Synapse Starts Journey (SSJ)

L3 SOMA — Read Layer · 12 Stages Sequential Monitoring · 2026-05-15 · Cell CB1R (CYCLICAL_BULL_sub1) (2025-Today)

L3 = Synapse Network의 "오늘 어디서 시작 → 어디로 전파 → 어디서 끝나는가" 진단.
Captain F-028 수계도(水系圖) 비유 (4/27): 발원지 → 지류 → 호수 → 댐 수위 → 방류 → 유역 전이.
Stage 1-10 (building blocks) → Stage 11-12 (결론 HAFS + HAL). L4 5 Doctors / L5 Cell의 input layer.

ARIS Neuron Network Neuron Monitor Timeline 1 (Verdict) Timeline 2 (Neuron Network)

HAFS (Today, Atlas SSOT)

+0.85σ

32y percentile 100th · 284 neurons

Path Probability (60d kNN)

A 70% / B 0%

SSOT Atlas Top 10 analogs · 60d forward

Today CAT Vote

13 /16/ 20

Bull / Neutral / Bear (49 CAT)

L4 Verdict

WEAK_BULL

SPY $739.17 · Div +0.121

Today Market Noteworthy Findings

구분	★ Thesis	Detail	Remark
① 시장 진단 L3 SOMA · HAFS	★ Bull mature peak — mean reversion 압력 강함	HAFS +0.85σ · 32y percentile 100th 32y rare extreme zone · CAT vote 13/16/20 (Bear-leaning 41%) · L4 WEAK_BULL · Div +0.121 → Surface Bull (CAT/L4) + Internal Bear (Neuron L0) — Maturity Pullback signal	✓ CRITICAL
② Neuron 발견 현상 origin	★ "Sector rotation/MAS Sector 표면 강세 + Credit/factor 가족 내부 약화"	Bull: NFP · RT_XLK_XLU · MAS_AAM_200_XLRE · Bear: HY_TR · IG_TR · SP_B_BB · Endpoint: Credit spread (OAS 확대) + Yield curve (Z_GT2/Z_GT5 음수)	✓ CAPTAIN V1
③ 변화 방향 Multi-window TE	★ Multi-window 모두 평균 부근 → 안정 / 횡보	Sign-flip: short-mid 5 · mid-long 5 · TE z-score: w20 +0.00σ / w40 +0.00σ / w100 +0.00σ / w250 +0.00σ · Multi-window 흐름 일관	✓ TRANSITION
④ 32y 과거 HAL kNN Top 10	★ Trump rally (2017) 후속 — 단기 횡보 → 중장기 강세 가속	Today PC: +1.90 / +0.04 / -0.79 · cluster: 2017 (10개) · 60d analog forward +2.9~+5.0% (100% pos) · 240d mean +18.51% (100%, 10/10)	✓ HIGH CONF

⑤ 종합 판단 Data-driven forecast

★ Maturity Pullback → V-bounce 회복 → 장기 강세 가속 진행 가능성 HIGH

PATH A · Rally >+5%

70%

PATH B · Pullback <−5%

0%

PATH C · Range ±5%

30%

단기 (5-15d): 횡보 +0.17/+0.06% (pos 60/50%) · 중기 (60d): +3.88% (pos 100%) · 장기 (240d): 강세 가속 +18.51% (pos 100%, 10/10 analogs)

Caveat: 32y rare zone (HAFS 100th percentile) → mean reversion 압력 동시 존재 (단기 pullback risk)

1 Source Neuron — 발원지 (Today firing) Bull 127Bear 45

Stage 1 = Source Neuron — "발원지" (Today firing 시작점)
"Captain F-028 수계도 비유 — 강이 시작되는 곳, 누가 먼저 깨어났나"
측정 방법: 284 daily neurons × HAFS (Status × Direction, σ 단위) · Today specific

정의 — 발원지 (Source): 강이 시작하는 곳, 누가 먼저 깨어났나

수계도 비유: 32y 시장 신호의 흐름이 시작되는 발원지. 오늘 어느 neuron이 가장 먼저 firing하는가?
284 daily neurons × HAFS (History Analog Finder Score) = Status × Direction (Sign-Corrected, z-score 기반 σ 단위).
Status = activation magnitude (방향 X, 강도). Direction = +1 Bull / 0 / −1 Bear (S6 sign + S7 vote).

Top 10 BULL (today firing)

Neuron	HAFS	Status	Top 10 streak
NFP	+9.45σ	9.45σ	NEW
RT_SPSM_SPY	+3.62σ	3.62σ	NEW
ICI_ETF_TOTAL	+2.61σ	5.23σ	NEW
XSMO	+2.35σ	2.35σ	NEW
BRENT	+2.18σ	2.18σ	1d
SLYG	+2.15σ	2.15σ	NEW
XMMO	+2.11σ	2.11σ	NEW
CLc12	+2.07σ	2.07σ	NEW
SOXX	+2.01σ	2.01σ	NEW
MAS_AAM_200_XLE	+1.97σ	1.97σ	NEW

Top 10 BEAR (today firing)

Neuron	HAFS	Status	Top 10 streak
RT_COPPER_GOLD	-1.70σ	3.39σ	NEW
RT_COPPER_BRENT	-1.64σ	3.28σ	NEW
COPPER	-1.50σ	3.01σ	NEW
RT_VTI_AAXJ	-1.36σ	1.36σ	NEW
RT_VTI_VXUS	-1.14σ	2.27σ	NEW
RT_VTI_ILF	-0.89σ	1.78σ	NEW
RT_MDY_SPY	-0.80σ	1.59σ	NEW
RT_SLYG_SLYV	-0.71σ	1.43σ	NEW
BTC	-0.64σ	1.27σ	NEW
RT_VTI_EFA	-0.62σ	1.23σ	NEW

Stage 1 측정 방법 (click to expand)

284 daily neurons × HAFS = Status × Direction
• Status = activation magnitude (|z| 강도)
• Direction = +1 Bull / 0 neutral / −1 Bear (S6 sign + S7 vote)
• HAFS = signed magnitude (σ 단위, +/−)
• Top 10 Bull/Bear 추출 + 30d streak 추적

Stage 1 종합 결론 (click to expand)

Today firing 시작점 식별 — 오늘 시장이 어디서 깨어나는가
→ Macro/Sector 강세 (NFP/RT_XLK_XLU/MAS_REIT) + Credit 약세 (HY_TR/IG_TR/SP_B_BB)
→ "Macro 표면 강세 + Credit 내부 약화" 패턴 발견
이 신호가 Stage 2 (분기) → Stage 4 (Path) → Stage 11 (HAFS scalar)로 propagate

Sources: dashboard_v2_hafs_today.json (today) + stage11_HAFS_daily.npz (8,385d × 284 neurons, last 30d streak) · HAFS σ 단위 Dashboard 전체 통일 의무 (Captain 5/16) · Atlas data layer 도착 시 32y context 추가

2 Sub-cluster — 지류 분기 (Active + Mixed-sign decomposition) 44 CATs · 5 mixed

Stage 2 = Sub-cluster — "지류 분기" (Active + Mixed-sign decomposition)
"49 CAT 가족 안에서 강줄기가 갈라지는가, 결속하는가"
측정 방법: PCA off-diagonal correlation × Mixed-sign vs Coherent · Today specific

정의 — 지류 분기: 49 CAT 안에서 강줄기가 갈라지는가, 합쳐지는가

수계도 비유: 발원지에서 시작한 신호가 49 CAT 가족 안에서 갈라지는지(Mixed-sign) / 한 방향으로 함께 흐르는지(Coherent) 측정.

