BullExpansion (32 weeks) — 90.6% LowFirst. The weekly Low prints before the High in 90.6% of cases. In 71.9% of BullExpansion weeks, that Low occurs on Monday.
BearExpansion (25 weeks) — 100.0% HighFirst. The weekly High prints before the Low in every BearExpansion week in this sample. In 68.0% of BearExpansion weeks, that High occurs on Monday.
Range premium: Expansion weeks average 211 pips vs 161 pips for all other types — a 31% range advantage. These are the highest-volatility weekly structures in the dataset.
NeutralRange precursor: After a neutral week, P(BullExpansion) = 22.7%, P(BearExpansion) = 22.7% — combined 45.4% expansion probability vs 28.5% base rate. Range compression appears to precede directional expansion.
Friday extremes: 22.0% of weekly Highs and 28.5% of Lows occur on Friday (random expectation: 20.0% each). Friday is a disproportionate extreme day across all candle types.
Data context: 200 weeks (3.8 years), single macro regime (post-COVID tightening, May 2022–Mar 2026). BullExpansion n=32, BearExpansion n=25. Treat all sub-group findings as indicative until n>50 per type across multiple regimes.
Candle Type Playbook
BullExpansion — 32 weeks (16.0% of 200 total)
Frequency
32 weeks — 16.0% of all weeks
Avg Range
221.9 pips (median 211.0 pips)
Sequence
HighFirst 9.4% | LowFirst 90.6%
Typical High Day
Friday (mode) | Mon 3.1% | Fri 68.8%
Typical High Hour
19:00 (mode) | Mean 15.4
Typical Low Day
Monday (mode) | Mon 71.9% | Fri 3.1%
Typical Low Hour
00:00 (mode) | Mean 5.7
Open Position
0.082 (0 = Low, 1 = High)
Close Position
0.868 (0 = Low, 1 = High)
Body
78.7% of range
Wick Profile
Upper 13.2% | Lower 8.2%
Structure
Opens near Low, expands upward, closes near High. Low set early (typically Monday). High set late. Strong committed bull structure.
ELI5 — How this might be traded
What the structure means
The market opened near the bottom of the week's range, climbed steadily, and closed near the top. The Low is set early (typically Monday morning) and the High arrives late in the week. The direction was clear and committed from the start — bulls were in control all week with minimal failed attempts.
Potential trading approach
Wait for the Monday Low to be established — ideally before 06:00–08:00 server time. Once Monday's session closes with no new lows being made, consider a long entry. Stop: below the Monday Low. Target: upper 80–90% of the week's range by Friday close (~180–200 pips on average). R:R estimate: if entry is 20–30 pips above the Monday Low, expected target is ~150–180 pips away — roughly 5:1 to 7:1 theoretical.
What would invalidate it
(1) Monday Low broken mid-week (new weekly Low prints on Tue–Thu) → structural failure; exit immediately. (2) Market extends higher on Monday but then immediately reverses — you are in a BearExpansion, not BullExpansion. (3) Major macro event Wednesday (Fed, NFP) can override weekly structure; reduce position into the event. (4) Multiple consecutive NeutralRange weeks prior suggest sustained range environment where directional expansions are less reliable.
BearExpansion — 25 weeks (12.5% of 200 total)
Frequency
25 weeks — 12.5% of all weeks
Avg Range
196.2 pips (median 168.5 pips)
Sequence
HighFirst 100.0% | LowFirst 0.0%
Typical High Day
Monday (mode) | Mon 68.0% | Fri 0.0%
Typical High Hour
03:00 (mode) | Mean 9.4
Typical Low Day
Friday (mode) | Mon 0.0% | Fri 80.0%
Typical Low Hour
17:00 (mode) | Mean 16.6
Open Position
0.832 (0 = Low, 1 = High)
Close Position
0.138 (0 = Low, 1 = High)
Body
69.4% of range
Wick Profile
Upper 16.8% | Lower 13.8%
Structure
Opens near High, expands downward, closes near Low. High set early (typically Monday). Low set late. Mirror of BullExpansion.
