early window is a useful predictor of final weekly structure.
first_extreme_type
HighFirst
LowFirst
Total
%_HighFirst
%_LowFirst
High
106
0
106
100.0
0.0
Low
0
93
93
0.0
100.0
Left: enriched Sequence breakdown per early-window direction. Right: alignment % between early-window and full-week sequence.
Hours from First Extreme to Weekly Close
Green = held weeks, red = broken weeks. Shows how far in advance the early-week extreme is set relative to Friday close.
Judgement — Can the Early-Week Extreme Anchor a Directional Strategy?
Q1 — Is the early-week extreme structurally stable? No — only 36.7% held (73/199). More than half of early-week extremes are exceeded later in the week. The early extreme cannot reliably anchor a directional strategy without major filters.
Q2 — High-First vs Low-First stability is symmetric. High First: 34.0% held, Low First: 39.8% held — only 6 pp difference. Direction alone does not differentiate stability. The filter must come from candle type or time-of-day, not direction.
Q3 — Expansion week stability: mixed (59.6% held). Expansion weeks show moderate structural stability. The ARM_HOUR window may be capturing the early extreme for only a subset of expansion weeks.
Q4 — Monday-set extremes: strongly structural (67.6% held). When the final weekly extreme (HighTime or LowTime from enriched data) is on Monday, the early-window extreme is also the final extreme in 67.6% of those weeks. Monday's first 12 hours are the highest-stability period for extreme anchoring.
Verdict — Structural stability supports sequence trading model design. Expansion-week held rate (59.6%) meets the minimum threshold for anchoring a directional bias. Recommended next test: Use candle type + early-window direction as a combined filter to isolate the highest-stability subsets (above 55% held rate individually). Then design the entry and stop around those weeks.