각 CAT은 여러 member neuron으로 구성. Cohesion = members 간 방향 일치도 (PCA off-diagonal correlation).
• Coherent CAT (high cohesion, +) — members 같은 방향 firing (family 일관)
• Mixed-sign CAT (low/negative cohesion) — members 방향 갈림 (분기 → sub-cluster 분해 신호)

Top 5 Mixed-sign CATs (Today, 방향 갈림)

CAT	Cohesion	n	Members
C5	-0.792	2	SLYV_SLP50, VRP
C39	-0.283	3	ANFCI, NFCI, Z_NFCI
C48	-0.256	3	UNEMPLOYMENT, BANK_RESERVES, FRED_TGA
C37	-0.048	4	GOLD, SILVER, COT_SILVER, COT_JPY_NCN
C2	-0.001	5	COT_DOLLAR, DXY, USDCAD, USDJPY...

Top 5 Coherent CATs (Today, 일관 family)

CAT	Cohesion	n	Members
C36	+0.988	4	BRENT, WTI, BRENT_DAILY, CRACK_SPREAD
C31	+0.971	3	COPPER, RT_COPPER_BRENT, RT_COPPER_GOLD
C43	+0.954	3	CCC_HY_YIELD, CCC_OAS, SP_CCC_BB
C22	+0.897	5	MAS_IT, MAS_AAM_56_XLK, MAS_AAM_200_XLK, MAS_AAM_56_XLY...
C1	+0.896	6	CVX, SLB, VDE, XOM...

32y Baseline Mixed-sign 4 (Sign-Correction 후 발견된 갈림 CAT)

CLAUDE.md 4/28 PM L3 Re-Test + Engine Key 발견 — Sign-Correction 후 32년 평균에서 members가 양/음 동시 firing으로 발견된 CATs. 원본 5개 발견 중 C3는 4/27 49 CAT rebuild로 다른 cluster에 merge. 현재 49 CAT 유효 baseline = 4개. Today와 비교하면 변화 추적 가능.

CAT	Members family	Today Cohesion	Status vs baseline
C5	SLYV_SLP50 + VRP	-0.792	still mixed
C0	Commodity / Inflation family	+0.049	now neutral
C2	Forex / Dollar family	-0.001	now neutral
C39	Financial Conditions (NFCI)	-0.283	still mixed

Stage 2 측정 방법 (click to expand)

49 CAT 가족 단위 결속도 측정 (PCA off-diagonal correlation)
• Coherent CAT = members 같은 방향 firing (high cohesion +)
• Mixed-sign CAT = members 방향 갈림 (low/negative cohesion)
• 32y baseline 4 Mixed (4/28 Sign-Correction 발견, 영구 reference)

Stage 2 종합 결론 (click to expand)

일관 family vs 분기 family 동시 존재 진단
→ Coherent (C36 Oil/C31 Copper/C43 CCC Credit) = 결속 유지
→ Mixed-sign (C5/C39 NFCI/C48/C37/C2) = 분기 진행
→ "어느 가족이 한 방향, 어느 가족이 갈리는가" 시장 sub-cluster 진단

Source: dashboard_v2_cat_cohesion_today.json (44 CATs Today) + CLAUDE.md 4/28 PM baseline (32y Sign-Corrected, 5개 → 4개 in 49 CAT). Atlas 작업 대기: dynamic within-CAT sub-cluster decomposition (k-means / spectral) + Sub vs full CAT 분기 daily metric.

3‍‍‍ Coherence Static — 가족 PHOTO (32y baseline 결속도) 1 diverged · 5 baseline

Stage 3 = Coherence Static — "가족 사진" (32y PHOTO baseline)
"Captain F-028 비유 — 32년 가족 사진 1장 (Long-term synaptic weight)"
측정 방법: Recent 10y PCA + mean off-diagonal correlation · 32y static reference

정의 — 가족 사진: 32년 강줄기 가족이 누구와 함께 흐르는가

수계도 비유 (Captain F-028): 강줄기 가족 사진 1장 — "누가 누구와 함께 움직이나" (Long-term synaptic weight, LTP 굳어진 상태).

Long-term synaptic weight (LTP 굳어진 상태). Recent 10y PCA + mean off-diagonal correlation.
32y baseline = static reference (CLAUDE.md 4/28 PM PostAudit 49 CAT). Today와 비교하면 "강한 가족이 유지되는가" 추적.

32y Baseline Top 5 Coherent CATs (CLAUDE.md 4/28 PM PostAudit)

CAT	4/28 PM Family	32y PC1	Today Cohesion	Status
C36	Commodity + Energy	0.988	+0.988	Stable
C39	Credit (4/28 PM 명명)	0.952	-0.283	Diverged (now mixed)
C43	CCC Credit	0.949	+0.954	Stable
C20	MAS Industry (4/28 PM)	0.937	+0.853	Slightly weakened
C30	Risk-off (4/28 PM)	0.931	+0.550	Significantly weakened

CAT Pair Divergence — Risk-on ↔ Risk-off (4/28 PM Captain V1 자동 분리)

CAT	Today Family	4/28 PM 명명	4/28 PM Pair Corr
C26	IAI_Capital_Markets, ITA_Aerospace_Defense, ITB_Homebuilders_Alt, IYT_Transportation...	Risk-on (4/28 PM)	−0.744
C30	FCG, OIH_Oil_Services	Risk-off (4/28 PM)	−0.744

Stage 3 측정 방법 (click to expand)

32y baseline 가족 사진 (Long-term synaptic weight, LTP 굳어진 상태)
• Top 5 Coherent CATs (32y 평균 PC1 + Today 비교)
• CAT Pair Divergence (Risk-on ↔ Risk-off 자동 분리, 4/28 PM)
• Stage 3 + Stage 9 = Captain V1 "Engine Key" 4-Layer Coherence

Stage 3 종합 결론 (click to expand)

32년 본질 가족 식별 + Today vs baseline 비교
→ C36 Commodity/Energy (32y 0.988, Today +0.988) = Stable 유지
→ C39 Credit (32y 0.952, Today -0.283) Diverged → Credit 가족 약화 진행
→ "32년 강한 가족이 지금도 강한가" 시장 본질 진단

Source: dashboard_v2_cat_cohesion_today.json (Today 44 CATs) + CLAUDE.md 4/28 PM L3 Re-Test PostAudit baseline. Caveat: 4/27 49 CAT rebuild 후 일부 CAT member identity 미세 shift (예: C20/C30). 32y cohesion 강도는 보존되나 family 의미 변동 가능. Atlas 작업 대기: 32y rolling coherence daily metric + CAT Pair divergence dynamic measurement.

4 Path — 지류 / 흐름 (Source → Receiver → Endpoint propagation) 5 Src · 5 Rcv · 4 End

Stage 4 = Path — "지류 / 흐름" (Source → Receiver → Endpoint)
"Captain V1 4/12 — "수원지 → 지류 → 본류 → 수계도 / 시작 → 어디로 → 끝""
측정 방법: Multi-window TE 11 windows × 4-tier directed propagation · Today specific

정의 — Captain V1 4/12 비전: "수원지 → 지류 → 본류 → 수계도" / "시작 → 어디로 → 끝"

Source = firing 시작 (오늘 가장 강한 신호) · Receiver = 영향 받는 next-tier · Endpoint = propagation 도달점 (leverage / macro structural).
Synaptic transmission pathway · Multi-window TE 11 windows × 4-tier directed (5/10/15/20/30/40/60/80/100/250/500d).