ELI5 — How this might be traded
What the structure means
Mirror image of BullExpansion. The market opened near the top of the week's range, sold off throughout the week, and closed near the bottom. The High is set early (typically Monday morning — 100% HighFirst in this dataset). Every single BearExpansion week in our sample made its High before its Low. The direction was committed from the opening bar.
Potential trading approach
Wait for the Monday High to be established — ideally before 06:00–08:00 server time. Once Monday's session shows the market weakening from the early High, consider a short entry. Stop: above the Monday High. Target: lower 10–20% of the week's range by Friday close (~160–190 pips on average). Note: Avg Open → High excursion is ~33 pips (the initial spike before reversal). The short opportunity usually appears after this spike fades.
What would invalidate it
(1) Monday High broken mid-week (new weekly High prints on Tue–Thu) → structural failure; exit immediately. (2) Strong bullish news event (e.g., ECB rate surprise) overrides weekly structure. (3) Market opens gap-down on Monday — this can set the week's Low early, producing a BullExpansion or NeutralRange week instead. (4) Sample warning: n=25 is small — this type's 100% HighFirst rate will not hold perfectly on larger samples. Apply the structural logic, not the exact statistic.
BullRejection — 47 weeks (23.5% of 200 total)
Frequency
47 weeks — 23.5% of all weeks
Avg Range
152.8 pips (median 151.8 pips)
Sequence
HighFirst 46.8% | LowFirst 53.2%
Typical High Day
Monday (mode) | Mon 31.9% | Fri 29.8%
Typical High Hour
17:00 (mode) | Mean 12.3
Typical Low Day
Wednesday (mode) | Mon 17.0% | Fri 4.3%
Typical Low Hour
17:00 (mode) | Mean 13.4
Open Position
0.568 (0 = Low, 1 = High)
Close Position
0.703 (0 = Low, 1 = High)
Body
20.0% of range
Wick Profile
Upper 26.5% | Lower 53.6%
Structure
Closes above midrange but body is narrow. Upper wick indicates at least one failed push higher. Directionally bullish but with friction.
ELI5 — How this might be traded
What the structure means
The market ended the week in the upper half of its range, but the body (Open–Close spread) is narrow. The week had a bullish direction overall but with detectable resistance — upper wicks indicate at least one failed push higher before settling with a bullish close. Think of it as a partial bull week with friction.
Potential trading approach
Not a primary entry signal. The narrow body means there is no clean directional commitment to trade against. Best used as prior-week context: if last week was BullRejection and this week shows early Monday LowFirst structure (Low prints before noon on Monday), the context strengthens a bullish bias for the current week. In isolation, do not trade this type.
What would invalidate it
Not applicable as a standalone entry. If using as prior-week context: the edge is weak if two consecutive BullRejection weeks occur (suggests resistance is building, not resolving).
BearRejection — 51 weeks (25.5% of 200 total)
Frequency
51 weeks — 25.5% of all weeks
Avg Range
162.9 pips (median 159.1 pips)
Sequence
HighFirst 74.5% | LowFirst 25.5%
Typical High Day
Tuesday (mode) | Mon 21.6% | Fri 2.0%
Typical High Hour
10:00 (mode) | Mean 12.3
Typical Low Day
Friday (mode) | Mon 11.8% | Fri 51.0%
Typical Low Hour
10:00 (mode) | Mean 12.1
Open Position
0.430 (0 = Low, 1 = High)
Close Position
0.259 (0 = Low, 1 = High)
Body
24.7% of range
Wick Profile
Upper 53.2% | Lower 22.1%
Structure
Closes below midrange but body is narrow. Lower wick indicates a tested support that partially recovered. Directionally bearish with friction.
ELI5 — How this might be traded
What the structure means
The market closed in the lower half of its range, but with a narrow body. The week had a bearish direction overall but with detectable support — lower wicks show at least one recovery attempt before the week settled lower. A partial bear week with friction.
Potential trading approach
Not a primary entry signal. Same logic as BullRejection in reverse. Best used as prior-week context: if last week was BearRejection and this week shows early Monday HighFirst structure (High prints before noon on Monday), the context strengthens a bearish bias for the current week. In isolation, do not trade this type.
