The Signal Room · Johor State Election

Know where Johor is moving before the headlines catch up.

Sistem risikan pilihan raya untuk membaca calon, isu dan kebarangkalian kemenangan setiap DUN Johor — dengan bukti sumber, bukan sekadar andaian.

56/56

DUN with live estimates

172

Candidates on file

246

Source receipts

9

Stored articles

Live dataSource receipts on every metric
Scroll — from noise to signal

From noise to signal

Beribu mention tidak bermakna apa-apa tanpa konteks kawasan.

Mandat Johor menapis noise menjadi signal yang boleh diaudit — setiap keluaran membawa resit sumber, skor keyakinan dan status pengesahan.

Official noticesRAW
Wire reportsRAW
Regional desksRAW
Public pagesRAW
RSS feedsRAW
Open datasetsRAW

Verification · Dedupe · Entity link

Candidate SignalSIGNAL

Verified identity + mention quality

Party MomentumSIGNAL

Coalition-level movement, source-weighted

Issue PressureSIGNAL

Narrative clusters mapped to seats

Seat ProbabilitySIGNAL

Estimate with stated uncertainty

Live seat forecast preview

Three seats, three different races.

Constituencies showing how the model reads an urban seat, a rural seat and the closest race. No candidate name is shown without a verified source receipt.

Tiram

Conf · Low

Race rating: Lean · 0.0pp over 7 days

BN33.3%
PH22.2%
PN22.2%

Probability estimate, not prediction guarantee. Interval and model features on the seat dossier.

Probability drivers

  • 012022 baseline: BN won this seat with 40.6% vote share
  • 024-candidate field across 4 parties raises volatility
  • 03No verified news or sentiment signals ingested yet
  • 04Candidate list partially verified (trusted secondary source)
Data mode · Live

Why this seat moved

Every shift comes with its reasons attached.

When a probability moves, the model publishes its drivers, its confidence, and the receipts behind each one. No silent updates, no unexplained jumps — the audit trail is part of the product.

Movement detected

No published movement in the current window

Baseline estimates are live for all tracked seats. Movement is published only when consecutive receipted model runs diverge — the system never animates a number to look busy.

Evidence Vault

No source, no metric. That rule is enforced in the schema.

Every fact-bearing row in the database carries a source receipt: name, tier, URL, retrieval time, confidence, verification status and audit history. These are the latest receipts on file.

SourceTierRetrievedConfidenceStatusReceipt
RTMTIER3 NEWS2 Jul, 03:43 pmConf · MediumUnverified
RTMTIER3 NEWS2 Jul, 03:43 pmConf · MediumUnverified
Utusan MalaysiaTIER3 NEWS2 Jul, 03:43 pmConf · MediumUnverified
RTMTIER3 NEWS2 Jul, 03:20 pmConf · MediumUnverified
RTMTIER3 NEWS2 Jul, 03:20 pmConf · MediumUnverified
RTMTIER3 NEWS2 Jul, 03:20 pmConf · MediumUnverified
Utusan MalaysiaTIER3 NEWS2 Jul, 03:20 pmConf · MediumUnverified
RTMTIER3 NEWS2 Jul, 03:16 pmConf · MediumUnverified
RTMTIER3 NEWS2 Jul, 03:16 pmConf · MediumUnverified
RTMTIER3 NEWS2 Jul, 03:16 pmConf · MediumUnverified
Utusan MalaysiaTIER3 NEWS2 Jul, 03:16 pmConf · MediumUnverified
Utusan MalaysiaTIER3 NEWS2 Jul, 11:51 amConf · MediumUnverified

Methodology

Built to be audited.

01

Estimates, not guarantees

Every probability ships with a confidence interval and a model confidence grade. The system never claims to know a result.

02

Official data wins

SPR and official notices override secondary sources for candidates, seats and results. Cross-checks run on every important fact.

03

Fact ≠ model ≠ rumour

A candidate name is a fact. A probability is a model output. An unverified claim is a rumour — each is labelled and stored separately.

04

Conflicts are shown, not hidden

When sources disagree, the UI says 'Source conflict detected' and the item routes to the Verification Desk for manual review.

The map is the product

Enter the Johor War Room.

Fifty-six seats, one command center. Probabilities, narratives, emerging signals and receipts — in one screen built for decision pressure.