Source Top 5 (firing 시작)

Neuron	z	Meaning
NFP	+6.79σ	고용 지표 강세 (Macro 강함)
HY_TR	-4.13σ	HY_TR
IG_TR	-3.95σ	IG_TR
SP_B_BB	-3.88σ	B-BB credit spread
RT_XLK_XLU	+3.83σ	Tech/Utility (risk on/off)

Receiver Top 5 (영향 받음)

Neuron	z	Meaning
IWF_SLP50	-3.76σ	IWF_SLP50
IWF	-3.76σ	IWF
B_OAS	-3.62σ	B_OAS
MAS_AAM_200_XLRE	+3.54σ	REIT sector 강세 (부동산 매수세 강력)
CRACK_SPREAD	+3.42σ	Crack spread (정유 마진)

Endpoint (propagation 도달점)

Neuron	z	Meaning
B_OAS	-3.62σ	B_OAS
MARGIN_DEBT	+2.97σ	Leverage 누적 (위험 신호)
Z_GT5	-2.81σ	5년 z-score
Z_GT2	-2.80σ	2년 z-score

Neuron Network Flow (M:N / N:M)

Today Flow: Source = macro · credit · ratios → Receiver = factor · credit · mas_aam → Endpoint = credit · margin · yield_curve

M:N = 각 Source가 여러 Receiver로 분기 · N:M = 각 Receiver가 여러 Endpoint로 합류 (수계도 비유)

Stage 4 측정 방법 (click to expand)

Source → Receiver → Endpoint 강줄기 전파 측정
• Source = firing 시작 (오늘 가장 강한 신호)
• Receiver = 영향 받는 next-tier
• Endpoint = propagation 도달점 (leverage / macro structural)
• M:N (source→multi receivers) + N:M (multi receivers→multi endpoints)

Stage 4 종합 결론 (click to expand)

오늘 시장 강줄기 흐름 진단
→ Source = macro/credit/ratios (NFP/HY_TR/RT_XLK_XLU)
→ Receiver = factor/credit/mas_aam (IWF/B_OAS/MAS_REIT)
→ Endpoint = credit/margin/yield curve (B_OAS/MARGIN_DEBT/Z_GT5)
→ "Macro/Sector 강세 → Tech/Credit 전파 → Leverage 누적 + Yield 음수" 도달

Source / Receiver: UV JSON strong_neurons_bull/bear magnitude proxy (Top by |z|). Endpoint: keyword filter (MARGIN_DEBT / Z_GT / VIX / MOVE / OAS / TGA / TLT). Today Flow narrative = data-driven (category prefix grouping).
Caveat: 현재 magnitude-based proxy (정확한 TE-based propagation X). 진짜 Source→Receiver→Endpoint = Atlas 작업 (공문 #144 P1-B): TE rolling 11 windows × 4-tier directed daily 측정.

3D PCA Path Visualization + Multi-window Forward 분포 (옛 L3 Pathway 통합, click to expand)

32y daily PC1/PC2/PC3 trajectory (monthly subsample, ~400 points). Color by HAL Stage K=10. Today + kNN Top 10 highlighted.

Today PC 좌표 (2026-04-24):

  PC1 = +1.90  ↑ Bull state magnitude 강함

  PC2 = +0.04  → 보조 차원 neutral

  PC3 = -0.79  ↓ tertiary mode 음수 (특정 패턴 활성)

= "강한 Bull magnitude + 특정 sub-pattern" 좌표
= 32y space에서 specific zone (다른 시기 거의 안 나타나는 좌표)

Source: dashboard_v2_pcs_all.json (32y daily PCs, 1993-2026) + dashboard_v2_knn_multiwindow.json (Top 10 analogs).
Multi-window Forward 분포 + Path A/B/C + 6중 검증 = Stage 12 (HAL)로 이동 ↓ (Captain 5/16 분리).

5 Convergence — 호수 (7 강이 한 호수로 / v5 8 dim Coincidence) 143 Bull · 26 Bear (5+/8 dim aligned)

Stage 5 = Convergence — "호수" (7 강이 한 호수로)
"Captain V1 4/27 — "7개의 강이 한 호수로 흘러든다. 그 호수가 시장의 답""
측정 방법: v5 8 dim per-neuron coincidence (5+/8 dim 같은 방향 firing) · Today specific

정의 — 호수: 여러 강줄기가 한 호수로 모이는 곳 (Multi-source Coincidence)

Captain V1 직접 발화 (4/27 Stage5_L3A_Convergence Final Report):
"7개의 강이 한 호수로 흘러든다. 그 호수가 시장의 답. 모든 source가 결국 거기로. 그 점이 phase change의 결론."

L3-A.5 method: "Sink + Path multi-source + Family". v5 8 dim (LEVEL_AVG / PMO / accel / ewma_z / accum_pos / accum_neg / peer_rel / abs_level) per-neuron 측정 — 각 neuron이 5+/8 dim에서 동일 방향 firing → strong coincidence.
Note: abs_level = magnitude (방향 ≥0). Hebbian "Fire together wire together" = L5 Synapse Cell 영역, L3 Stage 5에서는 동시성 측정만.
Captain Stage 5 통찰: Weekly PMO > Daily PMO (5/16 발견 — SS +83%, NP-Bull/Bear +20% — Pring Weekly PMO 우수).

Today 8 dim per-neuron distribution (294 neurons)

Dim	Pos count (>0.5)	Neg count (<−0.5)	Top 3 Bull-side neurons	Top 3 Bear-side neurons
LEVEL_AVG	119	58	CRACK_SPREAD +2.82 · RT_XLK_XLU +2.75 · SOXX +2.00	IG_TR -5.73 · HY_TR -5.16 · IWF_SLP50 -5.06
PMO	162	82	ICI_MF_HY +72788.44 · SKYY +10305.09 · SLYV_SLP50 +2653.18	NFP -26229.04 · PBJ_Food_Beverage -1213.70 · VRP -976.49
accel	58	35	NFP +1800.07 · VRP +37.08 · MA200D_SPX +36.28	COT_JPY_NCN -34.72 · SP_B_BB -31.35 · SLYV_SLP50 -25.23
ewma_z	161	50	NFP +3.89 · CRACK_SPREAD +2.83 · RT_XLK_XLU +2.66	HY_TR -2.40 · IG_TR -2.29 · SP_B_BB -2.24
accum_pos	200	39	MARGIN_DEBT +1220.00 · BANK_RESERVES +1103.00 · ISM_MFG +206.00	PCE_CORE_YOY -645.00 · CPI_YOY -496.00 · PPI_YOY -435.00
accum_neg	109	32	UMICH_SENTIMENT +913.00 · NYSE_MARGIN_FREE +480.00 · ISM_MFG +397.00	UNEMPLOYMENT -544.00 · ACM_TP10 -532.00 · ICI_MF_MIDCAP -178.00
peer_rel	94	78	NFP +2.94 · MARGIN_DEBT +2.85 · SOXX +2.60	IWF -4.52 · VRP -2.74 · FCG -2.62
abs_level	216	0	MAS_AAM_56_XLI +1.00 · MAS_AAM_200_XLK +1.00 · MAS_AAM_56_XLK +1.00	IG_TR +0.00 · FRED_RRP +0.00 · RT_COPPER_GOLD +0.00

Today Strong Coincidence — Neurons firing same direction across 5+/8 dims (v5 baseline)

8 dim = LEVEL_AVG · PMO · accel · ewma_z · accum_pos · accum_neg · peer_rel · abs_level (magnitude)
Note: abs_level은 magnitude (방향 ≥0). 5+/8 dim 같은 방향 firing = strong multi-source coincidence.

