Three machine layers that turn scattered data into edge. The Market Web shows how signals connect — tap a node to trace cause and effect. Forward View forecasts every driver behind your portfolio. Earth Signal reads physical activity from space and maps it to the SET names exposed to it.
LIVE PIPELINE · WHAT'S FLOWING INTO THE MARKET NOWas of 2026-06-24 · 2026-06-18
Each row is a real signal read from live data, translated into what it means for your money and the SET names it touches. The graph below is the full web of connections behind it.
Fed Funds Rate3.63% · 10y–2y +0.29HEADWIND
Fed at 3.63% — higher-for-longer US rates keep the dollar firm, which pressures the baht and risks foreign outflows from the SET. Banks earn wider margins either way.
LANDS ON →SETKBANKSCBBBL
Oil (WTI)$73 ▼ 0.5%TAILWIND
Oil $73 easing — Thailand imports its energy, so cheaper oil is a tailwind for transport, airlines and consumers. Energy producers move the other way.
LANDS ON →AOTMINTPTTPTTEP
SET ValuationP/E 22.3 vs 17.89 avgWATCH
SET trades at 22.3× — above its 17.89 long-run average. Less margin of safety; favour quality and patience over chasing.
LANDS ON →SET
Volatility & MoodVIX 16.4 · F&G 55NEUTRAL
Volatility is contained — no panic to fade right now.
LANDS ON →GoldSET
AI Capex BoomNASDAQ ▼ 2.1%WATCH
The hyperscaler spending super-cycle is the demand behind chips — it flows straight to Thailand's semiconductor names. When US tech runs, DELTA tends to follow; when it stumbles, watch for sympathy moves.
LANDS ON →DELTA
Gold$4,142 ▼ 0.5%NEUTRAL
Gold is the hedge that only helps if bought before the storm. Its bid signals de-dollarisation and real-rate fear — a check on how defensive the moment really is.
Readings computed from live caches (daily-brief + market-mood). The causal map is curated trader knowledge — not investment advice.
TREND TAPE· ranked by expected return
2026-06-04
ASSETSPOTEXP RETVOLMODEL
QQQ
US tech · Nasdaq-100
$716.38
+5.1%
18%
vol-cone · 90d
SOXX
Semiconductors
$625.2
+9.3%
40%
vol-cone · 90d
VNM
Vietnam equity
$18.16
+5.4%
27%
vol-cone · 90d
GLD
Gold
$369.46
+3.4%
28%
vol-cone · 90d
USD/THB
Baht vs dollar
33.45
+0.6%
8%
vol-cone · 90d
MCHI
China equity
$50.78
-1.3%
20%
vol-cone · 90d
SCB
SCB
฿141
-7.2%
18%
Kronos-mini · 20d
PTT
PTT
฿36.75
-9.0%
19%
Kronos-mini · 20d
BBL
Bangkok Bank
฿174.5
-8.7%
18%
Kronos-mini · 20d
ADVANC
Advanced Info
฿360
-10.0%
21%
Kronos-mini · 20d
KBANK
Kasikorn Bank
฿203
-11.9%
19%
Kronos-mini · 20d
SET
Thai SET
฿1,588.06
-10.3%
15%
Kronos-mini · 20d
Illustrative forecasts. Walk-forward calibration on ^SET.BK: APPROXIMATELY CALIBRATED — 80% band coverage 74% (target 80%), direction 51%, skill vs naive -11.9%. Kronos-mini is zero-shot on SET — not validated until walk-forward passes. Expected ranges, not promises.
· Walk-forward backtest, not live tracking — no guarantee of future calibration.
FORWARD CONE (ILLUSTRATIVE)· 90-day cone for every portfolio driver
vol-cone (drift+σ baseline) · 2026-06-26
^SET.BK
Thai SET
+6.5%
vol 18%
QQQ
US tech · Nasdaq-100
+5.1%
vol 18%
ACWI
Global all-world
+3.8%
vol 14%
URTH
Global brands / world
+3.4%
vol 13%
EWJ
Japan equity
+5.1%
vol 21%
SOXX
Semiconductors
+9.3%
vol 40%
VNM
Vietnam equity
+5.4%
vol 27%
GLD
Gold
+3.4%
vol 28%
BITQ
Blockchain equities
+4.4%
vol 57%
THB=X
USD/THB
+0.6%
vol 8%
MCHI
China equity
-1.3%
vol 20%
INDA
India equity
-2.1%
vol 15%
Each cone is a 90-day forward distribution for the index that drives a Dream-Portfolio holding — median path + 10/90 band. Baseline is a transparent drift+volatility cone; run ingestion/forward_view.py in the M3 TimesFM venv to swap in the learned model. Expected ranges, not promises — not investment advice.