What would invalidate it
Not applicable as standalone entry. If using as prior-week context: the edge is weak if two consecutive BearRejection weeks occur.
NeutralRange — 45 weeks (22.5% of 200 total)
Frequency
45 weeks — 22.5% of all weeks
Avg Range
166.1 pips (median 149.6 pips)
Sequence
HighFirst 42.2% | LowFirst 57.8%
Typical High Day
Monday (mode) | Mon 28.9% | Fri 15.6%
Typical High Hour
15:00 (mode) | Mean 11.6
Typical Low Day
Monday (mode) | Mon 33.3% | Fri 17.8%
Typical Low Hour
15:00 (mode) | Mean 13.4
Open Position
0.453 (0 = Low, 1 = High)
Close Position
0.455 (0 = Low, 1 = High)
Body
41.2% of range
Wick Profile
Upper 34.0% | Lower 24.8%
Structure
Open and close both near midrange. Wicks on both sides. No directional commitment — market was indecisive all week.
ELI5 — How this might be traded
What the structure means
The market went nowhere for the week. Both Open and Close are near the middle of the range, with wicks on both sides. Neither bulls nor bears won the week. This is a standoff week — the market was unable to commit to a direction.
Potential trading approach
Do not trade directionally on the NeutralRange week itself. The primary value of this type is as a precursor signal: in this dataset, after a NeutralRange week, the probability of the next week being a BullExpansion or BearExpansion nearly doubles compared to the base rate. Think of NeutralRange as a 'coiled spring' — the market is compressing before a directional release. The following Monday's early extreme (High or Low) is the trigger to watch.
What would invalidate it
Two or more consecutive NeutralRange weeks weaken the coiled-spring logic — this may be a sustained low-volatility environment rather than a temporary pause. Also: the NeutralRange → Expansion edge is directionally ambiguous; you still need the Monday early-extreme signal to determine which direction.
Weekly Extreme Timing Maps
All Weeks (n=200)
Weekly High timing — Day × Hour (counts in cells; colour intensity = % of 200 weeks)
Weekly Low timing — Day × Hour
High Peak Timing
Mode Day
Monday (28.5%)
Mode Hour
15:00
Mean Hour
12.3
Best 3-hour window
Friday 15:00–17:59 (15 / 7.5%)
Top 5 cells
Monday 03:00 (6 / 3.0%)
Monday 10:00 (6 / 3.0%)
Monday 15:00 (6 / 3.0%)
Wednesday 16:00 (6 / 3.0%)
Friday 15:00 (6 / 3.0%)
Low Peak Timing
Mode Day
Friday (28.5%)
Mode Hour
00:00
Mean Hour
12.2
Best 3-hour window
Monday 00:00–02:59 (23 / 11.5%)
Top 5 cells
Monday 00:00 (17 / 8.5%)
Friday 15:00 (8 / 4.0%)
Friday 17:00 (8 / 4.0%)
Tuesday 17:00 (6 / 3.0%)
Thursday 15:00 (5 / 2.5%)
BullExpansion (n=32)
Weekly High timing — Day × Hour (counts in cells; colour intensity = % of 32 weeks)
Weekly Low timing — Day × Hour
High Peak Timing
Mode Day
Friday (68.8%)
Mode Hour
19:00
Mean Hour
15.4
Best 3-hour window
Friday 14:00–16:59 (6 / 18.8%)
Top 5 cells
Friday 15:00 (3 / 9.4%)
Thursday 19:00 (2 / 6.2%)
Friday 11:00 (2 / 6.2%)
Friday 16:00 (2 / 6.2%)