Neuron	Direction	8 dim values (LEVEL/PMO/accel/ewma/acc+/acc−/peer/abs)
SKYY	BULL 6/8 dim	+0.6 \| +10305.1 \| -9.1 \| +0.8 \| +1.0 \| +6.0 \| -0.3 \| +0.6
SLYV_SLP50	BULL 6/8 dim	+1.8 \| +2653.2 \| -25.2 \| +1.8 \| +19.0 \| +0.0 \| +1.0 \| +0.9
GDX_Gold_Miners	BULL 5/8 dim	-0.0 \| +1337.7 \| +1.4 \| +0.2 \| +3.0 \| +20.0 \| -1.0 \| +0.5
MARGIN_DEBT	BULL 6/8 dim	+1.2 \| +14.4 \| +0.2 \| +2.1 \| +1220.0 \| +0.0 \| +2.9 \| +1.0
IGV_Software	BULL 7/8 dim	+0.5 \| +1040.5 \| +5.2 \| +0.8 \| +3.0 \| +8.0 \| -0.3 \| +0.5
UTES	BULL 6/8 dim	+0.2 \| +596.2 \| +9.7 \| +0.4 \| +11.0 \| +9.0 \| -0.7 \| +0.6
GEX	BULL 8/8 dim	+0.9 \| +557.8 \| +31.3 \| +1.1 \| +11.0 \| +21.0 \| +1.1 \| +0.5
NYSE_MARGIN_FREE	BULL 8/8 dim	+0.7 \| +4.8 \| +0.4 \| +0.8 \| +120.0 \| +480.0 \| +0.8 \| +0.6
VRP	BEAR 5/8 dim	-1.8 \| -976.5 \| +37.1 \| -1.9 \| -29.0 \| -0.0 \| -2.7 \| +1.0
SP_B_BB	BEAR 6/8 dim	-2.7 \| -735.6 \| -31.4 \| -2.2 \| -40.0 \| -0.0 \| -1.9 \| +0.9
B_OAS	BEAR 6/8 dim	-2.8 \| -318.4 \| -12.6 \| -2.1 \| -21.0 \| -0.0 \| -1.8 \| +0.9
COT_EUR	BEAR 5/8 dim	-0.4 \| -364.9 \| -0.9 \| -0.5 \| +47.0 \| +8.0 \| -0.4 \| +0.6
COT_EUR_NCN	BEAR 5/8 dim	-0.4 \| -364.9 \| -0.9 \| -0.5 \| +47.0 \| +8.0 \| -0.4 \| +0.6
COT_VIX_NCN	BEAR 6/8 dim	-0.8 \| -25.9 \| +0.7 \| -0.7 \| -93.0 \| -57.0 \| -0.9 \| +0.6
SP_CCC_BB	BEAR 6/8 dim	-1.2 \| -131.3 \| -0.1 \| -0.5 \| -39.0 \| -1.0 \| -0.6 \| +0.2
Z_GT5	BEAR 7/8 dim	-1.5 \| -8.5 \| -8.3 \| -1.9 \| -47.0 \| -5.0 \| -1.1 \| +0.7

Convergence Verdict

Partial Bull-leaning (Bull 143 / Bear 26 neurons 5+ dim aligned)

Our Findings:
26 neurons (5+ dim Bear aligned) — multi-source coincidence Bear side 강함, regime 안정 (Bear side). Weekly PMO 강도가 Daily PMO보다 forward predictive 우수 → Stage 5 weight = Weekly basis (Captain 5/16 발견).

Stage 5 측정 방법 (click to expand)

v5 8 dim per-neuron coincidence detection
• 8 dim = LEVEL_AVG · PMO · accel · ewma_z · accum_pos · accum_neg · peer_rel · abs_level
• 각 neuron이 5+/8 dim 같은 방향 firing → strong coincidence
• F-AI-001 정합 (per-neuron specific, aggregate X)

Stage 5 종합 결론 (click to expand)

여러 강줄기 한 호수로 모이는 강도 측정
→ 143 neurons Bull-side 5+/8 dim coincidence + 26 Bear-side
→ "Bull-side multi-source coincidence 우세" — 호수 한쪽으로 흐름 집중
→ Stage 6 (댐 수위)로 propagate → 압력 누적

Source: NPZ v5 8 dim SIGN_CORRECTED (8400 days × 294 × 8) · Today 2026-05-15 · F-AI-001 정합 (aggregate 폐기, per-neuron specific) · Captain 5/16 v5 8 dim baseline.

6 Pressure — 댐 수위 / 압력솥 (Bull − Bear + per-neuron extreme) Net -0.14 · 5 extremes

Stage 6 = Pressure — "댐 수위 / 압력솥" (Accumulated stress)
"Captain F-028 — 호수에 누적된 감정 압력 (Bull pressure vs Bear pressure)"
측정 방법: CAT 49 vote + per-neuron extreme (|z|≥1.5σ) · Today specific

정의 — 댐 수위 / 압력솥: 호수에 누적된 감정 압력 (Captain F-028)

수계도 비유 (Captain F-028 + _l3b1_deviation_map):
• 댐 수위 = Accumulated Stress (감정 누적량)
• 압력솥 + = accum_pos (좋은 감정 누적, Bull pressure)
• 압력솥 − = accum_neg (나쁜 감정 누적, Bear pressure)
• +60d 모두 좋은 감정 누적 = "압력솥 폭발 직전 (Bull)"

S6 sign(Bull pressure − Bear pressure) + per-neuron extreme detect (Captain V1 본체).
Extreme = leverage (MARGIN_DEBT/OAS) · sentiment (UMICH/PUTCALL) · stress (NFCI/VIX) · liquidity (TGA/RRP) · yield curve (GT*).

Today CAT 49 Vote Distribution

Bull pressure	13 / 49 (27%)
Neutral	16 / 49
Bear pressure	20 / 49 (41%)
Net pressure	-0.143

Per-neuron Extreme Detection (|z| ≥ 1.5σ)

Category	Neuron	z	Meaning
Leverage / Credit	MARGIN_DEBT	+2.97σ	Leverage 누적 (위험 신호)
Leverage / Credit	HY_TR	-4.13σ	HY_TR
Leverage / Credit	B_OAS	-3.62σ	B_OAS
Yield Curve	Z_GT5	-2.81σ	5년 z-score
Yield Curve	Z_GT2	-2.80σ	2년 z-score

Top Bull Pressure (|z| ≥ 1.5σ)

Neuron	z	Meaning
NFP	+6.79σ	고용 지표 강세 (Macro 강함)
RT_XLK_XLU	+3.83σ	Tech/Utility (risk on/off)
MAS_AAM_200_XLRE	+3.54σ	REIT sector 강세 (부동산 매수세 강력)
CRACK_SPREAD	+3.42σ	Crack spread (정유 마진)
RT_MDYG_MDYV	+3.26σ	RT_MDYG_MDYV
MAS_AAM_56_XLRE	+3.20σ	REIT 단기 강세 확인
XLK	+3.09σ	Tech sector 강세
MARGIN_DEBT	+2.97σ	Leverage 누적 (위험 신호)

Top Bear Pressure (|z| ≥ 1.5σ)

Neuron	z	Meaning
HY_TR	-4.13σ	HY_TR
IG_TR	-3.95σ	IG_TR
SP_B_BB	-3.88σ	B-BB credit spread
IWF_SLP50	-3.76σ	IWF_SLP50
IWF	-3.76σ	IWF
B_OAS	-3.62σ	B_OAS
Z_GT5	-2.81σ	5년 z-score
Z_GT2	-2.80σ	2년 z-score

Our Findings:
CAT vote: 13 Bull / 20 Bear → net pressure -0.143. Per-neuron extreme: 5 neurons |z|≥1.5σ. Leverage/Credit 극단 detect — 잠재 risk accumulation.

Stage 6 측정 방법 (click to expand)

댐 수위 = 호수 압력 누적량 측정
• CAT 49 vote (Bull / Neutral / Bear count)
• Bull pressure = accum_pos (좋은 감정 누적)
• Bear pressure = accum_neg (나쁜 감정 누적)
• Per-neuron extreme (|z|≥1.5σ, leverage/credit/yield 등)

Stage 6 종합 결론 (click to expand)

댐 수위 + 압력솥 진단
→ Today CAT vote 13 Bull / 16 / 20 Bear → Net pressure -0.143 (Bear-side 우세)
→ Per-neuron extreme 5개 detect (Leverage MARGIN_DEBT + Credit OAS 극단)
→ "댐 수위 Bear-side 누적 + 단기 polarization" → Stage 7 (방류) 임박

Source: master_db.csv (cat_bull_count / cat_neutral_count / cat_bear_count, 5/15) + UV strong_neurons (Top by |z|).

