Parliament — this week's report, explained

run #154Forecast updated 20 Oct 2026

Every number on this chamber's weekly report card, walked through in everyday words. Nothing technical without a translation.

The one-sentence version

If the election were held this week, Perikatan Nasional would most likely win the most seats — around 76. That is still short of the 112 seats needed to govern alone. The single most likely outcome right now is a hung chamber: a 100% chance that no bloc reaches the majority mark and partners have to be negotiated after the election.

What this report actually is

Think of it as a weather forecast for politics. The model replayed this election 10,000 times, nudging the real inputs each time — the last election result, how those voters should have shifted by now, opinion polls, and years of election history. Each replay is one possible election night.

The report is just a headcount across those 10,000 replays. When it says a bloc "leads with 76 seats", it means 76 was the most common outcome in the pile — not a promise, and not a guess.

Reading the card, line by line

Pakatan Harapan: 73 seats [66–81] — in the middle replay, Pakatan Harapan ended on 73 seats — and 8 out of 10 replays landed somewhere between 66 and 81. The wider the bracket, the less certain the model is.

Barisan Nasional: 35 seats [28–43] — in the middle replay, Barisan Nasional ended on 35 seats — and 8 out of 10 replays landed somewhere between 28 and 43. The wider the bracket, the less certain the model is.

Gabungan Parti Sarawak: 21 seats [19–24] — in the middle replay, Gabungan Parti Sarawak ended on 21 seats — and 8 out of 10 replays landed somewhere between 19 and 24. The wider the bracket, the less certain the model is.

Gabungan Rakyat Sabah: 7 seats [5–8] — in the middle replay, Gabungan Rakyat Sabah ended on 7 seats — and 8 out of 10 replays landed somewhere between 5 and 8. The wider the bracket, the less certain the model is.

Warisan: 4 seats [2–7] — in the middle replay, Warisan ended on 4 seats — and 8 out of 10 replays landed somewhere between 2 and 7. The wider the bracket, the less certain the model is.

Others / Independents: 5 seats [3–7] — in the middle replay, Others / Independents ended on 5 seats — and 8 out of 10 replays landed somewhere between 3 and 7. The wider the bracket, the less certain the model is.

Perikatan Nasional: 76 seats [69–83] — in the middle replay, Perikatan Nasional ended on 76 seats — and 8 out of 10 replays landed somewhere between 69 and 83. The wider the bracket, the less certain the model is.

"P(no single bloc reaches 112) = 100%" is the chance that no bloc wins a majority — a "hung" chamber where nobody governs alone and coalition partners have to be negotiated after the election.

"P(PH+BN majority) = 35%" is the chance the current government partners win enough seats between them to stay in government.

Who's up, who's down

These are the seats whose chances moved most since the previous weekly run — usually because news broke about them: a candidate announced, a local row, a shift in coverage. Read a move from X% to Y% as: out of 100 replays, this bloc used to win the seat in X of them; now it wins it in Y.

BENTONG — PH ▲ from 54% to 57%
BENTONG — BN ▼ from 35% to 33%
MASJID TANAH — BN ▲ from 14% to 16%
MASJID TANAH — PN ▼ from 86% to 84%
SEPANG — PN ▼ from 37% to 36%

The small print, translated

Volatility σ = 0.085 (horizon-scaled; see methodology §9).

"Volatility σ = 0.085" is how jumpy the model thinks this chamber is. A higher number means seat chances swing further week to week; a lower number means the model sees a steady race.

Last recorded turnout (GE15): 73.4% — differential turnout remains an unmodelled risk (methodology §11).

"Last recorded turnout (GE15): 73.4%" flags the model's blind spot: it cannot know who will actually turn up and vote. Last time, 73.4% did.

Forecast run #154; inputs hashed and immutable.

The note about inputs being "hashed and immutable" is a bookkeeping promise: the exact data that produced this report is archived, so anyone can re-run it later and get the same numbers back.

What this report is NOT

Not a prediction of the future — it is a snapshot of what would happen if the election were held today.
Not an endorsement, and not a view on who should win — it is arithmetic on public data.
Not precise to the seat — every number is a range, so treat close calls as genuinely close.

Full dashboard →Every seat, sorted →How the model works (plain language) →All weekly reports →