Friday 19:00 (2 / 6.2%)
Low Peak Timing
Mode Day
Monday (71.9%)
Mode Hour
00:00
Mean Hour
5.7
Best 3-hour window
Monday 00:00–02:59 (17 / 53.1%)
Top 5 cells
Monday 00:00 (13 / 40.6%)
Monday 01:00 (3 / 9.4%)
Monday 05:00 (2 / 6.2%)
Monday 15:00 (2 / 6.2%)
Tuesday 17:00 (2 / 6.2%)
BearExpansion (n=25)
Weekly High timing — Day × Hour (counts in cells; colour intensity = % of 25 weeks)
Weekly Low timing — Day × Hour
High Peak Timing
Mode Day
Monday (68.0%)
Mode Hour
03:00
Mean Hour
9.4
Best 3-hour window
Monday 01:00–03:59 (5 / 20.0%)
Top 5 cells
Monday 03:00 (3 / 12.0%)
Monday 11:00 (3 / 12.0%)
Monday 01:00 (2 / 8.0%)
Monday 14:00 (2 / 8.0%)
Monday 00:00 (1 / 4.0%)
Low Peak Timing
Mode Day
Friday (80.0%)
Mode Hour
17:00
Mean Hour
16.6
Best 3-hour window
Friday 17:00–19:59 (10 / 40.0%)
Top 5 cells
Friday 17:00 (5 / 20.0%)
Friday 19:00 (3 / 12.0%)
Friday 15:00 (2 / 8.0%)
Friday 16:00 (2 / 8.0%)
Friday 18:00 (2 / 8.0%)
BullRejection (n=47)
Weekly High timing — Day × Hour (counts in cells; colour intensity = % of 47 weeks)
Weekly Low timing — Day × Hour
High Peak Timing
Mode Day
Monday (31.9%)
Mode Hour
17:00
Mean Hour
12.3
Best 3-hour window
Friday 15:00–17:59 (7 / 14.9%)
Top 5 cells
Friday 17:00 (3 / 6.4%)
Monday 01:00 (2 / 4.3%)
Monday 10:00 (2 / 4.3%)
Monday 14:00 (2 / 4.3%)
Monday 17:00 (2 / 4.3%)
Low Peak Timing
Mode Day
Wednesday (34.0%)
Mode Hour
17:00
Mean Hour
13.4
Best 3-hour window
Wednesday 16:00–18:59 (6 / 12.8%)
Top 5 cells
Wednesday 09:00 (3 / 6.4%)
Wednesday 17:00 (3 / 6.4%)
Monday 17:00 (2 / 4.3%)
Monday 18:00 (2 / 4.3%)
Tuesday 14:00 (2 / 4.3%)
BearRejection (n=51)
Weekly High timing — Day × Hour (counts in cells; colour intensity = % of 51 weeks)
Weekly Low timing — Day × Hour
High Peak Timing
Mode Day
Tuesday (35.3%)
Mode Hour
10:00
Mean Hour
12.3
Best 3-hour window
Wednesday 16:00–18:59 (7 / 13.7%)
Top 5 cells
Wednesday 16:00 (4 / 7.8%)
Monday 15:00 (3 / 5.9%)
Tuesday 10:00 (3 / 5.9%)
Tuesday 22:00 (3 / 5.9%)
Wednesday 18:00 (3 / 5.9%)
Low Peak Timing
Mode Day
Friday (51.0%)
Mode Hour
10:00
Mean Hour
12.1
Best 3-hour window
Friday 09:00–11:59 (8 / 15.7%)
Top 5 cells
Friday 09:00 (4 / 7.8%)
Friday 10:00 (3 / 5.9%)
Friday 15:00 (3 / 5.9%)
Friday 17:00 (3 / 5.9%)
Friday 23:00 (3 / 5.9%)
NeutralRange (n=45)
Weekly High timing — Day × Hour (counts in cells; colour intensity = % of 45 weeks)
Weekly Low timing — Day × Hour
High Peak Timing
Mode Day
Monday (28.9%)
Mode Hour
15:00
Mean Hour
11.6
Best 3-hour window
Monday 03:00–05:59 (5 / 11.1%)
Top 5 cells
Monday 03:00 (3 / 6.7%)
Thursday 15:00 (3 / 6.7%)
Monday 05:00 (2 / 4.4%)
Tuesday 16:00 (2 / 4.4%)
Wednesday 15:00 (2 / 4.4%)
Low Peak Timing
Mode Day
Monday (33.3%)
Mode Hour
15:00
Mean Hour
13.4
Best 3-hour window
Monday 00:00–02:59 (5 / 11.1%)
Top 5 cells
Monday 00:00 (3 / 6.7%)
Thursday 21:00 (3 / 6.7%)
Monday 10:00 (2 / 4.4%)
Tuesday 17:00 (2 / 4.4%)
Wednesday 15:00 (2 / 4.4%)