7 Trigger — 방류 / 발사 (5 axis 3/5 threshold crossing + 32y events) 0.7:1 · 0 similar events

Stage 7 = Trigger — "방류 / 발사" (Threshold crossing → 방향 폭발)
"Captain F-028 — 댐 압력이 임계치 넘어 한쪽 방향으로 터지는 순간"
측정 방법: Today CAT vote ratio × 32y similar events catalog (444) · Today specific

정의 — 방류 / 발사: 댐 압력이 임계치 넘어 한쪽 방향으로 터지는 순간 (Captain F-028)

수계도 비유: 댐 수위(Stage 6)가 압력 임계치를 넘으면 방류 — 시장 한쪽 방향으로 흐름 폭발. 5 axis 중 3/5 동시 충족 → trigger.
Today CAT vote ratio (Bull:Bear) → 32y에서 같은 ratio 발생한 events 매칭 → per-event forward outcome 실측 (F-AI-001 per-event 정합).

Today Trigger Snapshot

Bull / Neutral / Bear count	13 / 16 / 20
Ratio (Bull:Bear)	0.65:1
32y similar events (±15% ratio)	0 / 444

32y Similar Events — Multi-window Forward Outcome (mean across 0 events)

Window	Mean return	Positive %	n
5d	—	—	n=0
15d	—	—	n=0
40d	—	—	n=0
60d	—	—	n=0
100d	—	—	n=0
240d	—	—	n=0

Top 10 Most Similar Events (sorted by ratio closeness to Today)

Date	Ratio	fwd 5d	fwd 60d	fwd 240d

Our Findings:
Today vote ratio 0.7:1 = 32y에서 0회 발생한 pattern. 60d data 부족.

Stage 7 측정 방법 (click to expand)

방류 = 댐 압력 임계치 넘어 한쪽 방향 폭발
• Today CAT vote ratio (Bull:Bear)
• 32y 444 events catalog 매칭 (±15% ratio range)
• Per-event forward outcome 실측 (Multi-window)
• F-AI-001 per-event 정합 (mean 추측 X)

Stage 7 종합 결론 (click to expand)

임계치 crossing 진단
→ Today 0.65:1 ratio → 32y similar events 0개 (unique pattern)
→ "32y 본 적 없는 패턴" — 단일 ratio bucket 부족
→ Stage 12 (HAL) PC 좌표 kNN으로 보완 (cosine similarity 다른 차원)

Source: dashboard_v2_events_41.json (444 events 32y catalog) + master_db Today CAT vote (5/15).

8 Spreading — 유역 전이 (Basin crossover / Metastasis, Multi-window TE) 13 short-mid + 9 mid-long sign-flip

Stage 8 = Spreading — "유역 전이" (Basin crossover / Metastasis)
"Captain F-028 (4/27) — 한 유역(단기)에서 다른 유역(중기/장기)으로 강줄기 넘어감"
측정 방법: Multi-window TE 5 windows × Cross-flip detection (Today specific 흐름 변화)

정의 — 유역 전이 (Basin crossover): 강줄기가 다른 유역으로 넘어가는 순간 (Captain F-028 = Metastasis)

수계도 비유: 한 유역(window short-mid)에서 다른 유역(mid-long)으로 sign-flip되며 흐름 방향 전환 — 단기 → 중기 → 장기 propagation chain (Basin crossover).

Multi-window TE × per-CAT cross-flip detect

Multi-window TE (w10/w20/w40/w100/w250d) per-CAT — short→mid→long propagation chain.
Cross-flip = window 간 sign 전환 (단기 → 중기 또는 중기 → 장기 propagation 진행).
Emerging = w20 Top 10에 신규 진입 / Fading = w20 Top 10에서 사라짐.

Multi-window TE Today (Atlas SSOT)

Window	System TE	z-score	Status
w10d	0.00	+0.000σ	BROKEN_NPZ_all_zero
w20d	1545.35	-2.116σ	OK
w40d	1124.26	-2.342σ	OK
w100d	668.08	-0.783σ	OK
w250d	368.56	+3.851σ	OK

Broken windows: [10] (all-zero, NPZ regen 의무). w10d NPZ all-zero (broken) — short_mid 비교 w20 사용. w10 재생성 요청 의무.

Cross-flip Detection — 단기/중기/장기 유역 전이

Short-mid (단기 w20 ↔ 중기 w40)

Sign-flip CATs (5개) — 단기/중기 방향 반대:

C0 (w20 +9.5 → w40 -12.1)
C1 (w20 -1.1 → w40 +0.8)
C10 (w20 +9.8 → w40 -11.6)
C12 (w20 -5.4 → w40 +3.6)
C17 (w20 -4.7 → w40 +6.8)

Emerging in w20 Top 10:
C10, C34

Fading from w20 Top 10:
C18, C11

Mid-long (중기 w40 ↔ 장기 w250)

Sign-flip CATs (5개) — 중기/장기 방향 반대:

C3 (w40 -8.7 → w250 +0.4)
C4 (w40 -4.3 → w250 +1.5)
C15 (w40 -8.5 → w250 +0.3)
C24 (w40 +5.5 → w250 -3.2)
C25 (w40 +2.9 → w250 -1.2)

Emerging in w40 Top 10:
C31, C9, C48, C32

Fading from w40 Top 10:
C17, C34, C12, C36

Our Findings:
Short-mid sign-flip 13개 / Mid-long 9개 — 단기 → 중기 propagation 진행 (regime transition). Emerging short-mid 2 / Mid-long 4.

Stage 8 측정 방법 (click to expand)

[1] Multi-window TE z-score (동일 신호 다른 window)
  w20 / w40 / w100 / w250 — 어느 stage에 강한가
  Today: w20 -2.12σ / w40 -2.34σ / w100 -0.78σ / w250 +3.85σ

[2] Cross-flip CATs (window 간 방향 반대)
  Short-mid (w20 ↔ w40) sign-flip 5개
  Mid-long (w40 ↔ w250) sign-flip 5개

[3] Emerging / Fading CATs (Top 10 진입/사라짐)
  w20 Top 10 신규 진입 / 사라짐
  w40 Top 10 신규 진입 / 사라짐

Stage 8 종합 결론 (click to expand)

유역 전이 진단 (Today specific)
[1] Multi-window 흐름 동시 변화 — 단기 약 + 중기 약 + 매우 장기 강 (+3.85σ)
[2] 10개 CAT 동시 sign-flip — short-mid 5 + mid-long 5 = 광범위 transition
[3] 매우 장기 (w250) 강한 결속 = 32년 본질 추세 강 (Bull regime active)

그래서 → "단기 transition (재정렬) + 장기 본질 유지"
= V-bounce 회복 가능성 + 단기 변동성 risk
= Stage 12 HAL kNN 2017 Trump rally 후속 패턴 정합

Source: rolling_te_E294_cat49_w{10,20,40,100,250}d.npz (w10 broken, fallback w20)

9 Coherence Dynamic — 가족 영화 / 강줄기 시계열 (252d rolling × 389 frames) 389 frames · 5 stable · 5 volatile

Stage 9 = Coherence Dynamic — "가족 영화" (32y MOVIE)
"Captain V1 (4/28 "Engine Key" 발견) — Stage 3 사진의 시간 시리즈"
측정 방법: 252d rolling × 21d step × 388 frames · 32y 통계 (시대 무관 식별)

정의 — 가족 영화 / 강줄기 시계열: 32년 강줄기 가족이 어떻게 진화했나

수계도 + Family 비유: Stage 3(가족 사진)의 시간 시리즈 — 252d × 21d step × 388 frames. 어느 가족 강줄기가 시대 무관 안정(Stable), 어느 가족이 시대별 변동 큰(Volatile), 어디서 시대 분기점(Era 진화).