FORECAST CALIBRATION^SET.BK · 120 windows · 2026-06-04
APPROXIMATELY CALIBRATEDtarget 80% | actual
80% BAND COVERAGE — amber line = target
74%
COVERAGE
74%
80% target
DIRECTION
51%
50% baseline
SKILL VS NAIVE
-11.9%
vs yesterday
LAST 20 WINDOWS — ● inside 80% band, ○ outside
· Walk-forward backtest, not live tracking — no guarantee of future calibration.
· ARIMA fitted on returns, compounded for terminal price bands.
· Model is refitted each window — real-time performance may differ.
· Small positive skill vs naive (yesterday's close) is expected; large skill claims should be suspect.
Statistical Decoupling. The Kelly sizing and Probability of Profit below are computed from the zero-shot expected return and volatility forecast. Since this model is not fine-tuned on SET-specific regulations, treat allocations as risk ceilings.
SET
SET Index
฿1,588.06
20d Drift-10.3%
Ann. Vol15.4%
HOVER TO INSPECT TRAJECTORY
REDUCE / SHORTPoP: 1%
Negative bias detected. Tighten stops or allocate to cash hedges.
REC. STOP (1D SD)
฿1568.28
KELLY STAKE
0.0%
PTT
PTT
฿36.75
20d Drift-9.0%
Ann. Vol18.9%
HOVER TO INSPECT TRAJECTORY
REDUCE / SHORTPoP: 5%
Negative bias detected. Tighten stops or allocate to cash hedges.
REC. STOP (1D SD)
฿36.19
KELLY STAKE
0.0%
KBANK
Kasikorn Bank
฿203
20d Drift-11.9%
Ann. Vol19.2%
HOVER TO INSPECT TRAJECTORY
REDUCE / SHORTPoP: 1%
Negative bias detected. Tighten stops or allocate to cash hedges.
REC. STOP (1D SD)
฿199.86
KELLY STAKE
0.0%
ADVANC
Advanced Info
฿360
20d Drift-10.0%
Ann. Vol20.6%
HOVER TO INSPECT TRAJECTORY
REDUCE / SHORTPoP: 4%
Negative bias detected. Tighten stops or allocate to cash hedges.
REC. STOP (1D SD)
฿354.02
KELLY STAKE
0.0%
SCB
SCB
฿141
20d Drift-7.2%
Ann. Vol18.5%
HOVER TO INSPECT TRAJECTORY
REDUCE / SHORTPoP: 8%
Negative bias detected. Tighten stops or allocate to cash hedges.
REC. STOP (1D SD)
฿138.90
KELLY STAKE
0.0%
BBL
Bangkok Bank
฿174.5
20d Drift-8.7%
Ann. Vol18.3%
HOVER TO INSPECT TRAJECTORY
REDUCE / SHORTPoP: 5%
Negative bias detected. Tighten stops or allocate to cash hedges.
REC. STOP (1D SD)
฿171.93
KELLY STAKE
0.0%
MATHEMATICAL SPECIFICATIONS
· Zero-shot transfer from global pretrained weights — SET-specific dynamics not learned.
· 3 sample paths averaged — reduces noise but not systematic bias.
· Not validated against SET forward returns yet. Treat as illustrative.
· Bands are vol-cone (95% CI), not model uncertainty — they frame the range, not confidence.
· Expected Returns & Volatility calculated over 20-day path. Continuous-time Kelly Criterion sizing assumes log-normal returns.
EARTH SIGNAL· AlphaEarth Foundations (DeepMind, 2025) → SET tickers
○ AWAITING EARTH ENGINE PULL
64-dimensional per-pixel embeddings fuse optical, radar, LiDAR and climate. Year-over-year embedding shift over an area = physical change on the ground. Each economic site below maps to SET tickers; the change magnitude is a leading, satellite-derived activity signal — alt-data that precedes the financial print.
Planted area + harvest cycle → sugar / cassava output
→ KSL · BRR · KTIS
Utility-scale solar sitesEnergy
—
New PV-array footprint → renewable capacity additions
→ GULF · GPSC · BGRIM
Dataset GOOGLE/SATELLITE_EMBEDDING/V1/ANNUAL. To populate the change metrics: pip install earthengine-api, earthengine authenticate, then run ingestion/alphaearth.py on the M3. No numbers are shown until real embeddings are pulled — never fabricated.