Expansion Strategy Research
Structural Distinctiveness
CandleType
N
AvgRange(pips)
CanonSeq%
CanonSeqNote
MondayExt%
AvgBodyPct
AvgOpenPos
AvgClosePos
BullExpansion
32
221.9
90.6
% LowFirst
71.9
0.787
0.082
0.868
BearExpansion
25
196.2
100.0
% HighFirst
68.0
0.694
0.832
0.138
BullRejection
47
152.8
—
n/a
—
0.2
0.568
0.703
BearRejection
51
162.9
—
n/a
—
0.247
0.43
0.259
NeutralRange
45
166.1
—
n/a
—
0.412
0.453
0.455
All Non-Expansion
143
160.6
—
n/a
—
0.283
0.482
0.467
Research Judgement
BullExpansion — proceed with dedicated research track.
Three structurally independent signals align:
(1) 90.6% canonical LowFirst sequence,
(2) 71.9% Monday Low concentration,
(3) early-morning timing cluster for the Low.
These are distinct from all other candle types simultaneously on range, position, and timing.
n=32 is below the ideal 50-count threshold —
continue collecting weekly data, but the direction justifies a dedicated research track.
BearExpansion — proceed, highest-signal type in this dataset.
100% HighFirst (n=25) is the strongest single structural signal here.
Wilson 95% CI lower bound is approximately 86% — directional edge is real even at this sample size.
Combined with 68.0% Monday High concentration,
BearExpansion is the most structurally distinctive type in the sample.
Caution: 100% will not hold on a larger dataset; the structural direction will remain valid
but the exact percentage will moderate.
Next research priorities before any EA consideration:
(1) Monday fail rate — how often does a Monday Low/High get broken later in the week?
(2) Entry precision — at what Monday hour exactly, and what confirmation is needed?
(3) Out-of-sample validation — test on 2010–2022 data (different macro regime).
(4) Extend dataset to 15-20 years for robust seasonality and conditional probability analysis.
Conditional Probabilities
Condition
CandleType
N_condition
N_match
Prob%
(base rate)
BullExpansion
200
32
16.0
(base rate)
BearExpansion
200
25
12.5
LowDay=Monday
BullExpansion
52
23
44.2
LowDay=Mon, LowHour<=06
BullExpansion
29
20
69.0
LowDay=Mon, LowHour<=12
BullExpansion
38
21
55.3
LowDay=Mon, LowHour<=18
BullExpansion
49
23
46.9
HighDay=Monday
BearExpansion
57
17
29.8
HighDay=Mon, HighHour<=06
BearExpansion
22
8
36.4
HighDay=Mon, HighHour<=12
BearExpansion
38
12
31.6
HighDay=Mon, HighHour<=18
BearExpansion
54
16
29.6
Sequence=LowFirst
BullExpansion
93
29
31.2
Sequence=HighFirst
BearExpansion
107
25
23.4
Seasonality: ISO Week Number
Thin-sample warning: With 200 weeks covering ~3.8 years,
the average observations per ISO week number is 3.8 (minimum 3).
This is insufficient for statistically reliable seasonality analysis.
To interpret these charts meaningfully, extend the dataset to 15–20 years
(750–1040 weeks), giving ~15–20 observations per week number.
Charts below are shown as-is for initial exploration only — treat individual week-number
readings as illustrative, not evidential.