252d rolling × 21d step × 389 sample frames (Atlas full timeline). per-CAT 32y rolling coherence — mean + std → "강한 family ≠ 안정 family" 새 차원.
Era 분기점 자동 발견 (사람 라벨 X). pre_2015 frame 251 / post_2015 252 / 2025 372.

Stable Top 5 Family Timeline (32y 시대 무관 안정 가족, 252d rolling)

Captain B option: specific 가족 강줄기가 32년 동안 어떻게 흘렀나 시각화. Stable family = 시대 변동 작은 가족.

Volatile Top 5 Family Timeline (시대 변동 큰 가족)

Volatile family = 시대 분기점 (2015 QE, 2025) 전후 가족 구조 큰 변화. log scale (큰 outlier 영향).

Stable Top 5 (low std, 시대 무관 안정)

CAT	32y std	Family
C31	1.193	Copper family (COPPER + Copper ratios)
C41	1.218	JPY Carry Trade (AUDJPY/EURJPY/NZDJPY)
C46	1.238	G3 Forex (EURUSD/AUDUSD/GBPUSD)
C14	1.275	Asia Pacific (HANG_SENG / NIKKEI)
C44	1.311	Treasury Yields raw (DGS2/5/10/30 + TYX)

Volatile Top 5 (high std, 시대 변동 큰)

CAT	32y std	Family
C47	14099.99	Volatility complex (VIX/VVIX/MOVE/IG_OAS)
C10	2703.08	Fed RRP Liquidity (singleton)
C29	634.23	Frontier Tech (ARKG/HACK/SKYY/SOCL + GEX)
C6	450.31	2Y Note Positioning (singleton)
C23	283.98	SPX MAs + AAII Bull + Gold/VIX

시대 분기점 Top 5 (PRE 2015 → POST 2015, 자동 발견)

CAT	PRE 2015	POST 2015	Δ (시대 강화/약화)	Family
C47	6.767	-7517.232	-7523.999	Volatility complex (VIX/VVIX/MOVE/IG_OAS)
C10	4.490	2129.506	+2125.017	Fed RRP Liquidity (singleton)
C29	-1.020	321.419	+322.439	Frontier Tech (ARKG/HACK/SKYY/SOCL + GEX)
C26	-2.751	122.171	+124.921	Cyclical Industries (Banks/Insurance/Homebuilders)
C23	-77.654	13.331	+90.984	SPX MAs + AAII Bull + Gold/VIX

Our Findings:
32y 388 frame rolling coherence — 5 stable family (시대 무관) + 5 volatile (시대별 변동). Era 분기점 5개 CAT (PRE/POST 2015 Δ 큰 family). 2015 QE 전후 시장 구조 변화 정량 검증.

Stage 9 측정 방법 (click to expand)

[1] Stable Top 5 (low std, 시대 무관 안정)
  "32년 내내 한 가족" — 시장 본질 부분

[2] Volatile Top 5 (high std, 시대 변동 큰)
  "시대마다 다른 가족" — 시장 진화의 증거

[3] 시대 분기점 Top 5 (PRE 2015 vs POST 2015 Δ 큼)
  "2015 QE 전후 시장 구조 변화" data 자체 발견 (사람 라벨 X)

Stage 9 종합 결론 (click to expand)

왜 9번이 중요한가 (Captain 4/28 "Engine Key" 평가)
Stage 3 (가족 사진) + Stage 9 (가족 영화) = Engine Key 후보
= "강한 가족 ≠ 안정 가족" 새 차원 (mean + std 결합)

예: C36 Commodity/Energy
  Stage 3 평균 0.988 (강한 결속) + Stage 9 std 0.093 (안정)
  → "32년 내내 강하고 안정"

vs C0 Energy + Inflation
  Stage 3 평균 0.110 (약) + Stage 9 PRE -0.201 → POST +0.295 (Δ +0.496)
  → "시대마다 다르게 firing" (2015 QE 전후 분기)

Wall Street Tier 1 진정한 차별화 — 32y rolling coherence + 시대 분기점 자체 발견

Source: multimodal v4 252d × 21d step (Atlas P1-C+ full timeline) · 389 frames × 252d window × 21d step

10 Divergence — 표면 흐름 vs 심층 흐름 (Surface CAT/L4 vs Internal Neuron/ASI) Δ +0.121

Stage 10 = Divergence — "표면 vs 심층 흐름"
"강 표면(CAT/L4) 방향과 심층 수류(Neuron/ASI) 방향 분기 측정"
측정 방법: L4 verdict (Surface) vs L0 Neuron mean signed z (Internal) · Today specific

정의 — 표면 흐름 vs 심층 흐름: 강 표면(CAT) 방향과 심층 수류(Neuron) 방향 분기

수계도 비유: 강 표면(CAT/L4 verdict) 방향과 심층 수류(Neuron/ASI) 방향이 일치하는가, 분기하는가? Surface-Internal divergence high → regime shift signal.

Surface (CAT-level vote, L4 verdict) vs Internal (Neuron-level ASI = mean signed z 294 axon).
Inner Acceleration = L1 (CAT) - L0 (Neuron) gap → 상위/하위 layer 가속 차이.
Divergence percentile high → regime shift signal (Captain V1 4/28: 95.9% percentile + 1개월 전 Big Event).

Today Divergence Breakdown (Atlas SSOT, 5/15)

L4 Verdict (Surface, CAT/5DWA)	WEAK_BULL
L0 Neuron (Internal, 294 mean signed z) — Atlas SSOT	-0.1529
L1 CAT-level (Phase 5) — Atlas SSOT	-0.0323
Divergence (\|L0 − L1\| 또는 L4 vs L0) — Atlas SSOT	+0.1206
Inner Acceleration (L1 − L0)	+0.1206

분기 진단

Surface Bull (CAT/L4) + Internal Bear (Neuron L0) — Maturity Pullback signal

Our Findings:
(1) Divergence magnitude: |+0.1206| (HIGH ≥0.10, regime shift 진행).
(2) Surface vs Internal: L4 verdict = WEAK_BULL (CAT 표면) / Neuron L0 (Internal, Atlas SSOT) = -0.1529.
(3) 분기 방향: Surface Bull + Internal Bear → Maturity Pullback signal (표면 강세 + 내부 약화, regime shift 진행).
(4) Inner Acceleration: L1 (CAT-level) − L0 (Neuron) = +0.1206 → 상위 (CAT) 가속 강함, 하위 (Neuron) 둔화 → 표면 momentum 일시 유지 BUT 내부 약화.
(5) 해석: Captain V1 (4/28: 95.9% percentile + 1개월 전 Big Event 사례) 정합 — 표면 (CAT/L4) + 내부 (Neuron L0) 분기 시 regime shift 진행. Stage 12 HAL 32y kNN과 cross-check 의무.

Stage 10 측정 방법 (click to expand)

Surface (CAT/L4 verdict) vs Internal (Neuron/ASI)
• Surface = L4 verdict + CAT-level mean (표면)
• Internal = 294 Neuron mean signed z (심층)
• Divergence = |Surface − Internal| (gap magnitude)
• Inner Acceleration = L1 (CAT) − L0 (Neuron) gap

Stage 10 종합 결론 (click to expand)

표면 흐름 vs 심층 흐름 분기 진단
→ Today L4 WEAK_BULL + L0 Neuron -0.153 → Divergence +0.121 HIGH
→ Surface Bull (CAT/L4) + Internal Bear (Neuron L0) = Maturity Pullback signal
→ Captain V1 (4/28: 95.9% percentile + 1개월 전 Big Event) 정합
→ "표면 강세 + 내부 약화 = regime shift 진행" 정량 검증

Source: Atlas SSOT stage_10_divergence (l4_verdict / phase5_l0_aggregate / level_1_cat / divergence, 5/15 fresh).