Expansion probability, average range, and sample count by ISO week number
Highest Expansion Probability (n≥2)
Week
Exp%
n
Week 20
75.0%
4
Week 27
75.0%
4
Week 32
75.0%
4
Lowest Expansion Probability (n≥2)
Week
Exp%
n
Week 2
0.0%
4
Week 4
0.0%
4
Week 8
0.0%
4
Extending the dataset to 15–20 years: Open MetaTrader 5, run the
WeeklyOHLC_Export script with InpWeekCount = 1040
(default updated in v5). This captures approximately 20 years of EURUSD weekly data,
subject to your broker's history depth. Re-run this Python script — all analysis
adapts automatically to the input file size.
Time-Decay Weighting Analysis
How to read this: Time-decay weighting assigns more influence to recent weeks.
A half-life of 52 weeks means a week from 1 year ago has 50% weight vs a current week;
a week from 2 years ago has 25%. If a candle type's weighted frequency is higher than
its equal-weight frequency, that type has been occurring more often recently.
Conversely, a lower weighted frequency suggests that type was more common in the past.
Differences here reflect regime shifts within the dataset window.
Equal-weight vs time-decay-weighted frequency and range by candle type
CandleType
N
UW_Freq%
UW_AvgRange
W52_Freq%
W104_Freq%
W52_FreqDelta
W104_FreqDelta
BullExpansion
32
16.0
221.9
15.7
15.9
-0.3
-0.1
BearExpansion
25
12.5
196.2
11.2
11.6
-1.3
-0.9
BullRejection
47
23.5
152.8
22.0
22.6
-1.5
-0.9
BearRejection
51
25.5
162.9
24.2
25.0
-1.3
-0.5
NeutralRange
45
22.5
166.1
26.9
24.9
+4.4
+2.4
Interpretation guidance: A positive FreqDelta (e.g., +3.2%) means that candle type
has been more frequent in recent weeks than its overall historical rate.
This can indicate: (a) a regime shift toward/away from expansion behaviour,
or (b) normal sampling variation at small n.
Treat changes below ±2% as noise at this dataset size.
Transition Matrix
PrevCandleType
BullExpansion
BearExpansion
BullRejection
BearRejection
NeutralRange
BullExpansion
12.5
9.4
25.0
31.2
21.9
BearExpansion
8.0
8.0
24.0
36.0
24.0
BullRejection
19.1
6.4
23.4
29.8
21.3
BearRejection
13.7
11.8
15.7
27.5
31.4
NeutralRange
22.7
22.7
31.8
9.1
13.6
Key transitions
BullExpansion → BullExpansion
12.5%
BullExpansion → BearExpansion
9.4%
BearExpansion → BearExpansion
8.0%
BearExpansion → BullExpansion
8.0%
NeutralRange → BullExpansion
22.7%
NeutralRange → BearExpansion
22.7%
NeutralRange → Expansion rate is elevated in both directions. Expansion types rarely follow each other directly (low persistence). This supports using prior-week type as a signal filter.
Next Steps & Hypotheses
Monday fail rate [critical]: How often does a Monday Low/High that forms before 06:00
get broken later in the same week? This is the primary entry robustness test.
Monday extreme vs prior-week close: Does a Monday Low below prior-week Close (or
Monday High above prior-week Close) add predictive power beyond the day signal alone?
Range compression precursor: If Monday H-L range < 40% of prior-week range
→ P(Expansion this week) = ? Quantify the NeutralRange → Expansion observation.
Hour 00:00 spike: ~11% of weekly Lows hit at hour 00 server.
Is this a Sunday-gap artefact? Inspect with H1 data directly.
Extend dataset to 15–20 years (re-run MT5 export with InpWeekCount=1040)
for: reliable seasonality analysis, regime comparison, and larger expansion sub-groups.
Regime split: Divide dataset by macro regime (trending vs. ranging).
Does expansion frequency, timing, and sequence clarity differ by regime?
ATR normalisation: Normalise all pip-based distances by 4-week ATR
to remove regime-level volatility bias from the statistics.
Thursday close signal: Does Thursday's D1 close position predict
whether Friday completes or reverses the expansion?
Multi-timeframe confirmation: Does H4 structure on Monday morning
(e.g., H4 close direction, H4 body size) align with the weekly expansion direction?