Stages 11-12 — 최종 통합 결론

11 HAFS — 본류 수위 (History Analog Finder Score, Daily 결론 scalar) +0.85 · 100th pct

HAFS Scalar

          +0.85σ
        

Status × Direction · signed

32y Percentile

          100th
        

32Y RAREST ZONE

So what?

★ Bull mature peak — mean reversion 압력 강함

32y 8,385일 중 약 8,385일보다 높음 → 단기 mean reversion risk 동시 존재 (Stage 6 압력솥 점검 의무) · 추세 지속 잠재력도 동등.

정의 — 본류 수위: 모든 강줄기가 합쳐진 최종 흐름의 수위 + 방향 (L3 결론 scalar)

수계도 비유: Stage 1-10이 모든 강줄기 / 호수 / 댐 / 방류를 측정 → Stage 11 = "본류 최종 수위 + 흐름 방향 1개 숫자". 32년 강줄기 historical 분포 안에서 오늘 본류 수위가 어디에 있나.

HAFS = Status × Direction (Stage 11 = L3 SOMA Daily 종합 결론 scalar, per-neuron 284 neurons)
• Status = 0.342 × L2_z + 0.435 × Mahal_z + 0.223 × PC1_z (Method C, 통계 가중)
• Direction = S6 sign(Bull pressure − Bear pressure) + S7 (3+ axes Bull/Bear vote)
• HAFS = Status × Direction (+/−, signed magnitude)

Today HAFS Breakdown (Atlas SSOT, 284 daily neurons)

HAFS scalar (Atlas SSOT Status × Direction)	+0.8518σ
Status mean (activation magnitude)	1.1078
Direction net (signed)	+0.6694
Direction 분포 (dashboard_v2 fallback)	+127 / 112 / −45
32y percentile (Today HAFS rank)	100.0th

32y Percentile (Today HAFS rank) 해석 — Today 100.0th (클릭하여 펼치기)

계산 방식: Today HAFS scalar 값이 32y (8,385 days) 전체 daily HAFS mean 분포에서 몇 % 의 days보다 높은가?

해석:
  • 50th percentile = 32y의 절반 (중간 위치)
  • 75th+ = Today가 32y의 75% days보다 높음 (상위 25%, Bull side mid-high)
  • 25th− = Today가 32y의 25%만 보다 높음 (하위 75%, Bear side mid-low)
  • 90th+ = Bull extreme (32y rare zone, mean reversion risk)
  • 10th− = Bear extreme (32y rare zone, strong Bear regime)

Today (100.0th): "Today HAFS가 32y 8,385일 중 약 8,385일보다 높음" = Bull extreme — 32y rare zone, mean reversion 압력

활용: Stage 11 결론 scalar의 32y 위치 → "지금이 32y에서 얼마나 강세인가" 정량 진단. 상위/하위 quartile 진입 시 = rare zone signal (mean reversion risk or strong regime persistence).

Top 10 Bull-side / Top 10 Bear-side Neurons (HAFS = Status × Direction, Atlas SSOT + Top 10 streak)

HAFS = Status × Direction 해석 가이드 — Stage 1 (raw z) vs Stage 11 (signed scalar) 차이 (클릭하여 펼치기)

기본 개념:
  • Status = activation magnitude (|z| 강도, 0~∞, sign 없음)
  • Direction = signed regime context (+1 Bull / 0 neutral / −1 Bear)
  • HAFS = Status × Direction → signed scalar (+/−, magnitude × sign)

Stage 1 (raw HAFS+) vs Stage 11 (HAFS):
  Stage 1 Top 10 BULL = z > 0 raw firing neurons (예: NFP +9.45σ, RT_SPSM_SPY +3.62σ)
  Stage 11 Top 10 BULL = HAFS > 0 = signed magnitude positive (regime-contextualized)

Credit spread / OAS가 Top 10 BULL에 보이는 이유:
  Credit spread/OAS는 통상 Bear signal (높을수록 risk-off).
  예: SP_B_BB raw z = +X (현재 spread 확대 = Bear regime contribution)
  BUT Stage 11 HAFS = Status × Direction → magnitude 크면 (regime contribution 강함) → Top BULL/BEAR 어느 한쪽 list에 등장.
  = "Bear regime의 강한 contribution"이 signed magnitude로 표현됨.

활용:
  • Stage 1 = 어떤 neurons이 현재 firing 시작 (단순 z magnitude)
  • Stage 11 = 어떤 neurons이 regime context 안에서 강한 contribution (signed scalar)
  • Stage 11 list에 Credit/OAS dominant = Credit regime contribution 강력 (Bull-Bear regime 어느 쪽이든 영향력 큼).

Top 10 BULL (today firing)

Neuron	HAFS	L2 z	Top 10 streak
SP_B_BB	+19.84σ	19.84	NEW
B_OAS	+11.94σ	11.94	NEW
SLYV_SLP50	+10.36σ	10.36	NEW
SKYY	+7.92σ	7.92	NEW
SP_CCC_BB	+5.57σ	5.57	NEW
RT_IWC_SPY	+2.80σ	2.80	NEW
RT_VTI_ILF	+2.53σ	2.53	NEW
CRAK	+2.45σ	2.45	NEW
CHFJPY	+2.40σ	2.40	NEW
XLRE	+2.36σ	2.36	NEW

Top 10 BEAR (today firing)

Neuron	HAFS	L2 z	Top 10 streak
EURUSD	-1.62σ	-1.62	NEW
DXY	-1.56σ	-1.56	NEW
EURGBP	-1.52σ	-1.52	NEW
HANG_SENG	-1.39σ	-1.39	NEW
XLY	-1.31σ	-1.31	NEW
GBPUSD	-1.29σ	-1.29	NEW
US_HY_YIELD	-1.21σ	-1.21	NEW
ENERGY_HY_OAS	-1.20σ	-1.20	NEW
RT_XLY_XLP	-1.17σ	-1.17	NEW
DAX	-1.09σ	-1.09	NEW

Per-neuron HAFS Distribution (284 neurons)

Bull-side (HAFS > 0)	127 / 284 (45%)
Bear-side (HAFS < 0)	45 / 284 (16%)
Quiet (HAFS ≈ 0)	112 / 284

32y Daily HAFS Timeline + Today percentile (Atlas SSOT mean_hafs series)

파란 선 = 32y daily HAFS mean (monthly subsample). 큰 다이아몬드 = Today position. Today HAFS +0.8518σ → 32y percentile 100.0th (32y 8,385일 중 약 8,385일보다 높음). 빨강 가로선 = Today HAFS level (그래프 어디 위치인지 표시).

구분	★ Thesis	Detail	Remark
Position 32y rank	★ 32y 최상위 zone (상위 5%) — rarest extreme	HAFS +0.8518σ · 32y percentile 100.0th · 8,385일 중 8,385일보다 높음	✓ CRITICAL
Component Status × Direction	★ Direction Bull lean 강함 + 강한 magnitude	Direction net +0.6694 · Status mean (activation) 1.1078 · 분포: +127 Bull / 112 / −45 Bear	✓ SIGNED
Per-neuron 284 dist.	★ Bull dominance 강함 (Bull/Bear ratio > 1.5x)	127 Bull (HAFS > 0, 45%) / 45 Bear (HAFS < 0, 16%) / 112 Quiet	✓ DOMINANCE
So What? Forward action	★ 단기 mean reversion risk + Stage 6 압력솥 점검 의무	32y rare zone Bull extreme → Stage 12 HAL kNN cross-check 의무 (multi-window forward + Path A/B/C 확률). Captain V1 직관 정합: Bull-asymmetry (39% Bull / 4.7% Bear / 56% mid) — extreme은 statistical rarity.	✓ ACTION

Source: multimodal v4 7dim 5/15 + 32y baseline (Atlas refresh) · Status × Direction proxy (Mahalanobis/PC1 추정, P1 정밀화 의무)
32y timeline: l3_daily_db.json (monthly subsample, Atlas refresh).

12 HAL — 32년 강줄기 비교 (Historical Analog Locator, forecast 결론) 60d mean +3.9% · Path A 70% / B 0%

32Y Cluster Located

★ 2017 (10개)

PC 좌표: +1.90 / +0.04 / -0.79 → Top 10 analog days (cosine similarity)

So what? — 그 후 어떻게 흘렀나

★ 단기 횡보 → 중장기 강세 가속 (240d +18.5% / pos 100%)

60d mean +3.9% (pos 100%) · 240d mean +18.5% (pos 100%, 10/10 analogs)

Multi-Window Path Probability — kNN Top 10 analog distribution

Window	PATH A · RALLY (Rally up)	PATH B · PULLBACK (Pullback down)	PATH C · RANGE (Sideways)	Threshold
15D 단기 (3주)	10% 1 / 10	70% 7 / 10	20% 2 / 10	±3%
40D 중기 (2개월)	40% 4 / 10	30% 3 / 10	30% 3 / 10	±5%
60D 중기 (3개월) · primary	10% 1 / 10	70% 7 / 10	20% 2 / 10	±5%

Window 흐름 진단: 단기→장기 모두 Pullback 우세 → 조정 패턴 일관 지속.

정의 — 32년 강줄기 비교: 오늘 강줄기 패턴과 가장 비슷한 32y 일자 + 그 후 어떻게 흘렀나

수계도 비유: 본 instance의 32y 강줄기 historical map에서 "오늘과 가장 비슷한 10개 일자" 찾기 → 그 일자들의 forward outcome 실측 = forecast 결론.
Today PC 3D 좌표 + kNN Top 10 historical analog days (cosine similarity) + 각 analog의 forward outcome 실측 (multi-window) + Path A/B/C probability (data-driven, F-AI-001 per-analog 정합).

Today PC 좌표 (2026-04-24)

  PC1 = +1.90  ↑ Bull state magnitude 강함

  PC2 = +0.04  → 보조 차원 neutral

  PC3 = -0.79  ↓ tertiary mode 음수 (특정 패턴 활성)

= "강한 Bull magnitude + 특정 sub-pattern" 좌표
= 32y space에서 specific zone (다른 시기 거의 안 나타나는 좌표)

kNN Top 10 Historical Analogs · Cluster: 2017 (10개)

SSOT cosine similarity (Atlas refresh) + multi-window forward (dashboard_v2 5/15/40/60/100/240d merge)

Date	Cosine sim	fwd 5d	fwd 15d	fwd 40d	fwd 60d	fwd 100d	fwd 240d
2017-04-04	0.416	-0.2%	+1.2%	+3.4%	+3.0%	+4.4%	+17.1%
2017-03-23	0.406	+1.0%	-0.7%	+1.8%	+5.1%	+5.9%	+18.3%
2017-03-24	0.406	+0.8%	+0.3%	+2.4%	+4.4%	+6.1%	+19.0%
2017-03-22	0.405	+0.5%	-0.1%	+1.1%	+4.1%	+5.8%	+18.2%
2017-03-27	0.405	+0.7%	+0.1%	+2.8%	+4.5%	+4.6%	+21.2%
2017-03-28	0.404	+0.1%	-0.8%	+2.2%	+3.7%	+3.6%	+20.1%
2017-03-21	0.403	+0.7%	+0.6%	+0.9%	+4.3%	+5.0%	+18.2%
2017-03-29	0.402	-0.3%	-0.1%	+2.6%	+3.7%	+3.6%	+19.2%
2017-03-20	0.400	-1.3%	-0.6%	+1.4%	+3.1%	+3.5%	+15.3%
2017-03-31	0.399	-0.2%	+0.6%	+2.4%	+2.9%	+4.2%	+18.4%

Multi-window Forward 통계 (mean ± range, kNN Top 10 종합)

Window	Mean	Median	Min ~ Max	Pos %	n
5d	+0.17%	+0.31%	-1.3% ~ +1.0%	60%	10
15d	+0.06%	+0.02%	-0.8% ~ +1.2%	50%	10
40d	+2.10%	+2.33%	+0.9% ~ +3.4%	100%	10
60d	+3.88%	+3.90%	+2.9% ~ +5.0%	100%	10
100d	+4.66%	+4.46%	+3.5% ~ +6.1%	100%	10
240d	+18.51%	+18.36%	+15.3% ~ +21.2%	100%	10

Multi-window Forward 분포 (kNN Top 10, mean ± range, ◆ median)

Path Probability — Data-driven (forward 60d 분포 기반)

Path A (Rally, 60d > +5%)	70%	0/10
Path C (Range, −5% ≤ 60d ≤ +5%)	30%	0/10
Path B (Pullback, 60d < −5%)	0%	0/10
Mean forward 60d	+3.88%	10 analogs
Mean forward 240d	+18.51%	10 analogs

6중 검증 종합 진단 (Stage 6 + 7 + 8 + 9 + 10 + 12)

• Stage 6 (Pressure): per-neuron extreme 5개 detect — Leverage / Credit 3 · Yield Curve 2
• Stage 7 (Trigger): Today 0.7:1 ratio = 32y 0 similar events, forward 60d mean —
• Stage 8 (Spreading): Atlas P2-C cross-flip detect (Captain V1 baseline: CAT 3 DIX 2배)
• Stage 9 (MOVIE): Atlas P1-C 252d rolling (4/28 baseline: Stable C43/C36, Volatile C12/C14, Era PRE→POST 2015 C8 Δ+0.496)
• Stage 10 (Divergence): Surface Bull (CAT/L4) + Internal Bear (Neuron L0) — Maturity Pullback signal (Δ +0.121, Inner Accel -0.032)
• Stage 12 (HAL): kNN Top 10 2017 (10개), Path A 70% / B 0% / C 30%, 60d mean +3.88%

구분	★ Thesis	Detail	Remark
Location 32y PC + kNN	★ Cluster "2017 (10개)" narrow — 강한 historical analog 신호	Today PC: +1.90 / +0.04 / -0.79 · "강한 Bull magnitude + 특정 sub-pattern" 좌표	✓ NARROW
Forward Path 6 windows × 10 analogs	★ 단기 횡보 → 중장기 강세 가속 패턴	5d +0.2% (pos 60%) · 40d +2.1% (pos 100%) · 60d +3.9% (pos 100%) · 100d +4.7% (pos 100%) · 240d +18.5% (pos 100%)	✓ HIGH CONF
Path 확률 60d 분포	★ Path A (Rally) 우세	A 70% (Rally >+5%) · B 0% (Pullback <−5%) · C 30% (Range ±5%) — 10 analog 분포	✓ DATA-DRIVEN
So What? Captain V1 직관	★ data-driven forecast 신뢰도 HIGH — Stage 1-12 일관	UV verdict: Maturity Pullback → V-bounce 회복 — 32y Multi-window forward corroborate. Stage 1-10 (Source / Coherence / Path / Pressure / Spreading / Divergence) → Stage 11 HAFS → Stage 12 HAL 모두 일관.	✓ CONSISTENT

Source: Atlas SSOT stage_12_hal (knn_forward_multiwindow + multiwindow_stats, 5/16 fresh, 공문 #145) · dashboard_v2_pcs_all.json (32y PC trajectory).